This is a great and common question, as there are many different myths and opinions as to how much disk space you...
should leave free.
What makes it such a slippery and evasive topic is that the answers will vary based on operating system, application, server and storage system type, configuration, personal experience and general industry folklore. For example, if your Windows-based laptop starts to slow down at 50% storage capacity usage, you might assume that all Windows-based systems slow down when 50% of their disk capacity is used, which may not be reality. The reality is that the performance will vary depending upon how the system and storage are configured, as well as how data and applications including the operating system utilize the disk storage that is available.
Monitor your systems and note how much free space is available at various times, as well as how the systems perform under different conditions. If you defragment and optimize your disk drives, you may get better performance as they fill up.
Many disk storage vendors, not surprisingly, will suggest that you keep your disks at or under 50% utilization as a conservative guideline. Disk capacity usage numbers in the 70 to 80% plus range are not uncommon for some Windows and open systems environments, depending on workload and I/O activity. For example, a database system may show disk usage in the 70 to 80% used range; however the database itself may not be using all of that space.
Take I/O activity into consideration when looking at how much free disk storage capacity to have as well. So if you want to be safe and have the disk space along with the budget, keep utilization low until you can determine for your self what the best rule of thumb is for different disks and applications. As you gain experience and can tune your systems, applications and storage, you can evolve to higher storage capacity utilization.
Greg Schulz is founder and senior analyst of The StorageIO group in Stillwater, Minn., (www.storageio.com) and author of the book Resilient Storage Networks (Elsevier).