Skip to main content
On this page

​
Coverage by target cloud

​
How the backends compare

Each row compares a capability of SNS with its adaptation on each backend. A dash means this row is not stated for that backend.

​
Max adaptation

​
On Google Cloud

​
Via Google Cloud Pub/Sub

​
How it works

Your application keeps its SNS client and topic ARNs. When it calls Publish, the message body is saved in PostgreSQL, then the deliveries still to be made are recorded: one to the Pub/Sub topic mapped to this SNS topic, and one to each SQS subscription. Publish then returns a MessageId. A successful response means the message is saved and its deliveries are queued. It does not mean that Pub/Sub or any consumer has received the message. A delivery worker in the appliance then publishes the message to the Pub/Sub topic, with the SNS MessageId in the sns.message-id attribute. Pub/Sub delivers it to the subscriptions on that topic. You create and manage those subscriptions; the appliance does not create, pull from or acknowledge them. SQS subscriptions do not go through Pub/Sub: the delivery worker sends to each SQS queue directly.
Publish saves the message and returns a MessageId. A delivery worker then publishes it to the Pub/Sub topic mapped to the SNS topic, and Pub/Sub delivers it to your subscriptions.Publish saves the message and returns a MessageId. A delivery worker then publishes it to the Pub/Sub topic mapped to the SNS topic, and Pub/Sub delivers it to your subscriptions.

Publish returns once the message is saved, before Pub/Sub has received it.

​
Delivery, retries and duplicates

Pending deliveries survive a restart: the delivery worker picks them up again when it starts. If Pub/Sub is unavailable, the worker retries with capped backoff and reports messages that have been waiting a long time to reach Pub/Sub. If a publish to Pub/Sub ends without a clear answer, the message can arrive twice with the same sns.message-id attribute, so consumers must handle duplicates. Each SQS delivery succeeds or fails on its own. Before sending, the appliance checks that the subscription still exists and that the queue’s access policy still allows the topic. After a subscription is deleted, no further deliveries are attempted for it, and a subscription created to replace it does not receive the old one’s pending messages. Once SQS has stored a message, SQS is responsible for it.

​
Limitations

  • FIFO topics are not supported. FIFO ordering, deduplication and MessageGroupId are rejected. Only standard topics work.
  • SNS delivers only to SQS subscriptions. Subscriptions using HTTP, HTTPS, Lambda, email, SMS or any other protocol are rejected.
  • You configure the Pub/Sub subscriptions. Acknowledgement deadlines, retries and dead-letter topics on Pub/Sub subscriptions are set in Pub/Sub, not through SNS.
  • Cross-account callers are not supported. The topic access policy is checked and enforced for supported callers. You can set it inline on the topic or with one aws_sns_topic_policy resource, which is applied with the topic.
  • SNS redrive policies are not supported. A subscription RedrivePolicy is rejected. Pub/Sub’s own redelivery and dead-letter settings are separate, and you configure them.

​
Other considerations

Before going live. Provision the Pub/Sub topic mapped to each SNS topic, and your Pub/Sub subscriptions, before sending production traffic. Confirm in this deployment that messages reach both your SQS queues and your Pub/Sub consumers before you move publishers. What to monitor. Watch how long messages wait to be sent to Pub/Sub, along with your Pub/Sub backlog and consumer health. An empty Pub/Sub backlog does not mean every published message has reached Pub/Sub; some may still be waiting to be sent. Moving from AWS. Messages still pending on AWS SNS are not transferred. They stay on AWS.

