This Article provides information for enhancement done in vSphere 7.0U2 improving read I/O performance for VM running on  a snapshot (SE Sparse).

VMware vSphere 7.0.x

SE Sparse is the default snapshot format on VMFS6. It's also the default format for VMFS5 when VMDK is equal to or greater than 2TB in size. Snapshots are heavily used in third party backup based work flows and by third party backup solutions. These workflows typically create a single level snapshot for a short period of time and after taking the backup, the files are deleted/consolidated into the base disk.

While the VM is associated with a snapshot, IO requests from guest operating system (read and write) happen to the base disk and the IO performance can get potentially impacted due to the presence of the snapshot layer.

The enhancement described here aims at improving read performance when VM is on SE sparse snapshot. This is done by using a probabilistic data structure like Bloom Filter and is targeted to optimize read work flow especially for first level snapshot.

Impact/Risks:
The enhancement aims at reducing the read IO latency and improving read IO performance when a VM is running on a SE Sparse snapshot.

You can expect better read throughput compared to earlier releases without using additional computing cost.

The quantum of improvement depends on the nature of work load ( read vs write), number of virtual disks per VM and underlying storage (all-flash, magnetic disk , NVMe etc). Based on our internal tests we see significant improvement (compared to earlier releases) for read IO on different backend storages.

Additional considerations:

• Improvement would be seen if size of the base disk is large enough (> 500GB). For very smaller disks there will be no significant improvement.
• Improvements are done specifically for one level of snapshot. 
• Improvements will be higher if the workload is predominantly read. For example database workloads which typically have 70% read-30% write mix will also see an improvement.