Servers & Systems: The Right Compute
ComputeExperts

Delivering hundreds of millions of reads per second

Find out what happened when a team did benchmark tests on HPE Superdome Flex running GigaSpaces InsightEdge in-memory data platform.

By Compute Experts guest blogger Noam Herzenstein, GigaSpaces Director of Product Marketing

HPE Superdome Flex-GigaSpaces.png

With enterprises requiring faster insights to unlock value for more efficient operations and superior customer experiences, accelerating and scaling data and analytics processing has paramount to business success.

To address these ever-increasing requirements, GigaSpaces, the leading provider of in-memory computing platforms that drive digital transformation, and HPE have partnered to provide a solution.

The two companies ran benchmark tests on an HPE Superdome Flex scale-up server running the GigaSpaces InsightEdge in-memory data platform. The results demonstrated superior performance, efficiency, and scalability of the solution. (Read the technical paper to step through a test and see some real-world use cases.)

The benchmark testing had two primary goals:

  • Demonstrate the superior performance that can be achieved by combining HPE Superdome Flex server with GigaSpaces InsightEdge in-memory technology
  • Show specific benefits of scale-up in-memory compute compared to scale-out and cloud infrastructure

Testing infrastructure included the following key components:

  • HPE Superdome Flex server with 12 Intel® Xeon® Platinum 8180M Processors 28 cores, 2.5 GHz, 205W, 38.5M and 4.5 TB of memory running GigaSpaces InsightEdge v15.2, scaling with the number of working nodes
  • Yardstick open-source testing framework

Testing processes included the following stages: 

An HPE and GigaSpaces engineering team led by a chief architect ran more than 600 sessions over the course of three weeks. Benchmark stages included the following:

  • Engineering kick-off: Server installation, benchmark tools calibrations, software deployment, add automation, and dry run
  • Microtuning: Handling each “performance cell” and optimizing batch size and op counts
  • Thread/node balance: Checking the optimum utilization between jobs/nodes and threads
  • Running distributed tasks that utilize the server and present a linear scale of read/write operations

The Results: Latency <1 ms in 99% of cases

HPE Superdome Flex and GigaSpaces InsightEdge deliver an ideal platform to support extreme performance in complex, high-performance database workloads. The GigaSpaces distributed architecture is an ideal match for the scalable, modular HPE Superdome Flex server when it comes to data-intensive applications and use cases. 

GigaSpaces utilizes the HPE Superdome Flex modular architecture and CPUs in the most optimal way. Instead of having a CPU sitting idle, as is the case with non-distributed database systems, GigaSpaces leverages the full compute power and unparalleled scale of HPE Superdome Flex.

Achieving 75–85% CPU utilization is considered optimal. Less than that means underutilizing the hardware, but at 90–100%, consistent CPU utilization is not stable and will tend to freeze.

Unlike traditional databases, which utilize a single processor and thus are limited in terms of scaling, GigaSpaces is distributed and can deploy additional partitions as needed to handle more data. Its in-memory technology powers incredibly fast performance on HPE Superdome Flex server, with query latency of less than 1 millisecond and the ability to run hundreds of millions of operations per second on a single server. The superior performance, efficiency, scalability, and RAS features delivered from this in-memory solution can help businesses meet the most demanding Service Level Agreements (SLAs) from their mission-critical applications.

Figure 1 shows the results of both read and write tests. The CPU utilization scaled up with the number of shards, indicating the HPE Superdome Flex server resources are utilized effectively to support demanding data throughputs.

Figure 1. Latency and operations per second test results.Figure 1. Latency and operations per second test results.

Significance for real-world use cases and applications 

HPE Superdome Flex server with GigaSpaces InsightEdge delivers a single platform that can address the challenges of multiple use cases across several industries (including financial services, transportation, retail). Regardless of industry or use case, the platform is designed to: 

  • Run digital applications at scale and speed 
  • Rapidly introduce new modern digital applications 
  • Modernize and offload from expensive legacy infrastructure 
  • Ensure always-on applications and services 
  • Develop once and deploy anywhere—on-premises, cloud, hybrid, and multi-cloud 

One use case example is elastic risk calculation engine for net present value (NPV) which involves calculating NPV for a massive set of financial transactions. NPV of a time series of cash flows—both incoming and outgoing—is defined as the sum of the present values of the individual cash flows of the same entity. This use case presents significant challenges:

  • Calculating NPV for a large number of items can take an excessive amount of time. 
  • Running calculations for multiple scenarios may take many hours to complete. 
  • Increasing or decreasing the calculation infrastructure capacity on-demand must not create downtime.

Additional use cases of extreme data processing include algorithmic trading, real-time payments, price forecasting, flight planning and rerouting, just to name a few.

To learn more about the joint HPE Superdome Flex and GigaSpaces solution that dramatically reduces the time and risk to complete large, complex calculations, read the technical paper here.


Meet our Compute Experts guest blogger Noam Herzenstein

Noam_Herezenstein.pngNoam Herzenstein is GigaSpaces Director of Product Marketing. He previously worked at NICE Systems and BMC Software. He has a LL.B. in Law from Tel Aviv University and an MBA from the University of Maryland. He and his wife, Michal, live in Tel Aviv with their two teenage sons. 

 

 


Compute Experts
Hewlett Packard Enterprise

twitter.com/hpe_compute
linkedin.com/showcase/hpe-servers-and-systems/
hpe.com/servers

About the Author

ComputeExperts

Our team of Hewlett Packard Enterprise server experts helps you to dive deep into relevant infrastructure topics.