SQL Server Performance tuning is always important and I can list off reasons why:
- Improved application response.
- Identifying issues before they become outages.
- Potential cost savings.
In terms of the cost savings however, that is always a little speculative when running on premises. I can take your 8 core box running at 70% CPU peak and get that down to 35%. Then you could potentially migrate to a smaller box with 4 cores and save up to $7,000 a year for standard edition or about $26,000 for enterprise. If you’re running on bigger kit then the potential savings are much larger. Often though SQL Server runs on physical machines so downsizing isn’t that straightforward and there isn’t the appetite for migrating to different kit.
In the cloud it’s different, you can change the size of the instance running your SQL Server quickly and easily with no real risk, so cost savings can be realised immediately. That applies whether you’re running SQL on cloud Virtual Machines or using a Platform as a Service (PaaS) offering such as Azure SQL Database or Amazon RDS. If you’re running the serverless option on Azure SQL Database then you can start seeing savings without even having to change configuration of the instance.
Let’s look at some figures showing the potential cost savings. The table below shows the cost of running 4, 8 or 16 vCPU instances in Azure and AWS. For Azure I’ve chosen provisioned compute on the General Purpose Service Tier running on Standard hardware. For AWS I’ve chosen M5 as the instance type with SQL Server Standard edition and Multi-AZ for high availability. The figures shown are on-demand costs per year:
|Number of vCPUs||Azure SQLDB||AWS RDS|
You can see the pricing calculators I’ve based these figures on here:
As a side note, we can see that AWS is a lot more expensive. There are a number of reasons why I see Azure as a better fit for running SQL Server workloads and price is one. That said, the database isn’t the only factor in choosing your cloud provider so it’s perfectly legitimate that you might be running on AWS.
When I’m asked to look at performance of a SQL instance, it’s rare that I can’t knock performance in half and often we can achieve much more. You can see the potential cost savings are significant. One large project I’ve worked on recently called me in where the application was struggling to perform running on 32 vCPUs in AWS RDS. I identified changes that meant it was performing perfectly running on 16 vCPUs, and soon after we were able to move it down to 8. A few more rounds of optimisation and now it can run on 4. That represented a saving of over $150,000 annually.
Get in touch today to discuss how I can help you with your cloud SQL Server performance and save you a lot of money year on year.
One thought on “The Importance of SQL Server Performance Tuning in the Cloud”