mirror of
https://github.com/vale981/ray
synced 2025-03-06 02:21:39 -05:00
1483c4553c
m5.16xlarge instances have 64 CPU and 256GB memory, which are overkill for scheduling tests that do not have a lot of computations. Use smaller instance m5.4xlarge to save cost and make allocating instances easier. |
||
|---|---|---|
| .. | ||
| distributed | ||
| object_store | ||
| single_node | ||
| app_config.yaml | ||
| distributed.yaml | ||
| distributed_smoke_test.yaml | ||
| many_nodes.yaml | ||
| object_store.yaml | ||
| README.md | ||
| scheduling.yaml | ||
| single_node.yaml | ||
Ray Scalability Envelope
Distributed Benchmarks
All distributed tests are run on 64 nodes with 64 cores/node. Maximum number of nodes is achieved by adding 4 core nodes.
| Dimension | Quantity |
|---|---|
| # nodes in cluster (with trivial task workload) | 250+ |
| # actors in cluster (with trivial workload) | 10k+ |
| # simultaneously running tasks | 10k+ |
| # simultaneously running placement groups | 1k+ |
Object Store Benchmarks
| Dimension | Quantity |
|---|---|
| 1 GiB object broadcast (# of nodes) | 50+ |
Single Node Benchmarks.
All single node benchmarks are run on a single m4.16xlarge.
| Dimension | Quantity |
|---|---|
| # of object arguments to a single task | 10000+ |
| # of objects returned from a single task | 3000+ |
# of plasma objects in a single ray.get call |
10000+ |
| # of tasks queued on a single node | 1,000,000+ |
Maximum ray.get numpy object size |
100GiB+ |