Back in the world of work, I spent an interesting day talking to a customer about scaling up their CLM estate. They have a user base of around 6,000 at the moment – but plan to ramp this up to about 18,000 in the next year or two. We will be using some of the excellent material on different deployment topologies on http://jazz.net/, as well as taking a careful look at how they currently use the existing infrastructure. Several things about the current setup will help us scale up:
- They have used proxy servers to front the different instances so they can move them to different (bigger) machines and even data centres.
- They are currently running on single app servers – so we can get a lot of mileage out of splitting out the different parts of RTC, etc on to their own application servers, databases and machines
- They are running on a good database infrastructure (in this case, Oracle) and we can look at improvements on the backend via tuning, etc.
Making changes to these systems while still maintaining full service is important – we can’t just shut down for a week while we experiment. Fortunately, they do have good development and UAT instances where we can try out any changes before they change production servers. We also have some good experiences from the infrastructure that supports the 56,000+ user population on our own CLM servers across IBM, and the jazz.net infrastructure that has users in the 100’s of 1000’s. In the meantime, we have the blessing and curse of corporate standards for everything from login names to database settings to content with – so we have to tread a fine line between what IBM recommends and what their architects will let us do.