Modern Hardware with 64 CPUs

In today’s world, more often then not you might run into systems that have large beefy hardware.  Hundreds of gigabytes of memory, ultra fast solid state disks and many CPUs to push your workloads.  These types of systems are no longer unicorns in the wild but are becoming more and more prevalent.  Further more, in today’s world, it’s also very likely that you will see the different facets of SQL Server, such as Integration Services, Reporting Services, or Analysis Services split out to better manage hardware resources.  These facets of SQL Server all compete for hardware resources against the database engine so having them reside on individual servers help to ensure better performance for each of them.

However, did you know that not everything related to SQL Server can utilize more than 64 CPUs?  It’s true, look for yourself.

The chart shown below, taken from this Microsoft article, breaks it down as to which component can use more than 64 CPUs.

According to that document, the only aspect of SQL Server that can take advantage of more than 64 CPUs is the database engine itself.  If you plan on using that many CPU’s for things like Integration Services or Analysis services, you might want to think again.

Before implementing new architecture, make sure to review all of the architecture documentation.  Otherwise you might be wasting resources on things that can’t even use them.

© 2019, John Morehouse. All rights reserved.

2 Replies to “Modern Hardware with 64 CPUs”

  1. 1. I’m assuming by “cpus” you mean cores?
    2. Can we affinitize different services on systems with >64 cores? Say if SSIS or SSRS had to coexist on the same box?

  2. Hey Rob – Yeah, I’m assuming CPU = cores here and I would imagine that you could affinitize the services as you mentioned, although to be honest I’ve never tried that. I’m a fan of splitting them out completely where ever possible but sometimes that’s cost prohibitive.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.