[Building Sakai] Successful upgrade @ AMC

Dave Ross dave.ross at gmail.com
Thu Jun 11 11:54:57 PDT 2009


Just a quick note that Albany Med has completed it's migration from Sakai
2.4.x to 2.5.x - and is planning on 2.6.x later in the summer. All
migrations completed without a hitch (the resources conversion took one of
our app servers about 8 hours to convert approximately 100k resources).

It's amazing how much capacity we've gained with a move to 64bit. Our load
test which works against a few common tools (resources, schedule, forums,
site info, and a few others) saw a max throughput increase of over 10x.
(There are also several performance improvements to resources and forums
that helped this value significantly). We remain running some contrib tools
such as SiteStats - and we have dropped JForum from our install. We have 8
homegrown Sakai tools - 6 of them are linked tools that rely heavily on core
Sakai services (authz, gradebook, etc). We are also looking at giving Tests
& Quizzes another try this year.

Here's some infrastructure details about prod:

*Old Environment:*

   - 2 App Servers
      - 2x Xeon 2.8ghz
      - 4gb ram
      - Windows Server 2003 Sp2
      - JDK 1.5.0_11
      - Tomcat 5.5.23 -Xmx1260m

      - Database Server
      - MySql 4.1.22
      - Windows Server 2003 R2 Sp2

      - File Server
      - 1x Xeon 2.8ghz
      - 2gb ram
      - Windows Server 2003 Sp2
      - direct attached storage

*New Environment:*

   - 2 App Servers
      - 2x Xeon 5160 (3.0ghz dual core)
      - 8gb ram
      - Windows Server 2008 x64
      - JDK 1.5.0_18 (AMD/EMT64)
      - Tomcat 5.5.26 -Xmx5000m

      - Database Server
      - 2x Xeon 5160 (3.0ghz dual core)
      - 8gb ram
      - Windows Server 2008 x64
      - MySQL 5.0.81 64bit

      - File Server
      - 2x Xeon 5450 (3.0ghz quad core)
      - 8gb ram
      - SAN attached storage
      - Windows Server 2008 x64 (failover cluster - other box is MSSQL)

We load balance the environment using Citrix Netscaler 7000 in a redundant
pair - we don't do full SSL offload (e.g we are https all the way through to
the app servers), and we also dropped APR from this config - so it's just
standard Tomcat/JSSE for SSL on the app servers. Since the load balancer
pools these connections, we do not see any discernable difference between
APR and SSL (on 64bit windows).

All in all - our customers are happy!

-Dave Ross
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://collab.sakaiproject.org/pipermail/sakai-dev/attachments/20090611/ae110e7b/attachment.html 


More information about the sakai-dev mailing list