​
Example monthly charges

This Tensor9 BYOC scenario estimates service charges for the named origin and target services. Adapter hosting and commercial Tensor9 fees need separate allocations. How Tensor9 estimates costs explains the method and what the estimates include.
SNS Standard · SNS publish billing units
Feature: SNS Standard. Estimated monthly service charges by SNS publish billing units. Adapter hosting and Tensor9 fees are excluded. Source: us-east-1; target: us-central1; currency: USD.
SNS Standard · SNS publish billing units: monthly service charges by SNS publish billing units. Exact values are in the following table.SNS Standard · SNS publish billing units: monthly service charges by SNS publish billing units. Exact values are in the following table.
Comparison pills describe all sampled scale points; amounts beside each service are for the named workload. Costs The Same means within 20% at every point. Costs Depend includes mixed results and unpriced samples. Sampled monthly charges in USD, rounded to cents. Difference means target minus origin; a positive difference means the target costs more. Estimated allocation, not benchmark calibrated: sized from 2 vCPU, 8 GiB memory and 100 GiB SSD at 0 per second to 32 vCPU, 128 GiB memory and 100 GiB SSD at 1,000 per second. From 10,000 per second the workload exceeds the planning range and the target is unpriced. Curve assumptions
  • Fan-out and retry assumptions remain explicit; this is not destination delivery TPS. Other native meters and workload assumptions retain their explicitly displayed default quantities. Workload translation and capacity assumptions are not benchmark measurements.
  • Native monthly usage priced at the cited provider list prices; listed target allowances apply where present, excluding invoice credits, taxes and negotiated discounts. Rates as of 2026-10-22: us-east-1 → us-central1, USD/month.
  • Source prices are reference charges, not an invoice. Target lines cover selected service usage/configuration only; adapter hosting and Tensor9 fees are excluded.
  • Samples of explicit prices are not performance measurements or capacity qualification. No interpolation or break-even guarantee is implied.
  • Variant assumptions show the first axis sample; any axis-dependent configuration is described by this curve’s workload note.
  • Rates are cited provider list prices, editable for discounts, commitments or another region. Workload translation and capacity coefficients remain editable planning assumptions, not benchmark measurements or a commercial quote.
  • Default-zero rates are explicit assumptions, not evidence that the deployed service or resource is free.
  • Provisioned-unit minimums use an editable approximate allocation floor. Modeled target service cost is max(usage/allocation, minimum), with billable network, durable staging and dependent sinks added separately; actual RU/TU sizing, scaling increments and throughput remain unqualified.
  • Durable staging is priced once: provisioned targets include adapter state in their selected allocation; native targets add a separate staging allocation when required. Adapter hosting is a separate allocation and must exclude resources already priced as staging, workers or target capacity.
  • A declared adapter mapping and cost model do not establish implementation availability or service equivalence.
  • This is a declared directed mapping; a planning estimate does not establish deployment or request-serving support.
  • Authored planning assumptions, not measured adapter capacity. Origin rates are cited first-tier list prices, before free allowances and discounts.
  • A cost scenario does not establish backend availability, protocol fidelity or production readiness. Registry serving status remains independent.
  • Native Pub/Sub and Service Bus backends also retain durable SNS staging. Price that allocation once; database targets already host the translated writes.
  • Native broker subscriptions and SNS delivery claims are distinct surfaces. Supported delivery protocols and delivery consent remain backend-specific.
  • Google Pub/Sub bills published and delivered bytes at the Message Delivery Basic rate and retained messages per GiB-month. Each request is billed at no less than 1,000 bytes; the family’s batching and delivery assumptions must represent that. The first 10 GiB of combined throughput a month is free per billing account and is not applied. BigQuery and Cloud Storage subscriptions bill USD 50 per TiB instead. Unacknowledged messages younger than one day bill no storage.
  • Row-store throughput (capacityOpsPerVcpuSecond, default 1500) is measured-by-tensor9 as a lower bound: the DynamoDB adapter on Cloud SQL db-custom-2-7680 (2 vCPU, PostgreSQL 16, us-west1) served 2,974 ops/s of YCSB-A at 16 threads without recorded saturation (src/main/kotlin/t9/prime/equiv/generated/eval/CloudsqlPostgresqlYcsbAT16GoYcsbV1_0_0.kt). Published CPU-bound pgbench reaches 3,101 read-write transactions and 20,289 point reads per vCPU-second (AWS z1d.2xlarge, PostgreSQL 14.5), so actual capacity is likely higher. Reads and writes share one coefficient: at the same adapter version, read-only YCSB-C was not faster than 50%-update YCSB-A.
  • Log-broker throughput (capacityStreamMiBPerVcpuSecond 20 at 0.5 utilization, 4 GiB RAM per vCPU) is provider-documented by Google Managed Service for Apache Kafka sizing (https://docs.cloud.google.com/managed-service-for-apache-kafka/docs/plan-cluster-size): write-equivalent bytes are replicated writes plus a quarter of reads. Producer batches below 10 KB lower per-CPU throughput; that penalty is not modeled.
  • Disk utilization 0.8 is provider-documented headroom below Cloud SQL automatic storage increase, which triggers near 91-95% full (https://docs.cloud.google.com/sql/docs/postgres/instance-settings). The 20 GiB reserve exceeds PostgreSQL’s 1 GB default max_wal_size; the remainder of that reserve is unmeasured.
  • CPU tiers double from 2 vCPUs to the target’s largest node at 4 GiB per vCPU; each target’s documented largest node bounds CPU, RAM, RAM per vCPU and disk, and the durable staging database uses Cloud SQL Enterprise (96 vCPU, 6.5 GiB per vCPU, 624 GiB, 64 TiB). Targets billed per capacity unit or throughput unit are sized by their backend instead.
  • No public evidence: peak/mean work (2) and the RAM working-set fraction (0.1). Proposed measurements: the p99 one-minute rate over the monthly mean from customer CloudWatch metrics; the buffer hit ratio while stepping RAM under a fixed YCSB-B load.
  • Node bounds: Cloud SQL Enterprise allows 1-96 vCPUs at 0.9-6.5 GB per vCPU and up to 64 TB with a dedicated vCPU (https://docs.cloud.google.com/sql/docs/postgres/instance-settings).
  • Stored GiB already includes the origin family’s physical footprint assumptions; no second storage amplification is applied.
  • The operations/vCPU coefficient supplies the compute requirement alongside explicit resource-hour minima.
  • Copies and allocation fractions change cost only; they do not increase modeled primary capacity.
  • AWS Price List API, AmazonSNS offer: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonSNS/current/us-east-1/index.json
  • Pub/Sub pricing: https://cloud.google.com/pubsub/pricing
Cost scope: Modeled target native service and explicitly modeled state allocation; hosting and Tensor9 fees only when supplied. Excluded costs and workload features:
  • Additional network transfer, backups, KMS, one-time migration and operational labor unless an explicit meter includes them.
  • Destination application execution and separately billed downstream services; supply those allocations separately.
  • HTTP/email/SMS/mobile delivery prices and destination-service charges; relay submission does not establish managed delivery reputation.
  • Taxes, discounts, credits and free-tier allowances not explicitly listed by this model.
  • Unmodeled dependent services, task execution and operations outside the selected workload family.
List prices read: 2026-10-22. Model: plan-5d3d0327635e1738d279839e37f45523e2cd9b39 / 2026-10-22.planning.2. Curve: workload-1. Sources: AWS Price List API, AmazonSNS offer, Pub/Sub pricing. SNS Standard → Google Cloud Pub/Sub: Authored native-usage planning example. Native monthly usage priced at the cited provider list prices; listed target allowances apply where present, excluding invoice credits, taxes and negotiated discounts. Scenario source region: us-east-1. Scenario target region: us-central1. Currency: USD. Estimated allocation, not benchmark calibrated: sized at 2 vCPU, 8 GiB memory and 100 GiB SSD for this scenario. Unestimated components: Adapter hosting, Tensor9 fees. A modeled adapted total and savings require these costs. Cost scope: Modeled target native service and explicitly modeled state allocation; hosting and Tensor9 fees only when supplied. Excluded costs and workload features:
  • Additional network transfer, backups, KMS, one-time migration and operational labor unless an explicit meter includes them.
  • Destination application execution and separately billed downstream services; supply those allocations separately.
  • HTTP/email/SMS/mobile delivery prices and destination-service charges; relay submission does not establish managed delivery reputation.
  • Taxes, discounts, credits and free-tier allowances not explicitly listed by this model.
  • Unmodeled dependent services, task execution and operations outside the selected workload family.
Scenario assumptions
  • Translated reads: 0 operations.
  • Translated writes: 0 operations.
  • Translated resident physical data: 0 GiB-month.
  • Native provider operations: 0 operations.
  • Published payload: 0 GiB.
  • Delivered payload: 0 GiB.
  • Allocated execution CPU: 0 vCPU-hour.
  • Allocated execution RAM: 0 GiB-hour.
  • Independent resource allocation: 0 resource-hour.
  • Independent inventory: 0 resource-month.
  • Input tokens: 0 tokens.
  • Output tokens: 0 tokens.
  • Billable network: 0 GiB.
  • Dependent object sink retained data: 0 GiB-month.
  • Dependent object sink writes: 0 writes.
  • Rates are cited provider list prices, editable for discounts, commitments or another region. Workload translation and capacity coefficients remain editable planning assumptions, not benchmark measurements or a commercial quote.
  • Default-zero rates are explicit assumptions, not evidence that the deployed service or resource is free.
  • Provisioned-unit minimums use an editable approximate allocation floor. Modeled target service cost is max(usage/allocation, minimum), with billable network, durable staging and dependent sinks added separately; actual RU/TU sizing, scaling increments and throughput remain unqualified.
  • Durable staging is priced once: provisioned targets include adapter state in their selected allocation; native targets add a separate staging allocation when required. Adapter hosting is a separate allocation and must exclude resources already priced as staging, workers or target capacity.
  • A declared adapter mapping and cost model do not establish implementation availability or service equivalence.
  • This is a declared directed mapping; a planning estimate does not establish deployment or request-serving support.
  • Authored planning assumptions, not measured adapter capacity. Origin rates are cited first-tier list prices, before free allowances and discounts.
  • A cost scenario does not establish backend availability, protocol fidelity or production readiness. Registry serving status remains independent.
  • Native Pub/Sub and Service Bus backends also retain durable SNS staging. Price that allocation once; database targets already host the translated writes.
  • Native broker subscriptions and SNS delivery claims are distinct surfaces. Supported delivery protocols and delivery consent remain backend-specific.
  • Google Pub/Sub bills published and delivered bytes at the Message Delivery Basic rate and retained messages per GiB-month. Each request is billed at no less than 1,000 bytes; the family’s batching and delivery assumptions must represent that. The first 10 GiB of combined throughput a month is free per billing account and is not applied. BigQuery and Cloud Storage subscriptions bill USD 50 per TiB instead. Unacknowledged messages younger than one day bill no storage.
  • Row-store throughput (capacityOpsPerVcpuSecond, default 1500) is measured-by-tensor9 as a lower bound: the DynamoDB adapter on Cloud SQL db-custom-2-7680 (2 vCPU, PostgreSQL 16, us-west1) served 2,974 ops/s of YCSB-A at 16 threads without recorded saturation (src/main/kotlin/t9/prime/equiv/generated/eval/CloudsqlPostgresqlYcsbAT16GoYcsbV1_0_0.kt). Published CPU-bound pgbench reaches 3,101 read-write transactions and 20,289 point reads per vCPU-second (AWS z1d.2xlarge, PostgreSQL 14.5), so actual capacity is likely higher. Reads and writes share one coefficient: at the same adapter version, read-only YCSB-C was not faster than 50%-update YCSB-A.
  • Log-broker throughput (capacityStreamMiBPerVcpuSecond 20 at 0.5 utilization, 4 GiB RAM per vCPU) is provider-documented by Google Managed Service for Apache Kafka sizing (https://docs.cloud.google.com/managed-service-for-apache-kafka/docs/plan-cluster-size): write-equivalent bytes are replicated writes plus a quarter of reads. Producer batches below 10 KB lower per-CPU throughput; that penalty is not modeled.
  • Disk utilization 0.8 is provider-documented headroom below Cloud SQL automatic storage increase, which triggers near 91-95% full (https://docs.cloud.google.com/sql/docs/postgres/instance-settings). The 20 GiB reserve exceeds PostgreSQL’s 1 GB default max_wal_size; the remainder of that reserve is unmeasured.
  • CPU tiers double from 2 vCPUs to the target’s largest node at 4 GiB per vCPU; each target’s documented largest node bounds CPU, RAM, RAM per vCPU and disk, and the durable staging database uses Cloud SQL Enterprise (96 vCPU, 6.5 GiB per vCPU, 624 GiB, 64 TiB). Targets billed per capacity unit or throughput unit are sized by their backend instead.
  • No public evidence: peak/mean work (2) and the RAM working-set fraction (0.1). Proposed measurements: the p99 one-minute rate over the monthly mean from customer CloudWatch metrics; the buffer hit ratio while stepping RAM under a fixed YCSB-B load.
  • Node bounds: Cloud SQL Enterprise allows 1-96 vCPUs at 0.9-6.5 GB per vCPU and up to 64 TB with a dedicated vCPU (https://docs.cloud.google.com/sql/docs/postgres/instance-settings).
  • Stored GiB already includes the origin family’s physical footprint assumptions; no second storage amplification is applied.
  • The operations/vCPU coefficient supplies the compute requirement alongside explicit resource-hour minima.
  • Copies and allocation fractions change cost only; they do not increase modeled primary capacity.
List prices read: 2026-10-22. Model: plan-5d3d0327635e1738d279839e37f45523e2cd9b39 / 2026-10-22.planning.2. Scenario: planning-example. Sources: AWS Price List API, AmazonSNS offer, Pub/Sub pricing.

