Date

Attendees

Discussion items

WhoNotes

Vadim Reutskiy

  • There was no spike in memory during testing of the fix for the synchronization after a big delay in block sync.

Mikhail Boldyrev

  • As expected, I am going to finish the environment variable configuration today in the evening.
    • The implementation will be reusable and configurable.

Andrei Lebedev

  • Will check the code from Mikhail tomorrow so we can merge it asap.
  • Yesterday finished the test run on Windows. There are some issues in the code.
    • There is some undefined behaviour. Hence, it is important to check the build with Microsoft compiler.
    • One issue is related to the Multihash code.
    • All the issue was fixed already, PR approved.
  • Now I need Mikhail to review 2 PRs: flat map and validation of the block store. Also, there is a PR for replacing Boost with std.
  • Regarding the RocksDB integration: I will continue working on it today. Will write the benchmark first for checking performance after the exchange.

Sara Garifullina

  • Andrey sent the data about configuration parameters. Will add the information to the documentation.
  • Will ask the typical hardware configuration from team leads.
  • I'll be on marketing call with HL
  • Regarding the "Thank you" letter for Indian student, we can do that.

Makoto Takemiya

  • After the 1.2.0 release, we need to concentrate on stability and performance.
  • We need to fix the Postgres bottleneck. Think about the optimization of the data model. Postgres is pretty fast in many cases, so we need to get familiar with techniques and best practices. Probably, in-memory databases will take even more development to be used and can provide not so many improvements.
    • Andrei Lebedev Regarding the Postgres: logs are saying that most likely the bottleneck is located in the relational constraints. There probably can be a way to cut the corner with the current situation and improve quickly. We can also disable some non needed features from Postgres (like multi-user access).
    • If we can make writing parallel it most probably will give us some improvements.
  • Having benchmarks is very important to have before starting the improvements.