@sebastienros also gave us on update on the addition of MongoDB scenarios on https://github.com/aspnet/benchmarks/. The results are inline with our relational database results. We didn't talk during the meeting about any follow up actions, but it seems we should share this with MongoDB team and see if they have any insights.
Brief update from @anpete: making progress on adding an ADO.NET facade on top of the Peregrine PoC, with the goal of measuring the overhead of ADO.NET abstractions/usage
We talked briefly about what else to add to our test suites. Currently we have a small set of repeatable benchmarks and we are using assumptions to extrapolate how our results compare with other results obtained from different environment. The more we add to our repeatable results the more confident we can be. Several dimensions we talked about (@divega created #20 to track this):
Include Redis case on TechEmpower and mini-benchmark
Add MongoDB on mini-benchmark (it was added already on TechEmpower)
Ability to run on physical hardware. All our results are currently "cloud"
SqlClient on our mini-benchmark
Non-raw, e.g. EF Core and Dapper on our mini-benchmark
Non-.NET tests that we can run alongside our mini-benchmarks to make sure we are comparing actual results on same hardware
Also talked about understanding the impact of batching and the overhead of detecting/checking rows affected in batching scenarios. I have updated the issue with more information.
We only made a bit of progress on #2 (the perf baseline for SQL Server). Still waiting to get more information from the SQL Server team. We are thinking about trying OLE DB.
Attendees: @ajcvickers @anpete @divega @geleems @sebastienros
@sebastienros gave us an update on the progress on Npgsql at https://github.com/aspnet/DataAccessPerformance/issues/5#issuecomment-349080515. The latest experiment shows very promising results.
@sebastienros also gave us on update on the addition of MongoDB scenarios on https://github.com/aspnet/benchmarks/. The results are inline with our relational database results. We didn't talk during the meeting about any follow up actions, but it seems we should share this with MongoDB team and see if they have any insights.
Brief update from @anpete: making progress on adding an ADO.NET facade on top of the Peregrine PoC, with the goal of measuring the overhead of ADO.NET abstractions/usage
We talked briefly about what else to add to our test suites. Currently we have a small set of repeatable benchmarks and we are using assumptions to extrapolate how our results compare with other results obtained from different environment. The more we add to our repeatable results the more confident we can be. Several dimensions we talked about (@divega created #20 to track this):
Also talked about understanding the impact of batching and the overhead of detecting/checking rows affected in batching scenarios. I have updated the issue with more information.
We only made a bit of progress on #2 (the perf baseline for SQL Server). Still waiting to get more information from the SQL Server team. We are thinking about trying OLE DB.