​
Via Google Cloud SQL for PostgreSQL

​
How it works

Your application uses the SNS API and keeps its topic and subscription ARNs. Google Cloud SQL for PostgreSQL stores each message body. A separate set of SNS records holds your subscriptions, queue permissions and the result of each delivery. Publish makes two writes, one to each store. It first saves the message body, then records the publication and one delivery for each subscription that should receive it. Because of this order, a recorded delivery never points to a message that was not saved. If the process stops between the two writes, the saved body is never delivered. Once both writes succeed, the delivery workers find the pending deliveries in the SNS records, even if the process that published the message has stopped.
Publish stores the message body in Cloud SQL, then records one delivery per matching subscription. Before each send, the queue's permission is checked, then the SQS queue receives the message.Publish stores the message body in Cloud SQL, then records one delivery per matching subscription. Before each send, the queue's permission is checked, then the SQS queue receives the message.

Publish returns after the first two steps. Your queues receive the message later.

​
Delivery and retries

When a message is published, SNS decides which active subscriptions receive it, applying their filter policies, and records one delivery for each. Before sending, the delivery worker checks that the subscription still exists and that the queue still allows the topic. Unsubscribe stops new deliveries; it cannot recall a message a queue has already stored. Each SQS subscription succeeds or fails on its own. Once SQS has stored a message, SQS is responsible for it. If SQS stores a message but the answer or the delivery record is lost, another attempt can deliver it again with the same MessageId. The code that processes messages must handle duplicates safely.

​
Limitations

  • FIFO topics are not supported. FIFO topics, ordering and deduplication are not provided. Only standard topics work.
  • Filter policies support a subset of SNS syntax. Message-attribute and message-body filter policies accept a reduced set of operators. A policy using any other operator is rejected.
  • You must keep the queue allowing the topic. The SQS queue’s access policy must allow the topic, and must still allow it at the time of each delivery.
  • Only SQS subscriptions are delivered; redrive is not supported. Subscriptions using HTTP, HTTPS, Lambda, email, SMS, mobile push or Firehose are rejected, and so is a subscription RedrivePolicy.
  • Topic policies are enforced. The topic access policy is checked and enforced. You can set it inline on the topic or with one aws_sns_topic_policy resource, which is applied with the topic. Unsupported policy statements are rejected.

​
Other considerations

Running the database. Google operates Cloud SQL maintenance and the database features chosen for the deployment. Configure database capacity, private connectivity, TLS and backups for your workload. A restore must bring back both the message database and the SNS records; a backup of only one of them is not enough. Before going live. Check queue permissions, confirm that your queues receive messages, and test recovery in this configuration before you move publishers. What to monitor. Watch database storage and connections alongside pending deliveries and delivery failures. Moving from AWS. Messages still pending on AWS SNS are not copied into Cloud SQL.

