From the testing results of the two use cases, we observed that the two-node RAC cluster, configured as the baseline setup on Dell EMC infrastructure, scaled well in terms of TPS throughput performance when tested with an increasing HLR OLTP benchmarking workload. However, during the baseline RAC-only tests, some performance impact was observed at higher throughputs due to increased latency for DML operations caused by an increase in ‘log file sync’ wait events.
When we deployed TimesTen on a Dell EMC PowerEdge server in front of the baseline RAC setup, the TimesTen cache was not only able to offload query and DML transactions from the backend RAC database, but it was also able to deliver the same TPS throughput with up to 37 times improvement in transaction response times. Note that these response times were delivered with the transactions initially being persisted to the local TimesTen server’s memory (DRAM and/or PMem) and disks before being asynchronously propagated to the RAC database. During the TimesTen tests, it was also observed that the backend RAC database node’s CPU utilization went down during the two use cases. The ‘log file sync’ wait events observed during the baseline RAC testing were significantly reduced. During the TimesTen tests, the RAC database resource utilization went up with the DB CPU (as reported by AWR) delta values increasing by 35 and 18 percent during Use Case 1 and 2, respectively, as compared to the baseline RAC tests. This testing shows that TimesTen deployed as a cache helps optimize the utilization of the backend database hardware resources.
Therefore, in environments where some RAC database queries or DML transactions are saturating the infrastructure utilization and impeding the overall performance, customers can take advantage of deploying a TimesTen cache on PowerEdge servers to offload those transactions. This will help to accelerate the overall database response times and to optimize the utilization of the Dell EMC infrastructure, thus yielding a better return on investment.
We also demonstrated that using Intel PMem, in conjunction with DRAM modules as front-end cache, can help to provide increased capacity for TimesTen. This enables a much larger volume of data to be cached within a single PowerEdge server. PMem can not only provide greater capacity than DRAM-only configurations but Use Case 2 results showed that it can do so without compromising the overall database performance. During their initial purchasing and capacity planning, customers can definitely consider using Intel PMem supported by PowerEdge servers for their TimesTen memory needs.
Note: The replication lags, latencies, and the maximum throughput or TPS rate we observed in this study are strictly applicable to our setup in the lab and are NOT indicative of the maximum performance capability of the individual hardware and software products, including the schema, dataset, and the workload used in the solution. Performance results will vary based on the deployed environment.