EMR Serverless Spark enforces the following resource quotas and performance limits per workspace. Review these limits before planning production workloads. To request a higher quota, submit a ticket.
| Limit | Value | Scope | Adjustable |
|---|---|---|---|
| Supported regions | Specific regions only. See Supported regions. | — | No |
| Resource quota (subscription) | 1,000 CUs (Compute Units) per workspace | Workspace | Yes — submit a ticket |
| Resource quota (pay-as-you-go) | 5,000 CUs per workspace | Workspace | Yes — submit a ticket |
| Shuffle data per job | 100 TB per Spark job | Per job | Yes — submit a ticket |
| Total shuffle data (all concurrent jobs) | 200 TB across all concurrent jobs in a workspace | Workspace | Yes — submit a ticket |
| Concurrent jobs | 1,000 jobs per workspace | Workspace | No |
Note
Shuffle data limits apply to data written during the shuffle phase of a Spark job. If either shuffle limit is exceeded — whether for a single job or across all concurrent jobs in the workspace — submit a ticket.
该文章对您有帮助吗？