Performance and Scalability Benchmark:
InterSystems IRIS and Intel Optane DC Persistent Memory
InterSystems and Intel recently conducted a series of benchmarks combining InterSystems IRIS® data platform with 2nd Generation Intel® Xeon® Scalable Processors, also known as Cascade Lake, and Intel® Optane™ DC persistent memory (DCPMM). The goal of these benchmarks was to demonstrate the performance and scalability capabilities of InterSystems IRIS with Intel’s latest memory technologies in various workload settings and server configurations. Along with various benchmark results, three different use cases of Intel DCPMM with InterSystems IRIS are provided in this report.
Two types of workload were used to demonstrate performance and scaling — a read-intensive workload and a write-intensive workload. The reason for demonstrating these separately was to show the impact of Intel DCPMM on different use cases specific to increasing database cache efficiency in a read-intensive workload, and increasing write throughput for transaction journals in a write-intensive workload. In both of these use-case scenarios, significant throughput, scalability, and performance gains for InterSystems IRIS were achieved.
- The read-intensive workload leveraged a four-socket server and massive long running analytical queries across a dataset of approximately 1.2TB of total data. With DCPMM in Memory Mode, benchmark comparisons yielded a significant reduction in elapsed runtime — it was approximately six times faster when compared with a previous-generation Intel Xeon E7 v4 series processor with less memory. When comparing like-for-like memory sizes between the E7 v4 and the latest server with DCPMM, there was a 20% improvement. This was due to both the increased InterSystems IRIS database cache capability afforded by DCPMM and the latest Intel processor architecture.
- The write-intensive workload leveraged a two-socket server and InterSystems HL7 v2 messaging benchmark, which consists of numerous inbound interfaces. Each message has several transformations and then four outbound messages for each inbound message. One of the stringent requirements in sustaining high throughput is the message durability guarantees of InterSystems IRIS for Health™, and the transaction journal write performance is crucial in that operation. With DCPMM in App Direct Mode as direct access (DAX) presenting an XFS file system for transaction journals, this benchmark demonstrated a 60% increase in message throughput.
To summarize the test results and configurations: DCPMM offers significant throughput gains when used in the proper InterSystems IRIS setting and workload. The high-level benefits are increased database cache efficiency and reduced disk I/O block reads in read-intensive workloads and also increased write throughput for journals in write-intensive workloads.
In addition, Cascade Lake-based servers with DCPMM provide an excellent update path for those looking to refresh older hardware and improve performance and scaling. InterSystems® technology architects are available to help with those discussions and provide advice on suggested configurations for your existing workloads.