​
Example monthly charges

This Tensor9 BYOC scenario estimates service charges for the named origin and target services. Adapter hosting and commercial Tensor9 fees need separate allocations. How Tensor9 estimates costs explains the method and what the estimates include.
SNS Standard · SNS publish billing units
Feature: SNS Standard. Estimated monthly service charges by SNS publish billing units. Adapter hosting and Tensor9 fees are excluded. Source: us-east-1; target: us-central1; currency: USD.
SNS Standard · SNS publish billing units: monthly service charges by SNS publish billing units. Exact values are in the following table.SNS Standard · SNS publish billing units: monthly service charges by SNS publish billing units. Exact values are in the following table.
Comparison pills describe all sampled scale points; amounts beside each service are for the named workload. Costs The Same means within 20% at every point. Costs Depend includes mixed results and unpriced samples. Sampled monthly charges in USD, rounded to cents. Difference means target minus origin; a positive difference means the target costs more. Estimated allocation, not benchmark calibrated: sized from 2 vCPU, 8 GiB memory and 100 GiB SSD at 0 per second to 32 vCPU, 128 GiB memory and 100 GiB SSD at 1,000 per second. From 10,000 per second the workload exceeds the planning range and the target is unpriced. Curve assumptions
  • Fan-out and retry assumptions remain explicit; this is not destination delivery TPS. Other native meters and workload assumptions retain their explicitly displayed default quantities. Workload translation and capacity assumptions are not benchmark measurements.
  • Native monthly usage priced at the cited provider list prices; listed target allowances apply where present, excluding invoice credits, taxes and negotiated discounts. Rates as of 2026-10-22: us-east-1 → us-central1, USD/month.
  • Source prices are reference charges, not an invoice. Target lines cover selected service usage/configuration only; adapter hosting and Tensor9 fees are excluded.
  • Samples of explicit prices are not performance measurements or capacity qualification. No interpolation or break-even guarantee is implied.
  • Variant assumptions show the first axis sample; any axis-dependent configuration is described by this curve’s workload note.
  • Rates are cited provider list prices, editable for discounts, commitments or another region. Workload translation and capacity coefficients remain editable planning assumptions, not benchmark measurements or a commercial quote.
  • Default-zero rates are explicit assumptions, not evidence that the deployed service or resource is free.
  • Provisioned-unit minimums use an editable approximate allocation floor. Modeled target service cost is max(usage/allocation, minimum), with billable network, durable staging and dependent sinks added separately; actual RU/TU sizing, scaling increments and throughput remain unqualified.
  • Durable staging is priced once: provisioned targets include adapter state in their selected allocation; native targets add a separate staging allocation when required. Adapter hosting is a separate allocation and must exclude resources already priced as staging, workers or target capacity.
  • A declared adapter mapping and cost model do not establish implementation availability or service equivalence.
  • This is a declared directed mapping; a planning estimate does not establish deployment or request-serving support.
  • Authored planning assumptions, not measured adapter capacity. Origin rates are cited first-tier list prices, before free allowances and discounts.
  • A cost scenario does not establish backend availability, protocol fidelity or production readiness. Registry serving status remains independent.
  • Native Pub/Sub and Service Bus backends also retain durable SNS staging. Price that allocation once; database targets already host the translated writes.
  • Native broker subscriptions and SNS delivery claims are distinct surfaces. Supported delivery protocols and delivery consent remain backend-specific.
  • Cloud SQL Enterprise edition (PostgreSQL) publishes separate vCPU, memory and SSD prices for zonal instances; a high-availability (regional) instance bills exactly twice each, which the standby copy reproduces. Enterprise Plus, committed-use discounts, backups (USD 0.08 per GiB-month) and idle IPv4 addresses (0.01 per hour) are not selected.
  • Row-store throughput (capacityOpsPerVcpuSecond, default 1500) is measured-by-tensor9 as a lower bound: the DynamoDB adapter on Cloud SQL db-custom-2-7680 (2 vCPU, PostgreSQL 16, us-west1) served 2,974 ops/s of YCSB-A at 16 threads without recorded saturation (src/main/kotlin/t9/prime/equiv/generated/eval/CloudsqlPostgresqlYcsbAT16GoYcsbV1_0_0.kt). Published CPU-bound pgbench reaches 3,101 read-write transactions and 20,289 point reads per vCPU-second (AWS z1d.2xlarge, PostgreSQL 14.5), so actual capacity is likely higher. Reads and writes share one coefficient: at the same adapter version, read-only YCSB-C was not faster than 50%-update YCSB-A.
  • Log-broker throughput (capacityStreamMiBPerVcpuSecond 20 at 0.5 utilization, 4 GiB RAM per vCPU) is provider-documented by Google Managed Service for Apache Kafka sizing (https://docs.cloud.google.com/managed-service-for-apache-kafka/docs/plan-cluster-size): write-equivalent bytes are replicated writes plus a quarter of reads. Producer batches below 10 KB lower per-CPU throughput; that penalty is not modeled.
  • Disk utilization 0.8 is provider-documented headroom below Cloud SQL automatic storage increase, which triggers near 91-95% full (https://docs.cloud.google.com/sql/docs/postgres/instance-settings). The 20 GiB reserve exceeds PostgreSQL’s 1 GB default max_wal_size; the remainder of that reserve is unmeasured.
  • CPU tiers double from 2 vCPUs to the target’s largest node at 4 GiB per vCPU; each target’s documented largest node bounds CPU, RAM, RAM per vCPU and disk, and the durable staging database uses Cloud SQL Enterprise (96 vCPU, 6.5 GiB per vCPU, 624 GiB, 64 TiB). Targets billed per capacity unit or throughput unit are sized by their backend instead.
  • No public evidence: peak/mean work (2) and the RAM working-set fraction (0.1). Proposed measurements: the p99 one-minute rate over the monthly mean from customer CloudWatch metrics; the buffer hit ratio while stepping RAM under a fixed YCSB-B load.
  • Node bounds: Cloud SQL Enterprise allows 1-96 vCPUs at 0.9-6.5 GB per vCPU and up to 64 TB with a dedicated vCPU (https://docs.cloud.google.com/sql/docs/postgres/instance-settings).
  • Stored GiB already includes the origin family’s physical footprint assumptions; no second storage amplification is applied.
  • The operations/vCPU coefficient supplies the compute requirement alongside explicit resource-hour minima.
  • Copies and allocation fractions change cost only; they do not increase modeled primary capacity.
  • AWS Price List API, AmazonSNS offer: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonSNS/current/us-east-1/index.json
  • Cloud SQL pricing: https://cloud.google.com/sql/pricing
Cost scope: Modeled target native service and explicitly modeled state allocation; hosting and Tensor9 fees only when supplied. Excluded costs and workload features:
  • Additional network transfer, backups, KMS, one-time migration and operational labor unless an explicit meter includes them.
  • Destination application execution and separately billed downstream services; supply those allocations separately.
  • HTTP/email/SMS/mobile delivery prices and destination-service charges; relay submission does not establish managed delivery reputation.
  • Taxes, discounts, credits and free-tier allowances not explicitly listed by this model.
  • Unmodeled dependent services, task execution and operations outside the selected workload family.
List prices read: 2026-10-22. Model: plan-56469bafbc69bb655a97cbd8145f3a1d5d0f6ce1 / 2026-10-22.planning.2. Curve: workload-1. Sources: AWS Price List API, AmazonSNS offer, Cloud SQL pricing. SNS Standard → Google Cloud SQL for PostgreSQL: Authored native-usage planning example. Native monthly usage priced at the cited provider list prices; listed target allowances apply where present, excluding invoice credits, taxes and negotiated discounts. Scenario source region: us-east-1. Scenario target region: us-central1. Currency: USD. Estimated allocation, not benchmark calibrated: sized at 2 vCPU, 8 GiB memory and 100 GiB SSD for this scenario. Unestimated components: Adapter hosting, Tensor9 fees. A modeled adapted total and savings require these costs. Cost scope: Modeled target native service and explicitly modeled state allocation; hosting and Tensor9 fees only when supplied. Excluded costs and workload features:
  • Additional network transfer, backups, KMS, one-time migration and operational labor unless an explicit meter includes them.
  • Destination application execution and separately billed downstream services; supply those allocations separately.
  • HTTP/email/SMS/mobile delivery prices and destination-service charges; relay submission does not establish managed delivery reputation.
  • Taxes, discounts, credits and free-tier allowances not explicitly listed by this model.
  • Unmodeled dependent services, task execution and operations outside the selected workload family.
Scenario assumptions
  • Translated reads: 0 operations.
  • Translated writes: 0 operations.
  • Translated resident physical data: 0 GiB-month.
  • Native provider operations: 0 operations.
  • Published payload: 0 GiB.
  • Delivered payload: 0 GiB.
  • Allocated execution CPU: 0 vCPU-hour.
  • Allocated execution RAM: 0 GiB-hour.
  • Independent resource allocation: 0 resource-hour.
  • Independent inventory: 0 resource-month.
  • Input tokens: 0 tokens.
  • Output tokens: 0 tokens.
  • Billable network: 0 GiB.
  • Dependent object sink retained data: 0 GiB-month.
  • Dependent object sink writes: 0 writes.
  • Rates are cited provider list prices, editable for discounts, commitments or another region. Workload translation and capacity coefficients remain editable planning assumptions, not benchmark measurements or a commercial quote.
  • Default-zero rates are explicit assumptions, not evidence that the deployed service or resource is free.
  • Provisioned-unit minimums use an editable approximate allocation floor. Modeled target service cost is max(usage/allocation, minimum), with billable network, durable staging and dependent sinks added separately; actual RU/TU sizing, scaling increments and throughput remain unqualified.
  • Durable staging is priced once: provisioned targets include adapter state in their selected allocation; native targets add a separate staging allocation when required. Adapter hosting is a separate allocation and must exclude resources already priced as staging, workers or target capacity.
  • A declared adapter mapping and cost model do not establish implementation availability or service equivalence.
  • This is a declared directed mapping; a planning estimate does not establish deployment or request-serving support.
  • Authored planning assumptions, not measured adapter capacity. Origin rates are cited first-tier list prices, before free allowances and discounts.
  • A cost scenario does not establish backend availability, protocol fidelity or production readiness. Registry serving status remains independent.
  • Native Pub/Sub and Service Bus backends also retain durable SNS staging. Price that allocation once; database targets already host the translated writes.
  • Native broker subscriptions and SNS delivery claims are distinct surfaces. Supported delivery protocols and delivery consent remain backend-specific.
  • Cloud SQL Enterprise edition (PostgreSQL) publishes separate vCPU, memory and SSD prices for zonal instances; a high-availability (regional) instance bills exactly twice each, which the standby copy reproduces. Enterprise Plus, committed-use discounts, backups (USD 0.08 per GiB-month) and idle IPv4 addresses (0.01 per hour) are not selected.
  • Row-store throughput (capacityOpsPerVcpuSecond, default 1500) is measured-by-tensor9 as a lower bound: the DynamoDB adapter on Cloud SQL db-custom-2-7680 (2 vCPU, PostgreSQL 16, us-west1) served 2,974 ops/s of YCSB-A at 16 threads without recorded saturation (src/main/kotlin/t9/prime/equiv/generated/eval/CloudsqlPostgresqlYcsbAT16GoYcsbV1_0_0.kt). Published CPU-bound pgbench reaches 3,101 read-write transactions and 20,289 point reads per vCPU-second (AWS z1d.2xlarge, PostgreSQL 14.5), so actual capacity is likely higher. Reads and writes share one coefficient: at the same adapter version, read-only YCSB-C was not faster than 50%-update YCSB-A.
  • Log-broker throughput (capacityStreamMiBPerVcpuSecond 20 at 0.5 utilization, 4 GiB RAM per vCPU) is provider-documented by Google Managed Service for Apache Kafka sizing (https://docs.cloud.google.com/managed-service-for-apache-kafka/docs/plan-cluster-size): write-equivalent bytes are replicated writes plus a quarter of reads. Producer batches below 10 KB lower per-CPU throughput; that penalty is not modeled.
  • Disk utilization 0.8 is provider-documented headroom below Cloud SQL automatic storage increase, which triggers near 91-95% full (https://docs.cloud.google.com/sql/docs/postgres/instance-settings). The 20 GiB reserve exceeds PostgreSQL’s 1 GB default max_wal_size; the remainder of that reserve is unmeasured.
  • CPU tiers double from 2 vCPUs to the target’s largest node at 4 GiB per vCPU; each target’s documented largest node bounds CPU, RAM, RAM per vCPU and disk, and the durable staging database uses Cloud SQL Enterprise (96 vCPU, 6.5 GiB per vCPU, 624 GiB, 64 TiB). Targets billed per capacity unit or throughput unit are sized by their backend instead.
  • No public evidence: peak/mean work (2) and the RAM working-set fraction (0.1). Proposed measurements: the p99 one-minute rate over the monthly mean from customer CloudWatch metrics; the buffer hit ratio while stepping RAM under a fixed YCSB-B load.
  • Node bounds: Cloud SQL Enterprise allows 1-96 vCPUs at 0.9-6.5 GB per vCPU and up to 64 TB with a dedicated vCPU (https://docs.cloud.google.com/sql/docs/postgres/instance-settings).
  • Stored GiB already includes the origin family’s physical footprint assumptions; no second storage amplification is applied.
  • The operations/vCPU coefficient supplies the compute requirement alongside explicit resource-hour minima.
  • Copies and allocation fractions change cost only; they do not increase modeled primary capacity.
List prices read: 2026-10-22. Model: plan-56469bafbc69bb655a97cbd8145f3a1d5d0f6ce1 / 2026-10-22.planning.2. Scenario: planning-example. Sources: AWS Price List API, AmazonSNS offer, Cloud SQL pricing.

