Concordium / Concordium/concordium-node
Slowdown when increasing capacity
- Dominant language
- Haskell
- Stars
- 108
- Forks
- 24
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 9
Description
**Task description**
The `execute-chain` binary executes blocks slower when given more resources with `-N?` RTS flag.
This is observed on Intel Core i9-9900K CPU.
When running with `-N1` a specific block import file, consisting of around 20 full blocks (~6000 simple transfers per block) takes about 1m30s to execute. During execution the process uses about 100% CPU, which is expected.
When running without any RTS flags (thus defaulting to `-N16`) the process takes around 2m30s to terminate and during execution uses around 600-700% CPU.
The execute-chain binary itself is not directly involved in the running node, but the fact that such a significant slowdown happens, as well as the 6 fold increase in CPU likely indicates an underlying issue that affects the node as well.
Profiling does not reveal anything useful, at least to the extend I had done it, but it does make execution a lot slower (around 20min to execute the same program) so this is not very surprising.
This task is to investigate why this is happening and whether it affects the running node as well.
Contributor guide
Assessment
This issue has not been assessed yet.