Overall Satisfaction with Sybase Replication Server
I used it to resolve resource issues where other teams were using the OLTP as a DW, and to resolve DR. Before RepServer, they had a live OLTP and a cold standby OLTP and transactions had to be updated using scripts we wrote running in the cron. Since these two systems were in geographically different parts of the country, latency played a role. We had application servers all over the nation where slow connections played a role in performance to the central OLTP. In some more remote locations, the application connecting to the central OLTP was very slow. Once RepServer was implemented, we set it up with by-directional replication, where we also had a active standby we built as well. I then built two more dbs, where those were used as DW so that engineers could scan any table they wanted and we didn't have to worry about table level locks. This improved performance substantially, brought in a near-real time standby DB, and utilized several DW's and made everyone happy...
- Training was excellent on the product.
- Great customer service when questions arose.
- Functionality was awesome.
- Bringing the main OLTP down for maintenance then allowed down time we could previously never afford to make SLA committments
- A few bugs existed in the specific implementation we created requirements for, and we had to wait for Sybase to fix those bugs before it went live to production
- Two way replication had issues where if the same table was being updated by both source and destination, a decision had to made on the "truth". We had to write some special software for specific business requirements.
- Sometimes RepServer would go down, and we didn't know about it until the next time a healthcheck was performed. We had to write some monitoring software to notify us if things went down.
- The positive impact was better performance from our application using these new databases available to the business.
- Negative impact was development time
- Another negative impact was that Sybase engineering had to patch RepServer to fix a bug we discovered.
I asked questions about a warm standby and multicast replication. We ended up using bi-directional replication and a warm standby db. Having 3 or more live OLTP DB's where each was read-write was going to be difficult to manage for contention if the same table was being updated. We stuck with an easier solution to have only two OLTP's, and then used additional DB's as read-only DW's.