Modern Middle Manager
Primarily my musings on the practical application of technology and management principles at a financial services company.
Project Kagome

Friday, May 16, 2003  

One of the more idiosyncratic practices we have is to name our Citrix Metaframe servers after women in sci-fi. Recently that's narrowed to female characters in anime. The latest Citrix project is named Project Kagome from Inu-Yasha. The name has no real significance other than as an ode to our goofiness; however, the project has been an interesting stretch for us. The overall project goals were:

1. Provide near-desktop performance for most applications.
2. Provide high availability.
3. Decrease instability and "quirks" in the user environment.

A number of choices led us to the Citrix environment for many of our desktop users. From that starting point we implemented Citrix Metaframe XPs on Vmware's ESX Server and got to about 22 users before performance problems finally caught us. That initial implementation has allowed us to really clamp down on quirks and bugs to the point of extremely stable sessions. However, we're not coming through on issue #1 -- performance. So how are we going to deal with that while preserving goals 2 & 3?

Our plan was to use a Dell PowerEdge 1655MC with three blades to serve as the Metaframe servers and use our core switch, a Cisco Catalyst 6509 to perform server load balancing (SLB). SLB is key here -- by publishing a virtual server that routes to a farm of real servers, we can keep the number of Citrix users per server down while also giving us the ability to take one out of service for work. In practice, we have three blades -- one is the baseline/test server, the others are the deployed servers doing time in the farm. SLB will balance between these two production blades and any additional ones we may add later. Finally, we are documenting each step of the way to make the process easily duplicated. With Dell's Remote Install software we're able to image the server at each step of the way. Today we performed some internal quality testing and beta deployment should start on Monday. We could be in full production by the end of next week. It looks like this will fulfill goals 1 through 3 while also giving us the scalability and disaster recovery capabilities we need. As for how many Citrix users each server will handle, well...we'll find out next week.

posted by Henry Jenkins | 5/16/2003 07:15:00 PM

Comments: Post a Comment
the author
open source