Exasol Accelerates In-Memory Data Analytics by up to 114 Percent1

Exasol offers a high-speed in-memory database, which enables organizations to make faster and smarter decisions. The company found that the Intel® Xeon® Platinum processor accelerated performance by up to 114 percent.1

Exasol is passionate about helping companies to run their businesses smarter and drive profit by analyzing data and information at speed. The company develops a fast database for analytics and data warehousing, and offers first-class know-how and expertise in data insight and analytics. The in-memory analytic database combines in-memory, columnar compression and massively parallel processing.

Handled well, data can be hugely empowering. It enables businesses to respond quickly to what’s happening in their organizations and industries; and helps them to make better decisions based on current information. The problem is that many organizations are overwhelmed with the volume of data they have, and the pace of its change. Every customer interaction, online or off, is a data point. The Internet of Things delivers streaming data, and market sectors and companies are in a constant state of change. Some organizations are still batch processing reports, leaving their analytics software to process vast databases overnight. The next day, the reports are ready. But they’re based on data that’s already one day old.

Exasol helps to solve these problems. It offers a high-performance database with a strong focus on analytical queries. One of its differentiators is that it’s an in-memory database that only needs a memory size of about 10 percent of the total database size. The software uses intelligent caching to keep frequently used data in memory. That means it can handle larger databases than solutions that require all the data to be stored in memory.

Optimizing Performance
The solution has been optimized to take advantage of the compute resources available. It’s typically hosted across a cluster. Queries are performed across all cores on all processors in parallel, with communication between the cores where necessary. If more processing power is needed, the solution can be scaled by adding more machines. 

The benchmarks showed a significant performance increase. If customers can increase their performance by simply replacing the processor, that’s an easy way for them to accelerate their analytics, and support more concurrent users.

Dr. Jens Graupmann, Product Manager, Exasol

The team at Exasol has also optimized its software, line by line, to take advantage of low level processor features. The company has used Intel® VTune™ Amplifier to identify CPU related performance bottlenecks and to tune performance. “We consider Intel® VTune™ Amplifier to be the most comprehensive and powerful tool for that purpose on the market,” says Dr. Jens Graupmann, product manager, Exasol.

“We’ve also used Intel® Trace Analyzer and Collector to visualize and optimize our multi-node behavior, says Graupmann. “The best thing is that you can actually see what your code does and how it performs. The tools assist you in finding bad computation patterns, lock stalls and other concurrency issues.”

Using Exasol’s solutions, customers are able to accelerate their data processing from hours to minutes. “Performance and speed change how people interact with data,” says Graupmann. “Before, they had to wait a day for their data. Now they can almost interactively do what they previously did on a day-to-day basis. This increases productivity, and means companies can solve problems they couldn’t before. We have one customer that had data from a fleet of vehicles. Using Exasol technology, they can, for the first time, respond to any issues while the vehicle is on the road.”

Testing the New Intel® Xeon® Platinum Processor
Exasol worked with Intel to benchmark the performance of the new Intel® Xeon® Platinum processor, and compared its performance to the previous generation Intel® Xeon® processor E5 family v4. Exasol ran its own internal benchmark on hardware provided by Intel. The benchmark simulates a real-world analytics scenario, and measures the speed of queries. The database size for the test was 100 GB.

The same test code was run on each processor configuration. The Intel® Xeon® Platinum 8168 processor was tested in a 2 socket configuration, which is how customers typically configure their servers for running Exasol software. This processor was compared with the previous generation Intel® Xeon® processor E5-2699 v4. Additionally, the Intel® Xeon® Platinum 8180 processor was tested in a 4 socket configuration, compared against the previous generation Intel® Xeon® processor E7-8890 v4.

Exasol found that the Intel® Xeon® Platinum 8168 processor delivered a 51 percent performance increase over the previous generation, and the Intel® Xeon® Platinum 8180 processor delivered a 114 percent speed-up.1

Core Count Increases Performance
The Intel® Xeon® Platinum 8168 processor has 24 cores, supporting 48 threads and the Intel® Xeon® Platinum 8180 processor has 28 cores, supporting 56 threads. By comparison the Intel® Xeon® processor E5-2699 v4 has 18 cores, and the Intel® Xeon® processor E7-8890 v4 has 24 cores.

Graupmann says: “Our software uses the improved memory bandwidth and reduced latency in the Intel® Xeon® Platinum processor to improve performance. We take advantage of the increased core count, as well as the improved performance of each core. Our software was designed to execute queries in parallel from the outset, so we can use new CPUs with a higher core count with minimal modifications to our software. It scales automatically.”

Conclusion
“The benchmarks showed a significant performance increase,” says Graupmann. “If customers can increase their performance by simply replacing the processor, that’s an easy way for them to accelerate their analytics, and support more concurrent users.”

Product and Performance Information

1

EXASOL* 6.0: custom TPC-H like workload. OS: custom CentOS* 6.8. Testing by Intel and Exasol May 2017.
Baseline: 2S Intel® Xeon® processor E5-2699 v4, 2.2 GHz, 22 cores, turbo and HT on, BIOS 338.R00, 64GB total memory, 2133 MT/s / DDR4 LRDIMM, 1 TB SATA HDD, custom CentOS* 6.8 kernel 2.6.32-642.
New: 2S Intel® Xeon® Platinum 8168 processor, 2.7 GHz, 24 cores, turbo and HT on, 192GB total memory, 12 slots / 16GB / 2666 MT/s / DDR4 LRDIMM, 1 TB SATA HDD, custom CentOS* 6.8 kernel 2.6.32-642.

Baseline: 4S Intel® Xeon® processor E7-8890 v4, 2.2 GHz, 24 cores, turbo and HT on, BIOS 338.R00, 256GB total memory, 32 slots / 8GB / 1600 MT/s / DDR4 LRDIMM, 300 GB SAS HDD + 1 TB SATA HDD, custom CentOS* 6.8 kernel 2.6.32-642.
New: 4S Intel® Xeon® Platinum 8180 processor, 2.5 GHz, 28 cores, turbo and HT on, 768GB total memory, 24 slots / 32GB / 2666 MT/s / DDR4 LRDIMM, 800GB Intel® SSD DC S3700, custom CentOS* 6.8 kernel 2.6.32-642.
The workload is not I/O and is not memory size bound.

2

Benchmark results were obtained prior to implementation of recent software patches and firmware updates intended to address exploits referred to as "Spectre" and "Meltdown". Implementation of these updates may make these results inapplicable to your device or system. 

Software and workloads used in performance tests may have been optimized for performance only on Intel® microprocessors. Performance tests, such as SYSmark* and MobileMark*, are measured using specific computer systems, components, software, operations, and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more complete information visit http://www.intel.com/benchmarks.