When looking at performance problems traditionally storage is usually the first place many of us would look and for a good reason. In many of the environments these storage problems are caused by misaligned filesystems. Why is that? Well fairly simple, misalignment causes overhead. Just look at this diagram:
If you look at the Guest OS and the Array, you will see that in order to read 1 Guest OS cluster the Array needs to read two chunks. With a single VM running on a datastore and your array the impact might not be severe but when you have dozens or hundreds running it will be substantial while it is completely unnecesarry.
Multiple vendors have released tools to solve this issue in the past and last week Nick Weaver (former EMC vSpecialist, now EMC Office of CTO) posted his tool to solve misalignment. Not only does the tool solve misalignment it also allows you to reclaim dead space within your guest. Is that cool or what? Keep in mind that this is an experimental tool, there’s no guarantee whatsoever… but I love these type of flings / tools and I hope you would be willing to try it and provide Nick feedback so he can bring it to the next level.
