Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- The durable-operation ceiling rose from 3,000 to 30,000 and checkpoint payloads from 256 KB to 1 MB, in AWS Regions where durable functions is available.
- Each step attempt, each wait-for-condition poll, and each parallel or map branch counts toward the operation budget.
- The current quotas table separately limits cumulative persisted payloads to 100 MB per execution. The 30,000-operation and 1 MB per-operation ceilings cannot be increased.
- The announcement describes a JavaScript/Python FileSystem serializer path for larger state. Higher limits alone do not establish lower total costs or an adaptation-free upgrade.
The change concerns operations and state within a durable execution
AWS announced on October 13 that the maximum durable operations per execution increased from 3,000 to 30,000, while the maximum checkpoint payload increased from 256 KB to 1 MB. These limits matter to multi-step workflows that save progress, wait for external results, retry work or coordinate branches.
Think of a checkpoint as the progress and state a workflow needs to retrieve when it resumes. A larger payload can hold a more complete event, model response or tool output. This change does not enlarge ordinary Lambda synchronous requests or allow one function invocation to compute continuously for a year.
Count step attempts, polling and branches first
The announcement says every durable operation counts, including each step attempt, every poll of a wait-for-condition, and each branch of a parallel or map operation. Estimate the operations produced by the actual workflow rather than treating the number of business records as the whole budget.
For a hypothetical workflow handling 5,000 records, each record might also involve retries, external-status polling and branches. Those 5,000 records therefore do not imply only 5,000 durable operations. Include these additional paths and leave room for extra attempts; this is design advice rather than a capacity test we have run.
A 1 MB payload still sits within a 100 MB cumulative ceiling
The current Lambda quotas table gives a maximum payload of 1 MB per durable operation, explicitly covering both checkpoint and callback payloads, and says it cannot be increased. The maximum of 30,000 operations per durable execution is also non-adjustable.
The table separately limits cumulative payloads persisted per execution to 100 MB, also non-adjustable. Multiplying 30,000 operations by 1 MB does not produce an allowable storage budget. Assess both individual payloads and cumulative writes. The 100 MB figure is a current documented constraint, rather than a change newly announced in this release.
Use the documented external-storage path for state above 1 MB
For state larger than 1 MB, the announcement says the JavaScript and Python SDKs include a FileSystem serializer. It writes the payload to an Amazon EFS or Amazon S3 Files mount and checkpoints a pointer instead. Separating the larger content from its reference does not remove the checkpoint limit.
Before choosing this path, check mounts, permissions and data lifecycle against the SDK and storage configuration being used. The announcement does not supply a minimum SDK version or a complete deployment procedure here. It also does not establish support for any arbitrary filesystem or automatic offloading of every oversized payload.
Check the Region and quota dimensions before changing a workflow
AWS says the new limits apply in all AWS Regions where Lambda durable functions is available. This does not mean the capability is available in every AWS Region. Concurrent executions, execution-start rates and checkpoint API request rates remain separate quota dimensions in the current table; the operation ceiling does not replace them.
As operational advice, have the workflow owner check the target Region, current operation counts and state sizes before a controlled change. Retain error handling, observability and rollback conditions, and start with a representative workflow containing no sensitive data. Higher ceilings do not guarantee an unchanged existing workflow, and the announcement does not describe the transition for every execution already in progress.
Revisit the cost estimate instead of treating higher limits as free
The announcement describes durable functions supporting long workflows and suspending execution during waits without compute charges. That specific statement does not make the whole workflow free. Evaluate step execution, durable state and external storage under the applicable service terms.
More operations and larger state can change an application’s usage profile. The announcement and quotas page do not provide a price table sufficient to calculate the total bill, announce a free upgrade or guarantee savings. Base the budget on actual billing items rather than only comparing the two new ceilings.
Sources
Frequently Asked Questions
Must an old workflow be redeployed for an in-progress execution to receive the limits?
The announcement confirms higher limits but does not specify activation times for every existing execution, a redeployment requirement or how in-progress state transitions. Avoid restarting a long workflow solely on that assumption. Confirm the target Region and current service behavior, then follow the application’s change and recovery process.
Does 1 MB now apply uniformly to all Lambda requests and responses?
No. This article concerns checkpoint and callback payloads for durable operations. The current quotas table separately lists 6 MB each for ordinary synchronous invocation requests and responses, 1 MB for asynchronous invocation payloads, and distinct streaming-response limits. Match the quota to the actual invocation type rather than applying the checkpoint ceiling to every interface.
Does a durable workflow lasting up to a year mean continuous function computation for a year?
No. A durable execution can coordinate a long workflow through checkpoints, waits and resumption. Individual function invocations still have their own timeout rules. The current quotas guide distinguishes ordinary invocations from certain Managed Instances invocations; workflow duration is not the continuous compute time of one invocation.