Meeting 10 - December 3rd, 2025¶

What Got Done¶

Thanksgiving Break Side Quest¶

• I recreated tms_min to use PostgreSQL instead of SQLite
• This significantly improved performance

• In these test, SQLite would start to have fails at 25 users doing writeheavy

• I also got postgres to fail in a test not shown that was massive

  • 250 users each doing 2500 iterations of both readheavy and writeheavy

Read Through Rich's Write Up¶

• Both good and bad news from this
• Good news: we had very similar findings
• He even had the same suspicion of SQLite being the problem
  • He proved SQL was not the problem by returning a dummy response instead of interacting with SQL
  • I proved SQL was not the problem with tms_min
  • We can be confident that SQL is not the problem
• Bad news: Still in the same situation, he gives suggestions that the problem is likely Poem/Tokio/TLS software stack

Starting stripping things out of tms_server¶

• I started taking everything out of tms_server so that only has the getclient functionality
• I think I'm very close to getting this to work

Next Steps¶

• Get tms_server_getclient_only working
• Run test to see if this still fails