​
On Azure, OCI, and Private Kubernetes

​
How it works

Your application keeps its SNS client and points it at an adapter in the target cluster. The message body is stored in PostgreSQL. A separate set of SNS records holds your topics, subscriptions, filter policies, queue permissions and the result of each delivery. Publish makes two writes. It first saves the message body without sending it, then records the publication and one delivery for each subscription that should receive it. Publish returns a MessageId only after both writes succeed. If the process stops between the two writes, the saved body is never sent to anyone. Deliveries that were recorded are picked up again after the delivery workers restart.
Publish saves the message in PostgreSQL and records one delivery per matching subscription. Before each send, the queue's permission is checked, then the SQS queue receives the message.Publish saves the message in PostgreSQL and records one delivery per matching subscription. Before each send, the queue's permission is checked, then the SQS queue receives the message.

Publish returns a MessageId only after the message and its deliveries are both saved.

​
Delivery and duplicates

When a message is published, SNS decides which active subscriptions receive it, applying their filter policies, and records one delivery for each. Just before sending, it checks that the queue still allows the topic to send to it. After a subscription is deleted, no further deliveries are attempted for it, and a subscription created to replace it does not receive the old one’s pending messages. Each SQS delivery succeeds or fails on its own. Once SQS has stored a message, SQS is responsible for it. If a send to SQS ends without a clear answer, the message can be delivered again with the same MessageId, so the code that processes it must handle duplicates.

​
Limitations

  • FIFO topics are not supported. FIFO ordering, deduplication and MessageGroupId are rejected. Only standard topics work.
  • Only SQS subscriptions are delivered. Subscriptions using HTTP, HTTPS, Lambda, email, SMS or any other protocol are rejected.
  • Filter policies support a subset of SNS syntax. Attribute and message-body filter policies accept a reduced set of operators. Raw message delivery works only for SQS subscriptions.
  • The queue must keep allowing the topic. The SQS queue’s access policy must allow the topic, and must still allow it at the time of each delivery.
  • Topic policies are enforced; redrive is not supported. The topic access policy is checked and enforced. You can set it inline on the topic or with one aws_sns_topic_policy resource, which is applied with the topic. Unsupported policy statements and a subscription RedrivePolicy are rejected.

​
Other considerations

You run the database. The deployment operator is responsible for PostgreSQL storage, replication, backups and recovery on the target cluster. Size the database for the messages you keep and the number of connections. Back up and restore the message bodies and the SNS records together. Before moving publishers. Check queue permissions and confirm that pending deliveries resume after a restart, then point publishers at the new endpoint. Deliveries still pending on AWS are not copied into PostgreSQL; let them finish or expire on AWS.

