Actian announced a new record TPC-H result with Hewlett Packard Enterprise and Intel to demonstrate the big data analytics performance of the new Intel Xeon processor E7-8890 v4. With over 2.14M QphH and $0.38/QphH, Actian Vector sets new records for performance and price/performance on non-clustered systems for the 3000GB scale factor. These numbers beat the previous record performance published on Cisco using Microsoft SQL Server by 2X, at a 36% lower cost per query.
What do these results show about the Actian Vector database performance that relates to a business benefit? Reducing the overall solution cost makes it possible to incorporate more data into the business analytics system to provide a broader perspective from which to gain business insights. Faster query performance shortens turn-around time on business questions and makes insight on bigger problems more reasonable. With better-informed and faster business insights, companies can increase revenue and profits, improve productivity and customer loyalty, and reduce operating costs and business risks.
The Transaction Processing Performance Council (TPC) develops objective, verifiable performance benchmarks. This decision support workload was designed for broad industry-wide relevance, with 22 different business-oriented queries simulating the analysis needed to make better business decisions around sales, inventory, production, etc. There are also multiple steps of mandated inserts and deletes that an independent auditor verifies to ensure the database maintains ACID semantics, even through a hardware failure.
These record-setting results show how well Actian Vector handles this workload – 5 times faster and 50% less expensive than an Oracle solution from 3 years ago, for example. Ad-hoc cost per query has dropped from $21 to just $0.38 in just the past six years—a 98% cost reduction, while queries-per-hour capability has increased by 19X.
Internally, we compared our performance running the same workload on exactly the same configuration using the prior generation processors (E7-8890 v3), which showed a 20% difference due to the processor change. Clearly, one large factor contributing to the doubling of TPC-H performance over the Cisco configuration was changing the database from Microsoft SQL Server 2016 to Actian Vector with our columnar data format and efficient data compression. Actian Vector takes direct advantage of the large caches and vector instruction extensions Intel has designed into the Xeon processors. Vector software features deliver on the performance of industry-standard CPUs and servers, eliminating the need for specialized hardware appliances to achieve capacity and performance. Other factors may have been the choice of operating system (RHEL 7 rather than Windows Server), and the different storage subsystems.
We have a history of top performance, showing a significant lead over other databases and query engines that are just starting to move in this architectural direction:
– Top result for SF3000 (published 2016)
– 4 of top 8 results for SF1000 (2011 – 2014, including historical results)
– Top 4 results for SF300 (2011 – 2013, all historical results)
– Top 5 results for SF100 (2011 – 2013, all historical results)
Since the X100 query engine that drives Actian Vector performance is the same query engine at the heart of Actian Vector in Hadoop, we look forward to sharing audited results on a Hadoop cluster in the near future.