Add "batch" requests submission to io-tester #2402
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
nit, s/smother/smoother/, because smother has some bad meaning. but this change is an improvement. |
avikivity
reviewed
Aug 25, 2024
| @@ -354,21 +354,11 @@ class class_data { | |||
| return do_until([this, stop] { return std::chrono::steady_clock::now() > stop || requests() > limit(); }, [this, buf, stop, pause, &intent, &in_flight] () mutable { | |||
| auto start = std::chrono::steady_clock::now(); | |||
| in_flight++; | |||
| return issue_request(buf, &intent).then_wrapped([this, start, pause, stop, &in_flight] (auto size_f) { | |||
There was a problem hiding this comment.
this patch splits it into different continuations to make further
patching simpler.
Consider also using coroutines.
There was a problem hiding this comment.
It's a separate branch in my queue, yes
avikivity
reviewed
Aug 25, 2024
|
Looks good. |
The rate-based submission of requests keeps track of the amount of outstanding requests with the in_flight counter. Also, requests in this job are cancellable and the cancellation exception should be properly handled. Currently all this logic sits in a single then-wrapped continuation, this patch splits it into different continuations to make further patching simpler. Signed-off-by: Pavel Emelyanov <[email protected]>
Both rate-based and continuous requests submissions carry rps and parallelism parameters all over their continuations, while both are available via this-> methods. Signed-off-by: Pavel Emelyanov <[email protected]>
The rate-based job can come with the parallelism parameter, so several fibers start and issue requests at the same rate. At the end all fibers cancel the only intent and wait for requests for finish. Despite cancellation is re-entrable, it's simpler and cleaner to cancel and wait only once. Signed-off-by: Pavel Emelyanov <[email protected]>
There's duplicating issue_request().then([] { add_result() }) chain in
rate-based and continuous submission helpers, generalize both. This
makes it less code and next patching simpler.
Signed-off-by: Pavel Emelyanov <[email protected]>
This is to make it possible to better simulate read-life workloads. Currently there are two ways to submit requests -- continuously via several fibers and at a given rate (the rate can be uniform, or can be disturned with random delays). In real life, however, there's often another pattern, namely -- requests come in batches at some rate. Say, once every 200 milliseconds 10 requests are queued shortly one after another. In fact, it's even more convoluted, once every 200 ms there comes a _sequence_ of, say, 10 requests almost one-after-another, but submitting them in a batch gives good approximation. This patch adds "batch" parameter to io-tester conf. Signed-off-by: Pavel Emelyanov <[email protected]>
xemul
force-pushed
the
br-io-tester-add-rps-batch
branch
from
e280656
to
83d431f
Compare
|
upd:
|
|
@avikivity , please consider merging |
3 similar comments
|
@avikivity , please consider merging |
|
@avikivity , please consider merging |
|
@avikivity , please consider merging |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Currently io-tester supports two kinds of IO workloads -- continuous and rated. In the former case requests come one after another from several fibers, in the latter case fibers submit requests at a given rate one at a time.
In real life the incoming requests pattern is not that simple. A nice approach to pretty common pattern can be to submit requests in batches at some rate. Say, once every 200 milliseconds 10 requests are queued at once.
Of course, in reality it's even more convoluted, e.g. once every 200 ms there can comes a sequence of, say, 10 requests with some small pauses in between, but submitting them in a one-time batch gives good approximation.
Also this PR simplifies the continuous and rate-based submission code to make the new batch parameter apply smoother.