[squeak-dev] Re: [Pharo-project] SqueakSource down... again

Andreas Raab andreas.raab at gmx.de
Mon Feb 21 19:49:31 UTC 2011


On 2/21/2011 1:31 PM, Miguel Cobá wrote:
> El lun, 21-02-2011 a las 19:25 +0100, Michael Haupt escribió:
>> Hi,
>>
>> at HPI, we have a SqueakSource with a GemStone backend. The only downtimes we experienced were due to necessary maintenance.
>
> I think the number of users that hit squeaksource.com vs HPS's own
> everyday is a key factor in the perceived performance and net uptime of
> each installation.

It's not. We used SS at Teleplace at first and it died every other day 
due to heavy commit/update frequency. When we switched to Gemstone we 
never had these problems again. The problem with SS is that it has very 
poor concurrency control. It just trips over its own feet when several 
people commit and update at the same time. I was actually looking at 
fixing these problems at first only to realize that SS was clearly 
designed as a single-user system with very little thoughts being put 
into where and how concurrency plays a role in various operations.

When we switched to Gemsource, these problems went away (though I'm not 
fully sure why). But they did get replaced with others, most notably the 
problem of our entire source code being locked in a system that we would 
be unable to access if our license ever expires. And although we had a 
sweet deal with Gemstone, due to the sweetness of the deal this happened 
every six months and a couple of times in critical release situations. 
After it happened once too many it was deemed to be no longer acceptable 
and we switched to a simple WebDAV based repository. Which has served us 
well.

Cheers,
   - Andreas



More information about the Squeak-dev mailing list