​
Example monthly charges

This Tensor9 BYOC scenario estimates service charges for the named origin and target services. Adapter hosting and commercial Tensor9 fees need separate allocations. How Tensor9 estimates costs explains the method and what the estimates include.
SNS Standard · SNS publish billing units
Feature: SNS Standard. Estimated monthly service charges by SNS publish billing units. Adapter hosting and Tensor9 fees are excluded. Source: us-east-1; target: us-central1; currency: USD.
SNS Standard · SNS publish billing units: monthly service charges by SNS publish billing units. Exact values are in the following table.SNS Standard · SNS publish billing units: monthly service charges by SNS publish billing units. Exact values are in the following table.
Comparison pills describe all sampled scale points; amounts beside each service are for the named workload. Costs The Same means within 20% at every point. Costs Depend includes mixed results and unpriced samples. Sampled monthly charges in USD, rounded to cents. Difference means target minus origin; a positive difference means the target costs more. Estimated allocation, not benchmark calibrated: sized from 2 vCPU, 8 GiB memory and 100 GiB SSD at 0 per second to 32 vCPU, 128 GiB memory and 100 GiB SSD at 1,000 per second. From 10,000 per second the workload exceeds the planning range and the target is unpriced. Curve assumptions
  • Fan-out and retry assumptions remain explicit; this is not destination delivery TPS. Other native meters and workload assumptions retain their explicitly displayed default quantities. Workload translation and capacity assumptions are not benchmark measurements.
  • Native monthly usage priced at the cited provider list prices; listed target allowances apply where present, excluding invoice credits, taxes and negotiated discounts. Rates as of 2026-10-22: us-east-1 → us-central1, USD/month.
  • Source prices are reference charges, not an invoice. Target lines cover selected service usage/configuration only; adapter hosting and Tensor9 fees are excluded.
  • Samples of explicit prices are not performance measurements or capacity qualification. No interpolation or break-even guarantee is implied.
  • Variant assumptions show the first axis sample; any axis-dependent configuration is described by this curve’s workload note.
  • Rates are cited provider list prices, editable for discounts, commitments or another region. Workload translation and capacity coefficients remain editable planning assumptions, not benchmark measurements or a commercial quote.
  • Default-zero rates are explicit assumptions, not evidence that the deployed service or resource is free.
  • Provisioned-unit minimums use an editable approximate allocation floor. Modeled target service cost is max(usage/allocation, minimum), with billable network, durable staging and dependent sinks added separately; actual RU/TU sizing, scaling increments and throughput remain unqualified.
  • Durable staging is priced once: provisioned targets include adapter state in their selected allocation; native targets add a separate staging allocation when required. Adapter hosting is a separate allocation and must exclude resources already priced as staging, workers or target capacity.
  • A declared adapter mapping and cost model do not establish implementation availability or service equivalence.
  • This is a declared directed mapping; a planning estimate does not establish deployment or request-serving support.
  • Authored planning assumptions, not measured adapter capacity. Origin rates are cited first-tier list prices, before free allowances and discounts.
  • A cost scenario does not establish backend availability, protocol fidelity or production readiness. Registry serving status remains independent.
  • Native Pub/Sub and Service Bus backends also retain durable SNS staging. Price that allocation once; database targets already host the translated writes.
  • Native broker subscriptions and SNS delivery claims are distinct surfaces. Supported delivery protocols and delivery consent remain backend-specific.
  • Self-hosted target: PostgreSQL (PostgreSQL License, https://www.postgresql.org/about/licence/). Selected CPU, RAM and physical disk allocation grow with workload; no per-operation price is assumed. No vendor charge: the target runs on the customer’s Kubernetes nodes, so CPU, RAM and disk bill at the Kubernetes node reference rates, Compute Engine N2 in us-central1 (USD 0.031611/vCPU-hour, 0.004237/GiB-hour) with pd-balanced disk (0.10/GiB-month) and Premium Tier egress. The cost-optimized E2 series lists 0.02181159/vCPU-hour and 0.00292353/GiB-hour; AWS m6i/r6i and Azure D2s v5/E2s v5 both decompose to 0.033/vCPU-hour and 0.00375/GiB-hour. Requested CPU and RAM are billed; node bin-packing waste, system reservations and the cluster control-plane fee are not.
  • Row-store throughput (capacityOpsPerVcpuSecond, default 1500) is measured-by-tensor9 as a lower bound: the DynamoDB adapter on Cloud SQL db-custom-2-7680 (2 vCPU, PostgreSQL 16, us-west1) served 2,974 ops/s of YCSB-A at 16 threads without recorded saturation (src/main/kotlin/t9/prime/equiv/generated/eval/CloudsqlPostgresqlYcsbAT16GoYcsbV1_0_0.kt). Published CPU-bound pgbench reaches 3,101 read-write transactions and 20,289 point reads per vCPU-second (AWS z1d.2xlarge, PostgreSQL 14.5), so actual capacity is likely higher. Reads and writes share one coefficient: at the same adapter version, read-only YCSB-C was not faster than 50%-update YCSB-A.
  • Log-broker throughput (capacityStreamMiBPerVcpuSecond 20 at 0.5 utilization, 4 GiB RAM per vCPU) is provider-documented by Google Managed Service for Apache Kafka sizing (https://docs.cloud.google.com/managed-service-for-apache-kafka/docs/plan-cluster-size): write-equivalent bytes are replicated writes plus a quarter of reads. Producer batches below 10 KB lower per-CPU throughput; that penalty is not modeled.
  • Disk utilization 0.8 is provider-documented headroom below Cloud SQL automatic storage increase, which triggers near 91-95% full (https://docs.cloud.google.com/sql/docs/postgres/instance-settings). The 20 GiB reserve exceeds PostgreSQL’s 1 GB default max_wal_size; the remainder of that reserve is unmeasured.
  • CPU tiers double from 2 vCPUs to the target’s largest node at 4 GiB per vCPU; each target’s documented largest node bounds CPU, RAM, RAM per vCPU and disk, and the durable staging database uses Cloud SQL Enterprise (96 vCPU, 6.5 GiB per vCPU, 624 GiB, 64 TiB). Targets billed per capacity unit or throughput unit are sized by their backend instead.
  • No public evidence: peak/mean work (2) and the RAM working-set fraction (0.1). Proposed measurements: the p99 one-minute rate over the monthly mean from customer CloudWatch metrics; the buffer hit ratio while stepping RAM under a fixed YCSB-B load.
  • Node bounds: Compute Engine n2-highmem-128 has 128 vCPUs and 864 GB (https://docs.cloud.google.com/compute/docs/general-purpose-machines); a Persistent Disk volume holds up to 64 TiB (https://docs.cloud.google.com/compute/docs/disks/persistent-disks).
  • Stored GiB already includes the origin family’s physical footprint assumptions; no second storage amplification is applied.
  • The operations/vCPU coefficient supplies the compute requirement alongside explicit resource-hour minima.
  • Copies and allocation fractions change cost only; they do not increase modeled primary capacity.
  • AWS Price List API, AmazonSNS offer: https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonSNS/current/us-east-1/index.json
  • Compute Engine general-purpose machine pricing: https://cloud.google.com/products/compute/pricing/general-purpose
  • Disk and image pricing: https://cloud.google.com/compute/disks-image-pricing
  • Network pricing: https://cloud.google.com/vpc/network-pricing
Cost scope: Modeled target native service and explicitly modeled state allocation; hosting and Tensor9 fees only when supplied. Excluded costs and workload features:
  • Additional network transfer, backups, KMS, one-time migration and operational labor unless an explicit meter includes them.
  • Destination application execution and separately billed downstream services; supply those allocations separately.
  • HTTP/email/SMS/mobile delivery prices and destination-service charges; relay submission does not establish managed delivery reputation.
  • Taxes, discounts, credits and free-tier allowances not explicitly listed by this model.
  • Unmodeled dependent services, task execution and operations outside the selected workload family.
List prices read: 2026-10-22. Model: plan-9063cd5055d0b6404b79cb2941f9fbd779c5669b / 2026-10-22.planning.2. Curve: workload-1. Sources: AWS Price List API, AmazonSNS offer, Compute Engine general-purpose machine pricing, Disk and image pricing, Network pricing. SNS Standard → PostgreSQL: Authored native-usage planning example. Native monthly usage priced at the cited provider list prices; listed target allowances apply where present, excluding invoice credits, taxes and negotiated discounts. Scenario source region: us-east-1. Scenario target region: us-central1. Currency: USD. Estimated allocation, not benchmark calibrated: sized at 2 vCPU, 8 GiB memory and 100 GiB SSD for this scenario. Unestimated components: Adapter hosting, Tensor9 fees. A modeled adapted total and savings require these costs. Cost scope: Modeled target native service and explicitly modeled state allocation; hosting and Tensor9 fees only when supplied. Excluded costs and workload features:
  • Additional network transfer, backups, KMS, one-time migration and operational labor unless an explicit meter includes them.
  • Destination application execution and separately billed downstream services; supply those allocations separately.
  • HTTP/email/SMS/mobile delivery prices and destination-service charges; relay submission does not establish managed delivery reputation.
  • Taxes, discounts, credits and free-tier allowances not explicitly listed by this model.
  • Unmodeled dependent services, task execution and operations outside the selected workload family.
Scenario assumptions
  • Translated reads: 0 operations.
  • Translated writes: 0 operations.
  • Translated resident physical data: 0 GiB-month.
  • Native provider operations: 0 operations.
  • Published payload: 0 GiB.
  • Delivered payload: 0 GiB.
  • Allocated execution CPU: 0 vCPU-hour.
  • Allocated execution RAM: 0 GiB-hour.
  • Independent resource allocation: 0 resource-hour.
  • Independent inventory: 0 resource-month.
  • Input tokens: 0 tokens.
  • Output tokens: 0 tokens.
  • Billable network: 0 GiB.
  • Dependent object sink retained data: 0 GiB-month.
  • Dependent object sink writes: 0 writes.
  • Rates are cited provider list prices, editable for discounts, commitments or another region. Workload translation and capacity coefficients remain editable planning assumptions, not benchmark measurements or a commercial quote.
  • Default-zero rates are explicit assumptions, not evidence that the deployed service or resource is free.
  • Provisioned-unit minimums use an editable approximate allocation floor. Modeled target service cost is max(usage/allocation, minimum), with billable network, durable staging and dependent sinks added separately; actual RU/TU sizing, scaling increments and throughput remain unqualified.
  • Durable staging is priced once: provisioned targets include adapter state in their selected allocation; native targets add a separate staging allocation when required. Adapter hosting is a separate allocation and must exclude resources already priced as staging, workers or target capacity.
  • A declared adapter mapping and cost model do not establish implementation availability or service equivalence.
  • This is a declared directed mapping; a planning estimate does not establish deployment or request-serving support.
  • Authored planning assumptions, not measured adapter capacity. Origin rates are cited first-tier list prices, before free allowances and discounts.
  • A cost scenario does not establish backend availability, protocol fidelity or production readiness. Registry serving status remains independent.
  • Native Pub/Sub and Service Bus backends also retain durable SNS staging. Price that allocation once; database targets already host the translated writes.
  • Native broker subscriptions and SNS delivery claims are distinct surfaces. Supported delivery protocols and delivery consent remain backend-specific.
  • Self-hosted target: PostgreSQL (PostgreSQL License, https://www.postgresql.org/about/licence/). Selected CPU, RAM and physical disk allocation grow with workload; no per-operation price is assumed. No vendor charge: the target runs on the customer’s Kubernetes nodes, so CPU, RAM and disk bill at the Kubernetes node reference rates, Compute Engine N2 in us-central1 (USD 0.031611/vCPU-hour, 0.004237/GiB-hour) with pd-balanced disk (0.10/GiB-month) and Premium Tier egress. The cost-optimized E2 series lists 0.02181159/vCPU-hour and 0.00292353/GiB-hour; AWS m6i/r6i and Azure D2s v5/E2s v5 both decompose to 0.033/vCPU-hour and 0.00375/GiB-hour. Requested CPU and RAM are billed; node bin-packing waste, system reservations and the cluster control-plane fee are not.
  • Row-store throughput (capacityOpsPerVcpuSecond, default 1500) is measured-by-tensor9 as a lower bound: the DynamoDB adapter on Cloud SQL db-custom-2-7680 (2 vCPU, PostgreSQL 16, us-west1) served 2,974 ops/s of YCSB-A at 16 threads without recorded saturation (src/main/kotlin/t9/prime/equiv/generated/eval/CloudsqlPostgresqlYcsbAT16GoYcsbV1_0_0.kt). Published CPU-bound pgbench reaches 3,101 read-write transactions and 20,289 point reads per vCPU-second (AWS z1d.2xlarge, PostgreSQL 14.5), so actual capacity is likely higher. Reads and writes share one coefficient: at the same adapter version, read-only YCSB-C was not faster than 50%-update YCSB-A.
  • Log-broker throughput (capacityStreamMiBPerVcpuSecond 20 at 0.5 utilization, 4 GiB RAM per vCPU) is provider-documented by Google Managed Service for Apache Kafka sizing (https://docs.cloud.google.com/managed-service-for-apache-kafka/docs/plan-cluster-size): write-equivalent bytes are replicated writes plus a quarter of reads. Producer batches below 10 KB lower per-CPU throughput; that penalty is not modeled.
  • Disk utilization 0.8 is provider-documented headroom below Cloud SQL automatic storage increase, which triggers near 91-95% full (https://docs.cloud.google.com/sql/docs/postgres/instance-settings). The 20 GiB reserve exceeds PostgreSQL’s 1 GB default max_wal_size; the remainder of that reserve is unmeasured.
  • CPU tiers double from 2 vCPUs to the target’s largest node at 4 GiB per vCPU; each target’s documented largest node bounds CPU, RAM, RAM per vCPU and disk, and the durable staging database uses Cloud SQL Enterprise (96 vCPU, 6.5 GiB per vCPU, 624 GiB, 64 TiB). Targets billed per capacity unit or throughput unit are sized by their backend instead.
  • No public evidence: peak/mean work (2) and the RAM working-set fraction (0.1). Proposed measurements: the p99 one-minute rate over the monthly mean from customer CloudWatch metrics; the buffer hit ratio while stepping RAM under a fixed YCSB-B load.
  • Node bounds: Compute Engine n2-highmem-128 has 128 vCPUs and 864 GB (https://docs.cloud.google.com/compute/docs/general-purpose-machines); a Persistent Disk volume holds up to 64 TiB (https://docs.cloud.google.com/compute/docs/disks/persistent-disks).
  • Stored GiB already includes the origin family’s physical footprint assumptions; no second storage amplification is applied.
  • The operations/vCPU coefficient supplies the compute requirement alongside explicit resource-hour minima.
  • Copies and allocation fractions change cost only; they do not increase modeled primary capacity.
List prices read: 2026-10-22. Model: plan-9063cd5055d0b6404b79cb2941f9fbd779c5669b / 2026-10-22.planning.2. Scenario: planning-example. Sources: AWS Price List API, AmazonSNS offer, Compute Engine general-purpose machine pricing, Disk and image pricing, Network pricing.

