Don't use cstruct - use string/bytes! #223
Closed
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
Co-authored-by: Hannes Mehnert <[email protected]> Co-authored-by: Reynir Björnsson <[email protected]>
|
With this branch and no-cstruct opam repository I see a significant speedup: cc628d5 (this branch + bench marking): fe92b0a (main): There is as well a higher number of reported allocations. It is unclear how much this is due to cstructs being underreported and how much it is an increase in allocations. |
|
After @hannesm fixed the bug in ocaml-tls#no-cstruct that caused invalid MAC I can now test against a VM on my machine running OpenVPN.?? and iperf3: Main branchThis branch & no-cstruct repo (with TLS fix)There is a performance improvement in sending (upload) and a performance regression in receiving (download). I think this is promising as the cstruct->string change in Miragevpn was somewhat naïve and there is room for some low-hanging improvements. It is also interesting to me the asymmetry goes from 1:2 to 2:3 in upload/download speeds. Note: the numbers are collected on a different machine than in #206 so they are unfortunately not comparable. If desired I can measure on that machine (which will require compiling for a different debian release). |
|
Below is a flamegraph of miragevpn while I run And while running |
When decoding operation return offset and length of the package. We avoid copying the packet only to copy it again without the first byte.
|
I pushed some commits that reduce the number of allocations and copying and I can observe a small improvement in the bechamel benchmark. I was struggling to observe improvements in the openvpn/iperf3 VM benchmark, and so I tried running the openvpn/iperf3 benchmark against |
|
For comparison here is 4 runs of iperf3 with this branch: |
Unfortunately, appending the empty string to another string results in a potentially expensive copy.
|
I discovered that This is a flamegraph before that commit. As you can see in Here is the flamegraph with that commit. The So it seems for TCP, when running iperf3 load at least, we often have lingering bytes, Maybe it is worth looking into a better suited data structure then. |
|
Oops I inadvertently pushed the octets commit |
Merged
|
Superceded by #279 |
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.
No description provided.