facebook / facebook/rocksdb

Throughput values too low

Open
#8,231 1 comment 0 reactions 0 assignees View on GitHub
question up-for-grabs
Dominant language
C++
Stars
32.1k
Forks
6.9k
Avg merge
32m
Merged PRs (30d)
1

Description

When I try to run the mixgraph benchmark using db_bench, I get unreasonably low throughput values. Normally when running other benchmarks I would get around 100k ops/sec but here I get only around 900-1500 ops/sec. This happens after I run fill the database. I used inspiration from the command in the appendix of [this paper](https://arxiv.org/pdf/2103.16267.pdf) and [this page](https://www.usenix.org/system/files/fast20-cao_zhichao.pdf) from the wiki, which are both in turn based on [this paper](https://www.usenix.org/system/files/fast20-cao_zhichao.pdf).

### Expected behavior
Higher throughput values. I tried playing around with the parameters but it doesn't do much.

### Actual behavior
I get very low throughput values.

### Steps to reproduce the behavior
I am using RocksDB version 6.19.0, Linux x86 Ubuntu version 20.04, and gcc compiler version 9.3.0. I am using the release build. I am also using 4 * Intel(R) Xeon(R) CPU @ 2.00GHz, 16GB RAM and a 500 GB SSD.

The command I used to fill the database:
```
\/db_bench --benchmarks="fillrandom" -num=50000000 -perf_level=3 -use_direct_io_for_flush_and_compaction=true -use_direct_reads=true -cache_size=268435456 -key_size=48 -value_size=43
```

The command I used to run the performance check:
```
\/db_bench --benchmarks="mixgraph" --use_existing_db -use_direct_io_for_flush_and_compaction=true -use_direct_reads=true -cache_size=268435456 -keyrange_dist_a=14.18 -keyrange_dist_b=-2.917 -keyrange_dist_c=0.0164 -keyrange_dist_d=-0.08082 -keyrange_num=30 -value_k=0.2615 -value_sigma=25.45 -iter_k=2.517 -iter_sigma=14.236 -mix_get_ratio=0.85 -mix_put_ratio=0.14 -mix_seek_ratio=0.01 -sine_mix_rate_interval_milliseconds=5000 -sine_a=1000 -sine_b=0.00000073 -sine_d=4500 --perf_level=2 -reads=420000000 -num=50000000 -key_size=48 --duration=300
```

Changing the number of reads, increasing sine_a, increasing since_b, increasing sine_d did not cause any cause any particularly noticeable difference.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.