With so many tools and features from many different vendors, it is almost impossible to research them all and find useful tools to make your job easier. Some features also provide a faster/cheaper way to solve common problems without spending a fortune, unfortunately, these “Golden Nuggets” are often underutilized. I’ll post a few quick tools that may make a big difference in someone’s environment. As always, test before deploying to production.
One of the cool features introduced in vSphere 5.5 was vFlash, which replaced swap to SSD from previous versions, but I won’t get into that. Essentially, this is flash-based read cache on the host that functions at the vmdk level for a specific VM. This feature works by adding flash-based resources such as PCIe cards or SSD drives to create a vFlash pool of resources at the host level, and configuring the amount of storage to be used for host swap cache. Such cache is placed on the data path of the vmdk between the host and the storage array.
Once the host is configured, you can expand the virtual disk of a VM’s properties in the Web Client and assign the amount of cache for that particular vmdk, as well as having the option to select the block size (4KB – 1024KB). So, for each pool, chunks are carved out or reserved for a specific vmdk on the host where the VM is located.
As far as data locality goes and features like HA, DRS, vMotion; it is possible to migrate the cached data to another host while migrating a VM, as long as the other hosts have also been configured with vFlash. You may also specify not to migrate the cached data during migration.
Requirements:
- Check HCL for compatible Flash devices
- vCenter 5.5 or later (VCSA or Windows)
- VM hardware version 10 or later
- vSphere vMotion if using DRS
- Requires vFlash on hosts within the cluster
Implementing vFlash can be beneficial for resolving or minimizing performance degradation for read intensive applications, or simply by utilizing local resources at the host level for read cache instead or in addition to storage read caching solutions. Having local cache eliminates the “extra hop” on the network to get to cached data at the storage array.
This is a high level view of vFlash but in my opinion, I think this is a nice feature that can get rid of some headaches and fire drills.
Image source – VMware doc (Rawlinson)