​
Costs

Estimated monthly service charges for each backend with a published cost model, at one sampled workload from that model’s cost curve. Adapter hosting and Tensor9 fees are excluded. How Tensor9 estimates costs explains the method, what the estimates include and how to read their qualifications. Difference means backend minus origin; a positive difference means the backend costs more. Workloads and target configurations differ by model, so compare each row’s difference at its stated workload rather than charges across rows. Cost curves across workload sizes, with every assumption and exclusion: SNS (Simple Notification Service) costs.

​
Service level agreements

These tables show Tensor9’s standard service levels and adapters whose standard terms are still being defined. Your signed agreement determines the SLAs, covered adapters and operations, limits, remedies and support terms that apply to your deployment.
Listed terms cover requests sent through this service adapter. They do not cover direct connections to the target service or replace the cloud provider’s own SLA. Each origin-to-target pair has its own terms or an explicit pending status. See the SLA tables, measurement rules, and scaling conditions.

​
SNS (Simple Notification Service) to Google Cloud SQL for PostgreSQL

Adapter-added latency. A + prefix marks time added by the adapter, not the total time your application waits. Total request time includes adapter processing, backend service time and network time. The backend provider’s SLA applies independently. Only separately measured target-service wait may be subtracted, and only when your agreement permits it; internal metadata, coordination and retry processing still count. For one request, 12 ms of adapter processing plus 90 ms waiting for the backend means at least 102 ms total. This example does not add monthly p99 values. What Tensor9 covers. These service levels cover the adapter between the origin API and target API. They do not replace the target provider’s SLA. How latency is counted. Adapter work, including metadata and coordination, counts. Only separately measured permitted target waits are excluded. Which terms apply. Your signed agreement names the covered operations, workload limits, remedies, and final service levels for your deployment.

​
What this adapter does

For google::1.0.0::cloudsql::postgresql, Publish persists accepted notifications and per-subscription delivery work in a PostgreSQL outbox. Workers deliver later and record their result. The latency row ends at API acceptance, not successful receipt by every subscriber.

​
How the latency targets were chosen

Bounded notification acceptance: +20 ms monthly p99 (adapter-added). Publish. The PostgreSQL path persists per-subscriber delivery work; native brokers fan out differently. Adapter subscription/authorization/outbox work stays on the fabric clock. Bounded notification batch acceptance: +50 ms monthly p99 (adapter-added). PublishBatch. Batch decoding, target calls and delivery scheduling are bounded by both entries and fan-out. Topic and subscription configuration: +20 ms monthly p99 (adapter-added). CreateTopic, DeleteTopic, GetSubscriptionAttributes, GetTopicAttributes, ListTagsForResource, SetSubscriptionAttributes, SetTopicAttributes, Subscribe, TagResource, Unsubscribe, UntagResource. Configuration and authorization state must be persisted; waiting for a human or endpoint confirmation is not promised within this request-response budget. Bounded topic and subscription pages: +20 ms monthly p99 (adapter-added). ListSubscriptions, ListSubscriptionsByTopic, ListTopics. Persistent catalog reads and token/result generation are included. Your signed agreement sets the terms for your deployment. A target does not add an operation or option that the compatibility tables mark unavailable.

​
What counts toward latency

For a request-response row, measure from the agreed ingress boundary to dispatch of the complete response. A row that explicitly names a first response chunk ends at that chunk; its number does not cover the rest of the stream. A long-poll row names the intentional wait and when adapter delay starts. Include parsing, authorization, admission, translation, serialization, adapter-owned storage and coordination, retries and response handling. Subtract only separately measured target-workload waits and external network segments allowed by the measurement rules. The adapter’s own response handling and dispatch remain covered. A database used for adapter metadata is still adapter work, even if a cloud provider hosts it. For concurrent calls, exclude the union of permitted wait intervals, not the sum of overlapping spans. Calculate each request’s adapter duration first, then the monthly p99. Do not subtract one service’s p99 from another’s. Known adapter timeouts are over-budget samples; failed or incomplete requests cannot disappear to improve the percentile. Missing measurements do not become zero latency. An SDK call span alone does not prove how much of its duration can be excluded.

​
Endpoint reachability and request failures

The endpoint-reachability target measures adapter health checks. Correctly forwarding a target quota or permission error is different from producing that error because the adapter sent the wrong request. Adapter-caused failures remain Tensor9’s responsibility even when the target is healthy. Correctness and data-integrity requirements apply to every supported request. Endpoint reachability does not measure request correctness or grant an allowance for incorrect results.

​
Scaling and target-service capacity

Bound subscriptions per topic, filter complexity, payload bytes and queued deliveries as well as publish rate. A publish with 20 subscribers creates up to 20 units of delivery work. Scale request handlers and delivery workers independently, and watch outbox age or broker backlog. Subscriber timeouts and provider quotas can limit delivery even when the adapter accepts new work quickly; the agreement must state a separate delivery/backlog objective if one is required. Tell Tensor9 the expected steady rate, bursts, concurrency, payload sizes and operation mix. Tensor9 sizes and scales the adapter for the agreed load; you choose and monitor the target service’s capacity with Tensor9’s help. A latency budget is not a requests-per-second rating. Larger requests and higher rates need explicit terms, not silent inheritance of a small-request SLA.

​
Data and behavior guarantees

A successful Publish must correspond to the documented durable acceptance point. Retries can duplicate downstream delivery, so consumers remain idempotent. Subscription activation must respect receiver consent and supported protocol rules. Filter policy, message attributes and FIFO behavior are limited by this target profile. Unsupported subscription or publish shapes are not implied by the numerical acceptance row.

​
Example and diagnosis

For topic order-created, publishing one 2 KiB event to ten subscriptions uses the single-publish row. The request can satisfy its SLA while an unreachable webhook accumulates retries; inspect delivery state separately. If adapter-owned outbox insertion is slow, that is fabric time rather than subscriber time and remains Tensor9’s responsibility. Use tensor9 explain and the documented explain headers to understand the selected adapter and its behavior. Correlate available request diagnostics with the target provider’s latency, throttling and capacity metrics. An explain report helps diagnose a request; it is not by itself a qualified SLA timing measurement. Share the operation, request shape, timestamps and request identifiers with support, with credentials and customer payloads removed.

​
SNS (Simple Notification Service) to Google Cloud Pub/Sub

Adapter-added latency. A + prefix marks time added by the adapter, not the total time your application waits. Total request time includes adapter processing, backend service time and network time. The backend provider’s SLA applies independently. Only separately measured target-service wait may be subtracted, and only when your agreement permits it; internal metadata, coordination and retry processing still count. For one request, 12 ms of adapter processing plus 90 ms waiting for the backend means at least 102 ms total. This example does not add monthly p99 values. What Tensor9 covers. These service levels cover the adapter between the origin API and target API. They do not replace the target provider’s SLA. How latency is counted. Adapter work, including metadata and coordination, counts. Only separately measured permitted target waits are excluded. Which terms apply. Your signed agreement names the covered operations, workload limits, remedies, and final service levels for your deployment.

​
What this adapter does

For google::1.0.0::pubsub, Publish maps to the native topic service, while adapter workers can pull broker subscriptions and forward notifications to the configured destination. Native topic acceptance and final HTTP/SQS/email delivery are different events; the request SLA covers the first adapter interaction.

​
How the latency targets were chosen

Bounded notification acceptance: +3 ms monthly p99 (adapter-added). Publish. The PostgreSQL path persists per-subscriber delivery work; native brokers fan out differently. Adapter subscription/authorization/outbox work stays on the fabric clock. Bounded notification batch acceptance: +10 ms monthly p99 (adapter-added). PublishBatch. Batch decoding, target calls and delivery scheduling are bounded by both entries and fan-out. Topic and subscription configuration: +20 ms monthly p99 (adapter-added). CreateTopic, DeleteTopic, GetSubscriptionAttributes, GetTopicAttributes, ListTagsForResource, SetSubscriptionAttributes, SetTopicAttributes, Subscribe, TagResource, Unsubscribe, UntagResource. Configuration and authorization state must be persisted; waiting for a human or endpoint confirmation is not promised within this request-response budget. Bounded topic and subscription pages: +20 ms monthly p99 (adapter-added). ListSubscriptions, ListSubscriptionsByTopic, ListTopics. Persistent catalog reads and token/result generation are included. Your signed agreement sets the terms for your deployment. A target does not add an operation or option that the compatibility tables mark unavailable.

