[OSGeo-Discuss] Fwd: [SAC] Upgrade Planning 2014 - Project Feedback

Alex Mandel tech_dev at wildintellect.com
Fri Apr 18 16:50:31 PDT 2014


Calling all Project Steering Committees, SAC is looking at the future of
OSGeo hosted services. Please chime in with your wants and needs for the
next 3-5 years. We want to maximize services while being efficient about
effort (pooling sys admin time amongst projects).

http://wiki.osgeo.org/wiki/Infrastructure_Transition_Plan_2014

More details below.

Thanks,
Alex
OSGeo Sys Admin Committee Chair


-------- Original Message --------
Subject: [SAC] Upgrade Planning 2014
Date: Thu, 10 Apr 2014 15:25:41 -0700
From: Alex Mandel <tech_dev at wildintellect.com>
Reply-To: tech at wildintellect.com, System Administration Committee
Discussion/OSGeo <sac at lists.osgeo.org>
To: System Administration Committee Discussion/OSGeo <sac at lists.osgeo.org>

In light of our recent hardware woes, the ending of osgeo1, and the age
of current machines (3-4yrs) I think it's time to start planning what to
do next.

We do have budget, and the board is interested in ideas and additional
funding is not out of the question.

>From my perspective, I think we need:
1. To survey PSC of all projects to assess what services we should
offer. Do we need buildbots, sphinx builds, mail service, issue
tracking, various CMS/wiki, mirrors, bandwidth for downloads? Now that
everyone is love with Github (any word on what the next hot host will
be), are there things we should retire?

2. Look at other hosting options besides physical machines in one place.
Renting space like QGIS or OSM, racking machines elsewhere, getting
OSGeo-ICA labs to mirror. If we do mirror look at GeoCDN and MirrorBrain
for geo-ip redirection balancing.

3. Look at new hardware that better meets the needs. RAID is nice but
not always the right answer to needs as we discovered recently. We also
didn't buy specific to Ganeti/Cloud style setups where hotcopy failover
works best with multiple identical machines, and lots of smaller disks
with single disks per VM keeps i/o from competing (RAID is handled in
mirror mode via DRBD over the network between nodes).

Anyone want to tackle making a short survey for Projects to describe
their needs and wishes?


All ideas welcome, we'll pool it all into proposal and a wiki page
before deciding on anything.

Thanks,
Alex
_______________________________________________
Sac mailing list
Sac at lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/sac





More information about the Discuss mailing list