Martin Dimitrov
Martin Dimitrov
Data is inserted into the time-series compressed by default. During certain use-cases, such as computing filtering operations (e.g. Max, min, etc.), decompressing the chunks can be a significant bottleneck -...
There is a potential memory corruption bug when allocating a new chunk, which is not aligned to 8 bytes. The alignment is checked in the trimChunk function, but not in...
ERROR {'github_org': 'redis', 'github_repo': 'redis', 'github_org/github_repo': 'redis/redis', 'deployment_type': 'oss-standalone', 'deployment_name': 'oss-standalone', 'triggering_env': 'ci', 'build_variant': 'gcc:8.5.0-amd64-debian-buster-default', 'running_platform': 'intel64-ubuntu22.04-redis-clx1', 'compiler': 'gcc', 'compiler_version': '8.5.0', 'os': 'debian-buster', 'arch': 'amd64', 'branch': 'unstable', 'deployment_name+branch': 'oss-standalone unstable',...
2023-08-30 09:48:55 CRITICAL Reading results json from /root/tmpa38dek3d/oss-standalone-2023-08-30-16-45-54-e792653753dc62b5a00822121e585511542a024b-memtier_benchmark-1Mkeys-load-stream-5-fields-with-100B-values-pipeline-10.json
Filipe, I wanted to simplify the code a bit. Lets discuss this next time? Some functions were way too long and hard to efficiently follow. I made main() and blocking_read()...
The load list commands: memtier_benchmark-1Mkeys-load-list-with-1KiB-values memtier_benchmark-1Mkeys-load-list-with-100B-values memtier_benchmark-1Mkeys-load-list-with-10B-values keep growing in memory capacity. If multiple memtier runs are specified, this would result in consuming more memory than specified.
In some test cases we are doing -c 50, but we are requesting only 2 CPUs in the resources section. Is that OK? For example, see: memtier_benchmark-1Mkeys-string-get-100B.yml Should that stay...
Still in progress...
Since we are comparing the value to a "window" of previous values during compression, I believe we may benefit from vectorizing the code - compare the value to multiple values...
(IN PROGRESS) (TODO - will test it out) Finding the distance at the bottom of the function should also be vectorizable. The right way to do this, is to first...