Slash VMware & SQL Server Costs with NVMe-oF

Smarter VMware and SQL Server Licensing

NVMe-oF and Pavilion To The Rescue

The demand for organizations to deliver more with less has never been more challenging. Challenges always land on IT and business-critical applications. A recent survey found that 80% of large enterprises use SQL Server and most have performance issues.

Latency is the bane of SQL Server. By cutting latencies 50% or more, SQL operations can recover valuable hardware and software expenses.

To understand how SQL Server benefits from low storage latency, we asked David Klee, a Microsoft Data Platform MVP, a VMware vExpert, and a 20-year consultant for global organizations. He summed up why latency matters:

David is not alone in his view. Microsoft's perspective is:

“On any server that hosts a SQL Server instance, it is very important that the server achieve the fastest response possible from the I/O subsystem.”

And VMware's view is:

“Poor storage performance is generally the result of high I/O latency.”

Typical Storage Bottlenecks and Limitations
Let’s look at the how different types of arrays will interact with SQL Server.

Direct Access Storage (DAS):
A server with NVMe SSDs will get the low latencies needed to boost SQL Server’s performance. However, limitations in the available PCIe lanes mean that most servers can utilize 6-8 NVMe SSDs at 100%. Add additional demand and the latency will increase, bottlenecking SQL Server. Every server is a single point of failure, so many customers use database-level availability technologies to keep their DB accessible during a failure. This reduces available storage by 2-3X. Enterprise storage operations (such as RAID, compression, encryption, and more) not only consume CPU, but slow SQL Server operations down resulting in IT deploying more SQL Server hosts, increasing server and licensing costs.

Software-Defined Storage (SDS):
An SDS has storage management and enterprise features, but bottlenecks SQL Server’s performance during high-IO and demanding workloads. The storage management overhead of SDS eats away at the CPU scheduling of any hypervisor. Network latency and metadata management contribute to significant penalties in write operations. Compression and deduplication not only slow SQL Server operations down, but consume more CPU, all of which collide and compete with your database and other workloads.

All-Flash Array (AFA)
An AFA with NVMe will get the low latencies needed for SQL Server. But the number of PCIe lanes on each storage controller limits the number of NVMe SSDs that can be 100% active, so writing concurrently to more SSDs will increase latency. To get around this bottleneck, many vendors write to multiple SSDs. This architecture works until the array encounters high IO or demanding workloads. You will then see latency increasing and SQL Server slowing.

Additionally, many storage vendors use a dual controller architecture, but under demanding IOs the controllers will bottleneck. What was a very low latency gets much higher, resulting in significant SQL Server performance bottlenecks.

Enter the Pavilion HyperParallel Flash Array (HFA)
To get around these limitations, Pavilion introduced a storage architecture designed like a network switch. Unlike most AFAs that have two storage controllers, it has up to 20, enabling each of its up to 72 NVMe SSDs to be 100% busy. The Pavilion HFA has ultra-low latency for reads and writes, and is linearly scalable. It reads at 100 µsec and writes at 25 µsec. Its enterprise features (like RAID, encryption, compression, snapshots, and more) result in consistent, blazing SQL Server performance while saving costs in server hardware and database licensing.

A Tool for Customers to Use:
To assist in this process, we built upon decades of configuration and optimization experience with the largest SQL Server Enterprise deployments and created a TCO tool that shows the savings in server hardware, VMware vSphere, and SQL Server Enterprise Edition licenses. Customers and partners can test drive the downloadable TCO tool.

By generating TCO results and visual graphs customers now have choice and control to optimize SQL Server and VMware storage. Once populated, this will show the value of Pavilion compared to alternative configurations in virtualized SQL Server environments.

If you have any questions about how to use the tool or to understand more about how the most performant, dense, scalable, and flexible storage platform in the universe can help you and your organization shatter expectations, simply reach out to Pavilion by filling out our Contact Us form or emailing us at We’d be happy to assist you!

Analysis by:
David Klee, Costa Hasapopoulos, and Gary Archer