Faster, Better and Cheaper (FBC) has long been the motto for Replication Server product development. FBC aims to improve replication performance and usability, reducing total cost of ownership for users. Sybase Replication Server 15.7, the newest release of Replication Server family, best illustrates this principle.
The crème de la crème of all RS 15.7 features is Multi-Path Replication (MPR). In our last MPR blog post, we introduced the MPR technology and its usages. But users traditionally have to acquire new expensive hardware in order to boost software performance to meet ever increasing business demands for low latency data movement.
In this post, performance improvement from a recent user’s testing will illustrate FBC through MPR without additional hardware CapEx. We will compare an existing single-path replication to the same replication done with increasing numbers of parallel replication paths (i.e., MPR) on the same set of hardware. User measured volume of replication data is in GB per hour, and latency is in minutes.
Increased performance with MPR*
The performance chart demonstrates near linear scalability when multiple parallel replication paths are employed in user’s test environment. User found 96 percent increase in the replication volume by using two parallel paths, and 173 percent increase by using three parallel paths.
Reduced latency with MPR*
Latency measurement goes hand-in-hand with volume to best demonstrate performance. The replication latency chart shows less time consumed when replication is done using multiple parallel paths. User found 44 percent latency reduction by using two parallel paths. With three parallel paths, latency is further reduced by 57 percent over single-path replication.
This user testing of MPR demonstrates Replication Server’s FBC development principle via increased replication volume and reduced latency without additional cost of new hardware. We expect other users will be able to benefit from this new feature of Replication Server to meet increasing demand for real time data.
* Your performance improvement may vary based on your own transaction profile and hardware employed.