-
Notifications
You must be signed in to change notification settings - Fork 301
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
Reading large data (binary, text) asynchronously is extremely slow #593
Comments
Cannot see much relation to #245 really. |
Here is the result of the test at NetCore & Netfx: BenchmarkDotNet=v0.12.1, OS=Windows 10.0.18362.657 (1903/May2019Update/19H1)
Intel Core i7-8665U CPU 1.90GHz (Coffee Lake), 1 CPU, 8 logical and 4 physical cores
[Host] : .NET Framework 4.8 (4.8.4121.0), X64 RyuJIT [AttachedDebugger]
Job-TUQHOY : .NET Framework 4.8 (4.8.4121.0), X64 RyuJIT
IterationCount=10 LaunchCount=1 WarmupCount=10
BenchmarkDotNet=v0.12.1, OS=Windows 10.0.18362.657 (1903/May2019Update/19H1)
Intel Core i7-8665U CPU 1.90GHz (Coffee Lake), 1 CPU, 8 logical and 4 physical cores
.NET Core SDK=3.1.202
[Host] : .NET Core 3.1.4 (CoreCLR 4.700.20.20201, CoreFX 4.700.20.22101), X64 RyuJIT
Job-LXIUHK : .NET Core 3.1.4 (CoreCLR 4.700.20.20201, CoreFX 4.700.20.22101), X64 RyuJIT
IterationCount=10 LaunchCount=1 WarmupCount=10
BenchmarkDotNet=v0.12.1, OS=Windows 10.0.18362.657 (1903/May2019Update/19H1)
Intel Core i7-8665U CPU 1.90GHz (Coffee Lake), 1 CPU, 8 logical and 4 physical cores
.NET Core SDK=3.1.202
[Host] : .NET Core 3.1.4 (CoreCLR 4.700.20.20201, CoreFX 4.700.20.22101), X64 RyuJIT
Job-OTAQAT : .NET Core 3.1.4 (CoreCLR 4.700.20.20201, CoreFX 4.700.20.22101), X64 RyuJIT
IterationCount=10 LaunchCount=1 WarmupCount=10
|
Working from memory here. Reading a MAX defined field will mean using chunked returns and the complete length of the field may not be known ahead of time so the optimization I put in place for reading won't help. If that's right then you'll see the same horrific escalating performance with increased data sizes as the reader continuously allocates and releases the temporary buffers, those GC numbers seem to support that. It might be possible to do adapt the cached buffer we've got for multipacket known lengths to work like a list and auto expand as we go. What I can't do is prevent the repeated rescanning of the async packet queue. |
I've had a look into the profiles and I was wrong. The optimization from #245 is being used because it's reading as PLP data. The excessive memory usage is caused mostly by input packet buffers on the TDS side being lost into the snapshot packet chain. The time escalation is cause by the snapshot chain repeatedly being replayed. The only currently available mitigations are to increase the packet size to as close to 16K as you can get or to use a stream reader mode which should work on a per-packet basis and not buffer the entire contents (I'm a bit hazy on that for async, it probably needs checking). The long term fix would be to rewrite async reads to avoid replaying the entire packet chain on every receipt. This is akin to brain surgery. I've looked into this while reviewing other performance issues around async and it isn't something to attempt lightly and if I managed it solo and dropped a PR for it the team they may all justifiably quit in protest. The |
Oh, and this in System.Data.Common causes about 1/3 of the allocations and really isn't helpful. We already use type workarounds with an overlay type to access the single value field directly to create SqlBinary instances so we could do the same to get it out but we'd then be handing the user a direct reference to a mutable array instead of a copy an I've no doubt that some idiot somewhere in the world is relying on the copy semantic to make their code work. |
You're referring to the fact that the binary data is copied out into an array that's given to the user, right? If so, isn't that the same between sync and async? In any case, that's indeed a lot of allocations, but the alternative would be to return memory that's owned by SqlClient, right? If I'm understanding things right, that would mean that if the user reads the next row, the reference they get would suddenly start pointing to a totally different value... I did think about this at some point as an advanced/high-perf/"unsafe" API in dotnet/runtime#24899 - but that's tricky because of the lifetime issue, and also because the entire binary data isn't necessarily in memory if CommandBehavior.Sequential is used (as it should if values are big). Anyway, all that stuff is less important than making async at least perform similar to sync... |
Yes, and yes.
