Webinar Follow-Up: Setting SQL Server Performance Baselines and Alerts
In January, I presented a webinar on Setting SQL Server Performance Baselines and Alerts with my esteemed colleague Richard Douglas.
If you haven’t already viewed the webinar, click here to view it on-demand for free.
During the presentation, Rich and I mentioned the following resources that can help round out your knowledge and dive deeper into baselines and alerts.
- Performance Baselines and Benchmarks for Microsoft SQL Server
- Alerting, Baselines, and Standard Deviations
- Investigate TempDB Like Sherlock Holmes
- Multiple Baselines & Power BI
- A Million Little Things: Global Baselines
- Troubleshooting SQL Server Performance
Q: I’m interested in SQL Sentry Essentials. Where can I learn more?
A: SentryOne SQL Sentry Essentials is our new, low-cost monitoring product. You can read more about it in my blog post. The advanced baselining features we demoed during the session are built into this product (as well as SQL Sentry and BI Sentry).
Q: Can you share the URL for your metrics document?
A: Check out our Performance Baselines and Benchmarks for Microsoft SQL Server white paper.
Q: How much resource does SQL Sentry, DOC xPress, and DBA xPress take on the SQL Server? Or do you recommend them on a separate server?
A: SQL Sentry’s resource overhead is well-documented at around 1%. Our white paper detailing our overhead is the only one of its kind in the industry. DOC xPress and DBA xPress are client apps and typically only query system tables. They might provoke a brief spike in resource consumption, but they don’t incur any long-term overhead.
Q: Not sure if I misread, but shouldn't SQL Sentry Essentials be free for 5 clients, or something like that?
A: The client application is free for SentryOne products, such as SQL Sentry, BI Sentry, and SQL Sentry Essentials. You only pay for the targets you monitor.
Q: When SQL Server returns "has encountered X occurrences of I/O taking longer than 15 seconds," where should the troubleshooting begin and what possible matrices should be used, if any?
A: You’re in luck! I used I/O troubleshooting as the main example in the webinar that's discussed in this blog post. (The blog post also includes links to T-SQL I/O tuning scripts). And the eBook Troubleshooting SQL Server Performance covers my general approach to troubleshooting performance issues on SQL Server.
Q: If I have SentryOne, do I need anything else for baselines (i.e., does it replace perfmon)?
A: SQL Sentry takes care of everything. You wouldn’t need any other tool.
Q: How long should you run baselines?
A: Great question! Like most technology questions, the true answer is “it depends.” For many situations, the best baseline would span the duration of a representative workload. However, you could collect a moderately useful baseline in only an hour, during your workload’s peak usage. If your application is heavily used during business hours, then you should at least collect a baseline during those hours. Note, however, that the results of that baseline would only be representative for that time span. So, you might run a baseline for a full week or even a full month with a different polling frequency for each, when collecting the performance metrics, so that you could see the normal peaks and valleys of usage over those longer trends. Although this can be a tough project to organize and run using only native tools, SQL Sentry can create a custom baseline over any time span with two clicks of your mouse. Hope that helps!
I’ve been working with the SentryOne marketing team to build out our 2019 webinar calendar; we also have lots of in-person events in the works. You can check out our webinar and event calendar here.
If you haven’t already, please connect with me online!