Home > Workload Solutions > Oracle > Guides > Reference Architecture Guide—Accelerate Oracle Database using Oracle TimesTen as an Application-Tier Cache > Use cases overview, results, and key observations
The study and benchmarking in this guide were based on a schema, dataset, and application that simulate the kind of processing performed in a telco Home Location Register (HLR) or Home Subscriber Server (HSS) environment. The schema has four tables— SUBSCRIBER, ACCESS_INFO, SPECIAL_FACILITY and CALL_FORWARDING—arranged in a parent-child hierarchy with SUBSCRIBER as the root table.
In this study, the total dataset generated in the RAC database consisted of data for 2,500,000,000 subscribers (2,500M subscribers). To assess the impact of TimesTen caching with a backend RAC database schema, the two use cases we tested targeted a subset of the total dataset: 400,000,000 subscribers (400M subscribers) for Use Case 1 and 1,800,000,000 subscribers (1,800M subscribers) for Use Case 2.
In Use Case 1, the 400M subscribers were cached in TimesTen running on 768 GB of DRAM-only modules. In Use Case 2, we added 3 TB of Intel PMem to the TimesTen server system, and loaded 1,800M subscribers into TimesTen cache. During Use Case 2, 768 GB of DRAM modules acted as the front-end cache to PMem. These DRAM modules were transparent to TimesTen.
For all tests, the purpose-built HLR application also generated the benchmarking workload. The HLR workload simulated an Online Transaction Processing (OLTP) workload environment with a read/write I/O distribution ratio of 90/10.
For each use case, we ran the HLR benchmark workload directly against the RAC database in order to establish a baseline and measured various performance metrics at the HLR benchmark application, database, and operating system levels. We then cached the use case data (sub)set into the TimesTen cache and ran the HLR workload against the cache. We measured the same performance metrics for both the TimesTen system and the RAC database so that we could compare them. The RAC database performance metrics during the TimesTen tests included the TimesTen replication agent asynchronously propagating the transaction updates and committing them to the backend RAC database. For more details on replication, refer to TimesTen cache AWT replication overview.
During Use Case 1 testing, we observed the following results:
During Use Case 2 testing, we observed the following results:
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.
The two use case results showed that, overall, the two-node RAC cluster, configured as the baseline setup, scaled well.
When we deployed a TimesTen cache in front of the baseline RAC setup, TimesTen could offload query and DML transactions from the backend RAC database. This is reflected in the RAC node’s CPU utilization and a reduction in the RAC 'log file sync' wait events. With the wait events significantly reduced, the RAC database’s CPU utilization efficiency also improved which is reflected by the increased DB CPU metric in the AWR report. This demonstrates that a TimesTen cache can help to improve RAC database consolidation and provide better return on RAC infrastructure investment.
In addition, these results show that when TimesTen is deployed as a cache, it greatly improves the database transaction response times while asynchronously persisting the transactions in the backend RAC database.
Using Intel PMem in Use Case 2 demonstrated that PMem can not only provide greater capacity than DRAM-only configurations for TimesTen within a single PowerEdge server, but it can do so without compromising the overall database performance.
Multiple independent TimesTen servers can be deployed to cache mutually exclusive partitions and to accelerate different datasets of a backend RAC database.
Customers can confidently use this reference architecture, based on their database size and needs, to deploy multiple independent TimesTen cache servers on multiple PowerEdge servers populated with PMem to offload and accelerate their backend RAC databases.