savannah-hackers-public
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Savannah-hackers-public] Re: [Savannah-help-private] Contribute


From: Sylvain Beucler
Subject: [Savannah-hackers-public] Re: [Savannah-help-private] Contribute
Date: Thu, 7 Jun 2007 20:28:03 +0200
User-agent: Mutt/1.5.13 (2006-08-11)

Hi,

> As far as I understand are we interested in getting patterns of access 
> behavior in order to optimize the server configuration. Could you tell me 
> more about this, what exactly can be optimized and based on which 
> information ?

Yes, a goal would be to known what kind of pages are most
requested. The individual path of each visitor is probably not worth
following, I think it's more important to check what kind of content
is requested and what could be potentially cached.

For example, ViewCVS can deliver both the latest version and old
versions of a document. If 90% of requests concern the latest version,
then we could think about caching it, and update the cache on CVS
commits. But if it's just 20% of requests this is probably not worth
it.

Stephan was also considering the setup of a reverse caching proxy -
would it worth it? That depends on whether the same pages are
frequently requested, or whether the requested pages are all very
different.

Eventually any stat that could show us how much the ViewVC service
weigh in the current Savannah load average. If we discovered that 80%
of the CPU is dedicated to the ViewVC service only, I guess we'd
consider switching to another CVS web browser, or disabling the
service, or in any case putting more effort in improving its
efficiency.

Cheers,

-- 
Sylvain

On Thu, Jun 07, 2007 at 01:48:00PM +0200, address@hidden wrote:
> Sorry, not much, I was out of office .....
> 
> Until now I have a parser for the log file written in Perl. The next step 
> is to understand what we can gather from this information.
> The log gives us among other things IP number, time, URL. 
> 
> I have seen we do not have any session logging, so we could distinguish 
> single user by their IP number. The IP number could be assigned 
> dynamically, so the user-IP coordination is limited to a time window. But 
> also several users could have the same IP number ( proxy server ).
> So taking into account this we could use the Browser Information plus IP 
> number to try a further distinction between users. On this point I am 
> working in this moment.
> 
> As far as I understand are we interested in getting patterns of access 
> behavior in order to optimize the server configuration. Could you tell me 
> more about this, what exactly can be optimized and based on which 
> information ?
> 
> Any comment ?
> 
> cheers,
> Reinhard
> 
> 
> "Sylvain Beucler" <address@hidden> wrote on 06/06/2007 11:35:07:
> 
> > Hi,
> > 
> > Did you progress?
> > 
> > -- 
> > Sylvain
> > 
> > On Wed, May 23, 2007 at 12:28:59PM +0200, address@hidden 
> wrote:
> > > Indeed, works fine.
> > > So this evening I can start work 
> > > Reinhard
> > > 
> > > 
> > > 
> > > 
> > > "Sylvain Beucler" <address@hidden> 
> > > 21-mag-2007 22:45
> > > 
> > > To
> > > address@hidden
> > > cc
> > > address@hidden
> > > Subject
> > > Re: [Savannah-help-private] Contribute
> > > 
> > > 
> > > 
> > > 
> > > 
> > > 
> > > Hopefully this will work this time.
> > > 
> > > -- 
> > > Sylvain
> > > 
> > > On Mon, May 21, 2007 at 10:05:15PM +0200, address@hidden 
> 
> > > wrote:
> > > > Sylvain,
> > > > 
> > > > I changed a lot of things, I send you the new key, sorry for the 
> delay 
> > > > ............
> > > > 
> > > > cheers
> > 




reply via email to

[Prev in Thread] Current Thread [Next in Thread]