[GH-ISSUE #184] `5afd9d1`: Force serial‑write & synchronous SHA256 on all snapshot packing paths, not only OSS publish path #59

New issue

Closed

opened 2026-10-29 14:18:58 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-29 14:18:58 +00:00

Owner

Originally created by @armBOY666 on GitHub (Aug 18, 2026).
Original GitHub issue: https://github.com/kvcache-ai/AgentENV/issues/184

Affected component or operation

Sandbox creation / boot

Regression summary

Commit 5afd9d1 (PR #177) wraps compact_to output with DigestTrackingFile + OrderedWriter. This change affects all snapshot packing paths, instead of only the OSS publish path it was intended to optimize.

Snapshot packing write concurrency drops from 32 to 1. Synchronous software SHA256 is added on hot‑path.
Snapshot creation latency rises significantly on aarch64 without SHA‑256 hardware acceleration.

Compared versions

Baseline: commit before 5afd9d1
Current: commit 5afd9d1
First known bad: 5afd9d1 (PR #177)
Local modifications: None

Host and workload

OS and kernel: Ubuntu 24.04 aarch64
Architecture and CPU:aarch64, no SHA‑256 hardware acceleration
Memory: 249GiB
Disk and filesystem: NVMe SSD, ext4

Benchmark command and procedure

Test workflow: Create local snapshot from running sandbox VM, no OSS upload.
1. Start AgentEnv service and boot sandbox VM(1‑4GiB guest memory).
2. Trigger snapshot packing API, repeat 20 times for latency statistics.
3. Warm page cache before each run.

Results

Runs:
Baseline:
- p50:
- p95:
- p99:
Current:
- p50:
- p95:
- p99:
Change:

Raw data

Profiles and diagnostics

No response

Measurement checklist

I ran the baseline and current versions on the same host or documented the differences.
I repeated the measurement and reported the sample count.
I controlled or documented cold/warm caches and other relevant variables.
I searched existing issues and did not find a duplicate.

Originally created by @armBOY666 on GitHub (Aug 18, 2026). Original GitHub issue: https://github.com/kvcache-ai/AgentENV/issues/184 ### Affected component or operation Sandbox creation / boot ### Regression summary Commit 5afd9d1 (PR #177) wraps compact_to output with DigestTrackingFile + OrderedWriter. This change affects all snapshot packing paths, instead of only the OSS publish path it was intended to optimize. Snapshot packing write concurrency drops from 32 to 1. Synchronous software SHA256 is added on hot‑path. Snapshot creation latency rises significantly on aarch64 without SHA‑256 hardware acceleration. ### Compared versions - Baseline: commit before 5afd9d1 - Current: commit 5afd9d1 - First known bad: 5afd9d1 (PR #177) - Local modifications: None ### Host and workload - OS and kernel: Ubuntu 24.04 aarch64 - Architecture and CPU:aarch64, no SHA‑256 hardware acceleration - Memory: 249GiB - Disk and filesystem: NVMe SSD, ext4 ### Benchmark command and procedure ```shell Test workflow: Create local snapshot from running sandbox VM, no OSS upload. 1. Start AgentEnv service and boot sandbox VM(1‑4GiB guest memory). 2. Trigger snapshot packing API, repeat 20 times for latency statistics. 3. Warm page cache before each run. ``` ### Results - Runs: - Baseline: - p50: - p95: - p99: - Current: - p50: - p95: - p99: - Change: ### Raw data ```text ``` ### Profiles and diagnostics _No response_ ### Measurement checklist - [x] I ran the baseline and current versions on the same host or documented the differences. - [x] I repeated the measurement and reported the sample count. - [x] I controlled or documented cold/warm caches and other relevant variables. - [x] I searched existing issues and did not find a duplicate.

bsakaguchi closed this issue

2026-10-29 14:18:58 +00:00

bsakaguchi commented

2026-10-29 14:18:59 +00:00

Author

Owner

@yingdi-shan commented on GitHub (Aug 18, 2026):

You can test the performance again now that PR#185 has been merged.

@yingdi-shan commented on GitHub (Aug 18, 2026): You can test the performance again now that PR#185 has been merged.

bsakaguchi referenced this issue

2026-10-29 14:20:39 +00:00

[PR #91] [MERGED] docs: add static multi-node deployment guide #173