Meeting notes for 11/6/2017

[divega]

Attendees: @ajcvickers, @DamianEdwards, @geleems, @divega, @sebastienros

• We have a repo! @sebastienros is going to initialize it with the console benchmark code an .gitignore file (we may want to take the changes in @damageboy's fork, as long as the tricks used there may have an impact on any ADO.NET provider).
• On the Npgsql front @roji and @damanageboy have been getting some good results in their benchmarks and at this point some numbers look quite similar to pgbench! @sebastienros is going to try their code in our test environment and do some profiling. We want to understand better the impact of reusing the user-facing connection and command objects. Is this a reason to makeover the ADO.NET API (Idea: fast data access APIs from the ground up  #4) or can we make the current abstractions faster transparently?
• @geleems is going to help us find a good baseline for SQL Server Perf baseline: Figure what to use for SQL Server #2
• @geleems also reports that they are making good progress on the networking perf issue SqlClient with local Linux. He will try to repro the remote case to see if the fix also has an impact there
• @divega will create an issue for the connection pool depletion issue that @brockallen has described with async. @sebastienros believes he has a repro already.
• @divega to create an issue for MySQL testing

[damageboy]

Hi, I've pushed my (latest) forked benchmarks to:

https://github.com/damageboy/benchmarkdb