Pavilion: Run Databases Up to 7X Faster and at Scale

Pavilion Data Systems, Inc. announced that its platform with unified high performance, sub-millisecond RW latency, intelligent architecture, patented data protection, and an end-to-end NVMe design makes any database, both on bare metal and virtualized, achieve material performance gains by operating up to 7x faster than using NVMe SSDs in a server.

The firm’s Data Acceleration Platform is composed of the HyperParallel Flash Array (HFA) and HyperOS. The performance density of the hardware and performance algorithms of the software enable the company’s platform to keep up with I/O demands from the databases, delivering a low write latency, and improving database transaction commit time and HA. Writes happen faster by allowing more databases on the same instance or more concurrent users on each database instance, so the ultra-low latency coupled with the high throughput and IO/s enable databases to analyze more data, faster.

Testing in the company’s solutions lab and deployments with customers and database consultants showed that the firm enabled databases like Microsoft SQL Server, Oracle, PostgreSQL, MySQL, Elasticsearch, and Cassandra to deliver more transactions at lower latency with the same hardware, whether on bare metal or virtualized servers. The solution lower latency also has a massive effect in VMware environments, requiring less memory from the database on each VM environment. This means more database files on the same virtualized database instance, and/or more VMs on the same physical server.

“I’ve worked on pretty much every storage array on the planet at this point, and it takes a lot to impress me. When I started learning about the Pavilion architecture, I was blown away by it. I just love it,” said David Klee, founder and CEO, Heraflux Technologies, and Microsoft MVP and VMware vExpert. “SQL Server on VMware comprises the bulk of the servers that we see in the wild. The unmatched performance that Pavilion offers allows me to reduce my SQL Server memory footprint, pack more SQL Servers on the host while improving performance, and squeeze more out of my expensive SQL server licensing – everyone wins!”

In the world of SQL Server on VMware, customers see an average of 30% to 50% reduction in memory required, which translates directly to 30% to 50% fewer SQL Server licenses. In virtualized SQL Server environments, vSphere licenses are cut as well.

Iqvia Inc., provider in connected intelligence, needed to scale its Microsoft SQLServer-based analytics platform, and NVMe SSDs inside their servers could not provide the speed required to meet their business expectations, so they used the company’s NVMe/RoCE support.

“We needed an external storage platform that matched the performance we get from internal NVMe SSDs to support a mission-critical SQL analytics use case running on VMware 7,” said Ken Boyer, director, global storage MGMT, Iqvia. “The Pavilion platform enabled us to meet this challenging requirement that could not be met with other options.”

“We have shown that ANY database, bare metal or virtualized, is faster on Pavilion than using NVMe SSDs in a server,” said Costa Hasapopoulos, chief field technology officer and WW VP, business development, Pavilion. “For testing each database, we leveraged the corresponding industry-accepted synthetic load-generation tool and compared the performance of each database on bare-metal and in virtualized environments using Pavilion’s certified NVMe/RoCE capability for Windows Server and VMware. For Linux we used CentOS and Oracle Enterprise Linux. For all tests the results were resoundingly conclusive, with up to 7X performance boost over using a server with NVMe SSDs, and there is zero virtualization tax.”

Using the company’s solutions enables customers, partners, and consultants to replace racks of database servers and minimize software license costs, reducing TCO. This makes the firm’s solutions for accelerating database operations.

Resources:
HyperParallel Flash Array DS
HyperOS 3.0 DS
Pavilion HyperParallel Data Platform for SQL Server DS
NVMe-oF Storage for Cassandra   
NVMe-oF Storage for Elasticsearch DS
NVMe-oF Storage for MySQL DS
NVMe-oF Storage for PostgreSQL DS
NVMe-oF Storage for Oracle DS