Optimize MySQL Bootstrap Chunking by Switching from MD5 to CRC32

[akshayrai]

Summary

• This PR introduces a performance optimization. The existing MD5-based partitioning strategy caused chunking and bootstrapping of large tables to be extremely slow and prone to connection timeouts.
• The partitioning strategy has been updated to use CRC32 instead of MD5 for MySQL bootstrapping.
• Benchmark results show a ~10x performance improvement together with few other optimizations as well (count star query from task 0 only, setting innodb_parallel_read_threads). For example, on a 0.64 GB table:
  • The previous MD5-based bootstrap took approximately 2 hours 20 minutes at a throughput of ~3 MB/s.
  • With CRC32, the bootstrap completed in ~20 minutes at ~30 MB/s throughput.
• Further, CRC32() is significantly faster than MD5() in MySQL. For large tables, MD5 can easily become a CPU bottleneck.

Release notes

• Need to ensure this is deployed when no existing bootstraps are running. Otherwise, it can lead to inconsistency in data.

Testing Done

• Verified on a 0.64 GB MySQL table in the EI environment.
• Detailed benchmark results are documented internally.

[kanishkjaiswal2015]

LGTM

[harshcum]

LGTM