Each row of data is stored in an array of SqlBuffer instances stored in the reader, The SqlBuffer type is a discriminated union and it'll hold a reference to an SqlBinary. When you move row or move field in sequential mode the buffer instance is dropped to the GC. The backing byte[] array inside SqlBinary is not re-used. If we returned the array directly it would only have an effect on multiple reads of the same field in the same row and the only way to observe a change of behaviour would be to change the contents of the array and then re-request it from the reader. It's a pretty clear optimization that I'd like to be able to make but I'm pretty sure I can't because someone will have relied on it returning a new array each time. As you suggest the ideal approach would be to use a ReadOnlyMemory then it'd all be safe but that'll require a runtime api change and for all providers to consume it.
The entire array is skipped in Sequential mode if the field is not requested, as you say.
I can sort of understand how it might be done but it's daunting and difficult. |
Wouldn't it be possible for a provider to simply support
I can certainly understand (and appreciate) that, and I'm definitely not suggesting anyone do anything specific (was just reporting the problem). Along with #547 and possibly other async-related issues, it seems that at least documenting these limitations could help users avoid pitfalls. BTW do you think there are any workarounds to this, besides switching to sync? |
I hadn't considered that, nice idea. Currently the only workaround I can think of is to use a stream read overload because I think those drive the parser per-packet directly so you don't have to replay the packet queue, that'll need verifying. |
That at least sounds like a good workaround if it works! Definitely worth benchmarking that. |
Unless I've got something very wrong it won't work because there's a bug. using var conn = new SqlConnection(ConnectionString);
using var cmd = new SqlCommand("SELECT foo FROM data", conn);
await conn.OpenAsync();
using var reader = await cmd.ExecuteReaderAsync(
System.Data.CommandBehavior.SequentialAccess
);
await reader.ReadAsync();
using var stream = reader.GetStream(0);
using var memory = new MemoryStream(16 * 1024);
await stream.CopyToAsync(memory);
return (int)memory.Length; This should work but freezes in a task wait after a single read cycle, the second read never completed. IF you change it to standard mode it'll work but it does so by fetching the entire field and giving you a reader over the byte[]. So no workaround, this needs fixing if the team agree that it's a bug. |
Thanks for testing this... Definitely looks like a bug to me.
