On Wed, Jun 15, 2011 at 1:02 AM, Mark Phippard <markphip_at_gmail.com> wrote:
> On Tue, Jun 14, 2011 at 6:33 PM, Johan Corveleyn <jcorvel_at_gmail.com> wrote:
>> On Tue, Jun 14, 2011 at 11:56 PM, Mark Phippard <markphip_at_gmail.com> wrote:
>>> On Tue, Jun 14, 2011 at 5:48 PM, Johan Corveleyn <jcorvel_at_gmail.com> wrote:
>>>>
>>>> Also, has anyone tested this on an NFS-working copy? Or CIFS?
>>>
>>> AFAIK, no one is doing any performance testing on NFS or CIFS. I have
>>> repeatedly invited users to run the benchmarks I wrote in this
>>> configuration but no one has bothered. This leads me to believe it is
>>> not important to people.
>>
>> Maybe not important enough to the few people who build trunk (or
>> alpha-1) themselves (or rather: not important enough relative to the
>> required effort at this time). But that doesn't mean it isn't
>> important to a lot of users (okay, it will be a minority, but still).
>
> I do not doubt it is important to some people but we cannot hold the
> release forever. If no one is going to stand up and test it then they
> are going to have to keep their fingers crossed that it will be OK. I
> did provide Windows binaries for people and TortoiseSVN has been
> providing nightlies. No one asked about any other OS. I am guessing
> for Linux at least we would have had someone that could have tarred up
> a build or something.
I understand that the release cannot be delayed indefinitely. But note
that I'm not asking to hold the release or anything (when suggesting
to make issue #3785 a release-blocker, I was mainly worried about the
Windows-perf of upgrade).
Also note that we've only just started to include the larger
user-community in testing the pre-releases of 1.7. I hope this is not
just a formality, but a genuine opportunity to get some more test
coverage, people trying it out on more different platforms, ... IMHO
this is still a good time for people (those that do not want to just
keep their fingers crossed) to test and report significant (perf)
problems if they find any.
>>> I tried to do it myself, but the NFS performance in my environment was
>>> so slow that I did not have the patience to wait for even the 1.6
>>> version of the tests to finish.
>>
>> I'd like to test NFS performance in our corporate environment
>> (Solaris10/sparc server talking to NAS over NFS). But I don't have a
>> build environment for that system. If someone can build alphas or
>> trunk or whatever for Solaris/sparc, let me know. I'll try to give it
>> a run.
>
> So it sounds like you want to test a server with repos on NFS? I do
> not think there is any reason to think 1.7 would have made that worse
> and hopefully with the caching it made it better. We mainly need
> people to test that normally locate their WC on NFS. That is where
> there might be concern over performance.
No no, I'm really talking about a client and a WC on NFS here. Sorry
if the "server" in my description threw you off. It's actually our
"build server", where we run our production and test builds from
different branches/tags (lots and lots of them), with WC's residing on
the NAS mounted through NFS.
> That said, given all the
> optimization we have already done it is hard to know what more we can
> do. If performance is bad with NFS then it will likely be a symptom
> of SQLite on NFS and we are not going to be able to do much about
> that.
Yeah, I must admit that I'm a lot less worried about it than I was a
couple of months ago. Given that Windows/NTFS perf is good, I think
it's pretty likely that NFS (and CIFS) perf will be quite reasonable
as well. It's just that it would have been nice to see some real
numbers, if someone did a real test.
I'm willing to do some tests, but I can't build it, sorry (@danielsh,
sorry, can't cross-compile either). If someone can build/publish a
pre-release for Solaris/sparc, I'll give it a try, but if not, too bad
... then I'll just keep my fingers crossed :-).
--
Johan
Received on 2011-06-15 01:34:49 CEST