​
What counts toward latency

For a request-response row, measure from the agreed ingress boundary to dispatch of the complete response. A row that explicitly names a first response chunk ends at that chunk; its number does not cover the rest of the stream. A long-poll row names the intentional wait and when adapter delay starts. Include parsing, authorization, admission, translation, serialization, adapter-owned storage and coordination, retries and response handling. Subtract only separately measured target-workload waits and external network segments allowed by the measurement rules. The adapter’s own response handling and dispatch remain covered. A database used for adapter metadata is still adapter work, even if a cloud provider hosts it. For concurrent calls, exclude the union of permitted wait intervals, not the sum of overlapping spans. Calculate each request’s adapter duration first, then the monthly p99. Do not subtract one service’s p99 from another’s. Known adapter timeouts are over-budget samples; failed or incomplete requests cannot disappear to improve the percentile. Missing measurements do not become zero latency. An SDK call span alone does not prove how much of its duration can be excluded.

​
Endpoint reachability and request failures

The endpoint-reachability target measures adapter health checks. Correctly forwarding a target quota or permission error is different from producing that error because the adapter sent the wrong request. Adapter-caused failures remain Tensor9’s responsibility even when the target is healthy. Correctness and data-integrity requirements apply to every supported request. Endpoint reachability does not measure request correctness or grant an allowance for incorrect results.

​
Scaling and target-service capacity

Bound subscriptions per topic, filter complexity, payload bytes and queued deliveries as well as publish rate. A publish with 20 subscribers creates up to 20 units of delivery work. Scale request handlers and delivery workers independently, and watch outbox age or broker backlog. Subscriber timeouts and provider quotas can limit delivery even when the adapter accepts new work quickly; the agreement must state a separate delivery/backlog objective if one is required. Tell Tensor9 the expected steady rate, bursts, concurrency, payload sizes and operation mix. Tensor9 sizes and scales the adapter for the agreed load; you choose and monitor the target service’s capacity with Tensor9’s help. A latency budget is not a requests-per-second rating. Larger requests and higher rates need explicit terms, not silent inheritance of a small-request SLA.

​
Data and behavior guarantees

A successful Publish must correspond to the documented durable acceptance point. Retries can duplicate downstream delivery, so consumers remain idempotent. Subscription activation must respect receiver consent and supported protocol rules. Filter policy, message attributes and FIFO behavior are limited by this target profile. Unsupported subscription or publish shapes are not implied by the numerical acceptance row.

​
Example and diagnosis

For topic order-created, publishing one 2 KiB event to ten subscriptions uses the single-publish row. The request can satisfy its SLA while an unreachable webhook accumulates retries; inspect delivery state separately. If adapter-owned outbox insertion is slow, that is fabric time rather than subscriber time and remains Tensor9’s responsibility. Use tensor9 explain and the documented explain headers to understand the selected adapter and its behavior. Correlate available request diagnostics with the target provider’s latency, throttling and capacity metrics. An explain report helps diagnose a request; it is not by itself a qualified SLA timing measurement. Share the operation, request shape, timestamps and request identifiers with support, with credentials and customer payloads removed.

​
SNS (Simple Notification Service) to PostgreSQL

Adapter-added latency. A + prefix marks time added by the adapter, not the total time your application waits. Total request time includes adapter processing, backend service time and network time. The backend provider’s SLA applies independently. Only separately measured target-service wait may be subtracted, and only when your agreement permits it; internal metadata, coordination and retry processing still count. For one request, 12 ms of adapter processing plus 90 ms waiting for the backend means at least 102 ms total. This example does not add monthly p99 values. What Tensor9 covers. These service levels cover the adapter between the origin API and target API. They do not replace the target provider’s SLA. How latency is counted. Adapter work, including metadata and coordination, counts. Only separately measured permitted target waits are excluded. Which terms apply. Your signed agreement names the covered operations, workload limits, remedies, and final service levels for your deployment.

​
What this adapter does

For postgres::1.0.0::database, Publish persists accepted notifications and per-subscription delivery work in a PostgreSQL outbox. Workers deliver later and record their result. The latency row ends at API acceptance, not successful receipt by every subscriber.

​
How the latency targets were chosen

Bounded notification acceptance: +20 ms monthly p99 (adapter-added). Publish. The PostgreSQL path persists per-subscriber delivery work; native brokers fan out differently. Adapter subscription/authorization/outbox work stays on the fabric clock. Bounded notification batch acceptance: +50 ms monthly p99 (adapter-added). PublishBatch. Batch decoding, target calls and delivery scheduling are bounded by both entries and fan-out. Topic and subscription configuration: +20 ms monthly p99 (adapter-added). CreateTopic, DeleteTopic, GetSubscriptionAttributes, GetTopicAttributes, ListTagsForResource, SetSubscriptionAttributes, SetTopicAttributes, Subscribe, TagResource, Unsubscribe, UntagResource. Configuration and authorization state must be persisted; waiting for a human or endpoint confirmation is not promised within this request-response budget. Bounded topic and subscription pages: +20 ms monthly p99 (adapter-added). ListSubscriptions, ListSubscriptionsByTopic, ListTopics. Persistent catalog reads and token/result generation are included. Your signed agreement sets the terms for your deployment. A target does not add an operation or option that the compatibility tables mark unavailable.

​
What counts toward latency

For a request-response row, measure from the agreed ingress boundary to dispatch of the complete response. A row that explicitly names a first response chunk ends at that chunk; its number does not cover the rest of the stream. A long-poll row names the intentional wait and when adapter delay starts. Include parsing, authorization, admission, translation, serialization, adapter-owned storage and coordination, retries and response handling. Subtract only separately measured target-workload waits and external network segments allowed by the measurement rules. The adapter’s own response handling and dispatch remain covered. A database used for adapter metadata is still adapter work, even if a cloud provider hosts it. For concurrent calls, exclude the union of permitted wait intervals, not the sum of overlapping spans. Calculate each request’s adapter duration first, then the monthly p99. Do not subtract one service’s p99 from another’s. Known adapter timeouts are over-budget samples; failed or incomplete requests cannot disappear to improve the percentile. Missing measurements do not become zero latency. An SDK call span alone does not prove how much of its duration can be excluded.

​
Endpoint reachability and request failures

The endpoint-reachability target measures adapter health checks. Correctly forwarding a target quota or permission error is different from producing that error because the adapter sent the wrong request. Adapter-caused failures remain Tensor9’s responsibility even when the target is healthy. Correctness and data-integrity requirements apply to every supported request. Endpoint reachability does not measure request correctness or grant an allowance for incorrect results.

​
Scaling and target-service capacity

Bound subscriptions per topic, filter complexity, payload bytes and queued deliveries as well as publish rate. A publish with 20 subscribers creates up to 20 units of delivery work. Scale request handlers and delivery workers independently, and watch outbox age or broker backlog. Subscriber timeouts and provider quotas can limit delivery even when the adapter accepts new work quickly; the agreement must state a separate delivery/backlog objective if one is required. Tell Tensor9 the expected steady rate, bursts, concurrency, payload sizes and operation mix. Tensor9 sizes and scales the adapter for the agreed load; you choose and monitor the target service’s capacity with Tensor9’s help. A latency budget is not a requests-per-second rating. Larger requests and higher rates need explicit terms, not silent inheritance of a small-request SLA.

​
Data and behavior guarantees

A successful Publish must correspond to the documented durable acceptance point. Retries can duplicate downstream delivery, so consumers remain idempotent. Subscription activation must respect receiver consent and supported protocol rules. Filter policy, message attributes and FIFO behavior are limited by this target profile. Unsupported subscription or publish shapes are not implied by the numerical acceptance row.

​
Example and diagnosis

For topic order-created, publishing one 2 KiB event to ten subscriptions uses the single-publish row. The request can satisfy its SLA while an unreachable webhook accumulates retries; inspect delivery state separately. If adapter-owned outbox insertion is slow, that is fabric time rather than subscriber time and remains Tensor9’s responsibility. Use tensor9 explain and the documented explain headers to understand the selected adapter and its behavior. Correlate available request diagnostics with the target provider’s latency, throttling and capacity metrics. An explain report helps diagnose a request; it is not by itself a qualified SLA timing measurement. Share the operation, request shape, timestamps and request identifiers with support, with credentials and customer payloads removed.

​
Debug this service

For requests through a service adapter, use Explain and request diagnostics to investigate the selected mapping. Follow the BYOC service-adapter debugging runbook to capture and interpret the diagnostic evidence. Service Catalog.