I was hoping someone could help me out here, I am setting up the integration cluster side and the documentation seems to point to an active/active cluster. But after reading the posts most seem to say that loadbalancing is that way to go not is clustering. How is the active/active clustering handled, ,locking files etc. Especially if I use a shared storage. Should I use third pary like the VCS, but then the application should be responsible for controlling the locking and access to the files/database.
My questions is how to we setup the IS architecture, do we relay on the integration server and run mutple instances with seperate storage or purchase a load balancer. Any guidance would be appreciated.
Thanks
Purchase a load balancer*. Use multiple IS instances in active/active mode. There is no need for the IS instances to share files to function. Any file sharing will be introduced by integration processing needs that you define.
Use a database for shared storage if needed. Accessing files from multiple systems can be done but you need to make sure your solution accounts for all scenarios.
There are several posts that discuss the pros and cons of using the IS clustering facilities. My advice is to not use the IS clustering feature unless what you need to do cannot be addressed in some other way.
I will second Rob’s suggestion. Use Load Balancer instead of Clustering.
Please note that Load Balancing and Clustering are entirely different concepts. You would also want to go over all your integration points and see how they will behave with LB vs C. Do understand that some integration solutions will work better with LB while others may require changes if you are using C.
Let us know if you have any specific question on how an integration will behave with LB vs C.
Strictly speaking, you’re right that they are different but the concepts are closely related. A cluster of servers (or applications) provides fault tolerance and/or load balancing. There are different types of clustering solutions, some which provide one or the other feature or both. A load balancing cluster provides workload distribution and a level of fault tolerance (if one server goes down, the others pick up the work).
If your statement “Use Load Balancing instead of Clustering” really meant use a load balancing cluster instead of Integration Server clustering then we are in complete agreement (with the exceptions noted in this post).
It is generally always beneficial to use “external” Load Balancer than use any Clustering solution. (when compared wrt pain points of implementing/maintaing LB vs C solution)
I just wanted to clarify the point to “zeus_zot” that your suggestion assumes that his integration points will behave the same with either using LB or C. Which is not always true.
He still needs to go over all his integration points and decide if he can use LB over a C solution.
I again second your suggestion of “My advice is to not use the IS clustering feature unless what you need to do cannot be addressed in some other way.” And no one would have summed pros/cons better as you did in this post you mentioned
You’re absolutely right that the style of cluster may have an impact on how the integration solution behaves. It was good advice to encourage zeus_zot to review the integration touch points to confirm desired behavior.
My post was more about the notion that load balancing and clustering (using general definitions) are completely different. IMO, they are not. Load balancing is a feature enabled by clustering. One cannot load balance without a cluster–again, cluster here meaning the generic “group of resources.”
The “Integration Server clustering” facility has very specific features. IMO, one should not refer to this simply as “clustering” but should always qualify it as “IS clustering” for clarity.
Again, your advice to make sure that the integration points behave properly based on the clustering solution is right on. One does indeed need to consider the implications of the use of load balancing and/or IS clustering.
