In our project (in my day job), we have just implemented the first version of replication.
This is a pretty tough challenge as we are taking replication beyond mere reporting. We have a very large system, with millions of transactions per day. We also have multiple applications connected to the system.
This has meant to distribute load, we are using a subscriber database for non time latency critical tasks, and the publisher for the real time aspects. So we have this interleaving between the publisher and the subscriber based on the application at that point in time.
Although we are following a publisher-subscriber model, we will certainly be looking to use multiple publishers in the future.
To be honest this is pretty exciting stuff, and we are really pushing the boundaries of how you use replication and the performance benefits for us are incredible.
It does lead me to believe that replication is an underused feature in performance systems, and i can see as time goes by, even more use of it.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment