 |
|
Say my performance is approx 50ms per request (request initiated from one port on local machine to another port SAME machine), each request contains about ten rows of data (2 columns only).
What's your thought on this performance level?
(I know some algo trading execution say from London to NY can be done in say 50-60ms ...)
dev
|
|
|
|
|
Is this a C# issue?
One of these days I'm going to think of a really clever signature.
|
|
|
|
|
it's not an issue, just testing water see how people feels about 50ms per call if it's considered slow/acceptable/fast
dev
|
|
|
|
|
|
i did thought of that but it's a programming question - question for an opinion on a general IPC performance - in particular, for those .NET developers who has tried to push performance on IPC/socket
dev
|
|
|
|
|
devvvy wrote: question for an opinion on a general IPC performance
Which has nothing to do with C#.
One of these days I'm going to think of a really clever signature.
|
|
|
|
|
dude, there isn't a forum by the name "Socket" or ".NET IPC", so this i figure is the closest.
dev
|
|
|
|
|
You could have chosen this[^] forum, or this[^] one. Richard is right - as your question doesn't have any code element, i.e. you haven't actually asked why your code is slow, then it's not for this forum. Your question is more general and, dare I say it, architectural.
|
|
|
|
|
i see, you are correct Peter, in absence of C# code, this is an architectural issue
dev
|
|
|
|
|
And neither "Socket" nor "NET.IPC" are specific to C#.
One of these days I'm going to think of a really clever signature.
|
|
|
|
|
The question isn't really valid. What is acceptable for your app depends on your requirements, not some general opinion.
|
|
|
|
|
no, if you wire one simple request with a small payload and it takes one second, we know "Something is Wrong", regardless of your application.
In fact I think fact our code now do one request in 50ms local-to-local tells me it's not tuned enough
dev
|
|
|
|
|
As others have said, you're asking for other peoples opinion about what is acceptable performance for a single request. We REALLY can't answer that. Only your requirements can.
Then, you toss all of that discussion and replace other opinions with what YOUR opinion is and that the performance is unacceptable. You answered your own question!
Now, if you're asking about how to improvie your code performance, you can restate your question and supply the relevant code snippets to show what you've got.
|
|
|
|
|
|
Reported as abusive.
Have a nice life.
|
|
|
|
|
Dave Kreskowiak wrote: Have a nice life.
Go get one buddy
dev
|
|
|
|
|
devvvy wrote:
What's your thought on this performance level?
What is the performance measurement if you open 50 threads at the same time and run for 5 minutes and average the result?
|
|
|
|
|
Yes threaded but test done single threaded manner - this seems to tell me our code can be further tuned, guess to <10ms for local-to-local (simple request not a big payload)
dev
|
|
|
|
|
devvvy wrote: Yes threaded but test done single threaded manner
Which is why I suggested that you do a threaded test.
|
|
|
|
|
btw, removed logging one place it's not 6ms per call! Hurrah!
dev
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Rant 
Admin 
Submit an article or tip
Post your Blog
