When is one Broker or more is better


I’m trying to figure out if one Broker is sufficient or would it be better to have multiple brokers for increase in performance. Just interested in hearing everyone thoughts and inputs on singe vs multiple brokers. Currently, we have multiple IS publishing to a single broker and it’s working fine for us. However, we have an opportunity to improve our current architecture with an upgrade to v8.x. We’re also scaling horizontally vs vertically as well. Just trying to get everyone experience from a architecture perspective to see if there’s a solid architecture components for B2B (IS, Broker, TN, EDI). Also, which hardward is the best running wM for Solaris env.

The best boost I usually see is in moving the Broker off to its own machine and not sharing the same machine with IS.

It depends on traffic volumes, of course, but usually the Broker is never even close to being a bottleneck. The pokey slow bits are the components hosted on IS (doing way more work than Broker does).

The Broker is primarily bound by disk i/o. So moving it’s own box will help. If the broker’s data files live on shared disk file system (as would be needed in broker cluster) then ensure the shared file system’s mount has high bandwidth, like a fiber LUN.

Before v 8, the webMethods broker only supports failover clustering (active/passive) I believe v 8 supports (active/active).

i believe active/passive is the way to go in v8 as well… apparently software ag will release v.8.1 in april along with updated clustering guide (supporting windows server 2008 r2)