Is this at least fast - in the sense that it resembles the sync perf rather than the async? If so it's at least a workaround, even if not ideal... |
So no. The only way to get the speed benefit Is to put the reader in sequential mode. SqlClient/src/Microsoft.Data.SqlClient/netcore/src/Microsoft/Data/SqlClient/SqlDataReader.cs Line 1557 in f2fd632
There's also an interesting comment elsewhere in the file that says streaming isn't supported when using encryption. I think that makes streaming unusable for a generic consumer like EFCore because you'll have to special case it per driver. We need to fix the sequential stream issue and then we need to rewrite async reads. |
Thanks for the details @Wraith2. |
With the fix in #603 the numbers are better:
|
That looks quite good, at the very least you can point people to a workaround. |
@SeanFeldman Is your app running on Azure App Service? It looks like it could be SNAT exhaustion. The maximum SNAT connection limit depends on the VM size. In our case, it was 128. We resolved the issue by integrating our App Service into a VNet. @Wraith2 Great job on async-cm32. We had numerous issues and downtimes with our app, and we had to switch to synchronous. Sinc 5.2.0 we'Re back on async. It's working much better now. Are there any improvements expected in this area for version 6? |
Watch Wraith's #2714 PR. Just trying to work out the bugs over there. There seem to be some edge case failures that are difficult to pin down. |
@inloox-dev it doesn't. It's doesn't. However, the issue is not the number of connections. Notice that the connections went down when we switched from async to sync API. |
Finally it seems to bubble up: https://www.youtube.com/watch?v=7iuUh0BIAuE |
We're in the process of fixing the issue. What more attention do you want? |
@Wraith2 I'm sorry, I'm not complaining about anybody who helps with this issue. I am grateful that with the modifications you made sync and async are getting more equal in terms of performance. However I had a number of jobs and somehow it is a constant struggle convincing people to not do async blob related work. The main thing is, MSSql is a Microsoft product, the .Net stack is driven by Microsoft, but somehow there exists a huge performance issue that takes years to plan/resolve... I think there is an issue with not recognizing the impact of this issue. That youtube post at least gives me something where I can redirect colleagues to. |
I could write up a blog post with the history and details if you want. |
I'd rather you just keep working on the (super duper amazing) pull's to fix the perf. 😄 |
What's the consensus around backporting cm32 to v5.1? It feels like a lot of the times this issue is re-raised come from using v5.1 because it's the most recent LTS release (either directly, or via EF Core.) Even assuming that v6.0 will be an LTS release, there's still another year of support left for v5.1. I'm happy to perform that backport, but I can see a chain of nine PRs which look like they contributed to (or were prerequisites of) the v5.2 fix. That's a larger set of changes than v5.1's been serviced with before, and I'm conscious that soon the situation might change and people could update to v6.0 without stepping off the LTS train. The PRs I'm looking at are #1866; #1544; #2121; #2122; #2126; #2132; #2144; #2157; #2164. Would this safely cover cm32 and its prerequisites @Wraith2 ? |
Don't do it. Users expect 5.1.x to be stable. Spend effort on 6.0 and 6.1 / 7.0 instead. Users can easily opt in to use 5.2.2 or 6.0 |
What users need is a stable version with the cm32 changes and 6 seems like the best way to do that. Unless there is some reason user may not be able to migrate from 5.1 to 6.0 I would suggest letting the changes go stable with 6.0. |
That makes sense, thanks. I can't think of any supportable speedbumps between 5.1 and 6.0, and an upgrade between two LTS versions seems like a fair enough way to get the changes... |
The proposed fix for this is #3161 and should be included and enabled by default in preview 7. |
… jako workaround dotnet/SqlClient#593 + publikace
@mdaigle Could this be closed now? |
Preview 7 isn't released and there is no stable build with the fix so it makes sense to me to keep it open for a while. |
Will that make it into EF Core 10 (LTS)? |
@janseris if it is not included with EF Core 10 you can just add an explicit reference |
is there an ETA on the V7 release date? |
@SimonCropp Preview 1 is just around the corner. |
@ErikEJ thanks |
The PR is currently assigned to this milestone https://github.com/dotnet/SqlClient/milestone/65 which has no due-data. It was prevously in the 7.0 milestone which had 2025-03-31 due date. |
Looks like 7.0 efforts were renamed to 6.1... |
Reading a 10MB
VARBINARY(MAX)
value asynchronously takes around 5 seconds my machine, while doing the same synchronously takes around 20ms. Increasing the data size to 20MB increases the running time to around 52 seconds.These numbers were with Microsoft.Data.SqlClient 1.1.3, but 2.0.0-preview4.20142.4 has the same issue (it actually seems slightly slower). Note that I'm not posting final BDN results because the benchmark ran extremely slowly.
Note that this looks very similar to #245, but that issue was fixed for 1.1.0. The difference may be that here we're writing binary data - or possibly the bigger size.
Benchmark code
Originally raised in dotnet/efcore#21147
The text was updated successfully, but these errors were encountered: