Re: repospanner and our Ansible repo
On Tue, Sep 17, 2019 at 6:47 PM Randy Barlow
<bowlofeggs(a)fedoraproject.org> wrote:
stickster asked me today how these numbers would compare to
Git{Hub,Lab}. I did a bit of testing with GitLab just now.
Note that this isn't a particularly apples to apples test, because my
repospanner nodes were on the same virtual host, and my git client was
on a 1 Gbps LAN with them. My GitLab test results are from my house,
where I only have a 60x6 Mbps connection to the Internet, and of
course, higher latency.
I considered testing from batcave01 to get higher bandwidth, but I
didn't want to try to figure out a safe way to use my GitLab
credentials on a shared server and I didn't want to make a throw away
account just to test this.
On Mon, 2019-09-16 at 18:51 -0400, Randy Barlow wrote:
> I pushed the Ansible repository into it. This took a very long time:
> 298m2.157s!
This took 6m44.705s to get to GitLab. However, since I only have 6 Mbps
outbound and the repository is 268.43 MiB, I calculate that almost all
of this time was just due to waiting on my outbound pipe.
> The next test was to see how long it takes to clone our repo. I did
> this on another machine on the same LAN (so again, ideal network
> latency) and it took 2m27.433s.
This took 0m40.359s, and again, almost all of the time was just due to
how long it would take to send that much data over a 60 Mpbs link.
> Next, I made a small commit (just added/deleted some lines) and
> pushed
> it into the cluster. This went reasonably quick at 0.366s, which I
> think we would be OK with.
This took 1.443s to GitLab, and I bet most of it was just latency/round
trip crypto setup time.
> The last test I performed was to see how quickly another checkout
> could
> pull that commit, and this was again a speed I might consider to be a
> bit slow at 4.931s, especially considering that the commit was small
> and was only one.
This took 0m1.523s to GitLab, and I bet most of it was just
latency/round trip crypto setup time.
I don't expect it would be useful to perform this test with GitHub
since I'd expect essentially the same results (bottlenecked on my home
internet connection).
Out of curiosity, do we know where the bottlenecks are in repoSpanner?
In theory, the architecture of repoSpanner isn't supposed to be too
different from gitaly, so I'm curious where we're falling down.
--
真実はいつも一つ！/ Always, there's only one truth!