High speed storage is back in the VMware Host with both HCI and Host Side Caching, but the similarities end there.
The main advantage that hyper-converged infrastructure(HCI) has over traditional SAN based infrastructure is that HCI has put high speed storage back in the compute nodes. This is also true for Host Side Caching software, with the added benefit that host side caching maintains the flexibility that SAN based infrastructure always had over HCI, that of being able to scale and do maintenance on compute and storage hardware independently of each other.
Other pros of host side caching versus hyper-converged infrastructure are:
- Ease of deployment. Hardware refresh not needed. Cheaper.
Host side caching can be deployed in minutes and without any maintenance mode/downtime for VMware or storage. It only requires an SSD (or spare RAM) in the ESXi host. On the other hand, deploying HCI is more complicated, and most likely will require new server hardware.
When using host RAM, Host Side Caching is the fastest storage on the market.
This is simply because there are no HCI solutions on the market that are capable of using large amounts of RAM in the storage IO path. And RAM is the fastest storage media there is.
Faster recovery from failure situations.
In HCI, because the same physical server is responsible for storage and compute operations, host failure or maintenance mode also results in degraded storage, and vice-versa. Host shutdown takes longer since all the local data needs to be redistributed to other hosts. In general, recovery times are longer. On the other hand, in case of host side caching, since in-host cache has transient data only, with the backend storage array still being the system of record, recovery times are not affected.
Less vendor lock-in.
SAN based infrastructure has always had less vendor lock-in than HCI because compute connects to storage using standardized protocols like iSCSI or FC, as a result, a storage array or server vendor can be swapped for another. Host side caching doesn’t change this dynamic. However in HCI, since vendor specific protocol is used to connect locally attached storage to ESXi, you are locked-in to the HCI vendor for both storage and compute. So if you want to add another HCI node, you’d need to buy it from the same HCI vendor.