There are many storage technologies available today from various vendors. The storage technology and configuration best for your application depends on the application access patterns and workloads.
The attached document discusses the various design considerations and recommendations for various technologies. This guide is to help you during discussions with your storage vendor to determine the appropriate storage technologies and products that will work best to meet the performance goals for your applications.
Thanks Mark. Cool.
The document recommends a range for Max Response Times but certain operating systems like Red Hat do not report Max times instead they only report average times. How do we achieve these recommendations if we can’t compute them in the first place? Are there any third party tools used by InterSystems to measure Max Response Times on Red Hat?
Thank you for your comment. You will need to establish you own monitoring and ultimately range of IO response times for your application using tools like iostat. This article is used to give you a starting point for monitoring. Your specific application may need higher or lower requirements.
Using iostat, you want to continuously monitor storage device performance (specifically the iostat -x <device> <time between sample in seconds> <number of iterations> command) and monitor it for a particular range of time. For example, if you want to only monitor during peak business hours from 8am-12pm. What is mostly important is average response times - typically I like using iostat -x <devices> 2 1000 to report 1000 2-second samples. This is useful when diagnosing a performance issue.
To reduce the amount of data collected you can use a higher time between samples such as iostat -x <devices> 5 1000 for 5 second samples or even higher if you wish. It's really a function of what reasons you are monitoring - if doing an in-depth performance analysis you would want a small time between samples to better observe spikes in response times, or if you are doing just daily statistic collection you could go for a higher time between samples. The objective here is to get familiar with your specific application's needs and this article just provides a baseline for what is typical for most applications.
Kind regards,
Mark B-