Algorithm that limits disk space usage in pure SSD SAN solutions

04 Dec

With cheaper SSDs, consumers have more choices of SSD SAN and many traditional and upstart SAN storage vendors are competing to offer a better SSD SAN product. Features like increased 100,000 (32K) IOPS versus traditional SAN and high compression and deduplication ratio, meaning a 10 TB of SSD storage can be as effective as 30 to 40 TB are no common. Those are, of course, vendor literature.

Recently we found out that despite the promise 3-4 times more virtual storage, there are limits. For example, a vendor may promise up t0 40 TB for every 10 TB of SSD. In an actual environment, one may see performance degradation as soon as 50 – 60 % of the physical SSD storage space is utilized. A few reasons like the current limitation of CPU speed and resources, the vendor’s own algorithm for compression and de-duplication, speed of internal memory and the SSD themselves comes together to form a limitation of usage. Basically, the more data you have, the more work have to be done to compress, uncompress, de-duplication and recollect the data by the CPU, the memory subsystem and the physical disks.

The case study: 

Having deployed an SSD SAN storage solution as datastores to our VMware estate not too long ago, we are indeed seeing improved IOPS performance in the estate, especially within  VDI estate.  We also saw good compression and dedup ratios. However, the performances from the SAN hit a bottleneck when we see physical used disk space hitting around 60%. At 60% disk used,  CPU utilization on all the array heads hit 100% constantly, especially when we try to deploy VDIs attempting to us more disk space above that 60% threshold. This brought the VMs connected to the affected datastores in the estate almost down to its knees.

After investigation with the vendor, it was determined that this is the real world limit of the product for our estate. This may not be the real world limit for the product itself as this limit will depend the kind of data and usage. This means we can effective use approx. 50% of the available storage per array only.

This by no means represents all SSD SANs out there and the limit may vary from vendor to vendor. And it may reduce over time when with fast CPU and better algorithms and even faster SSDs. The lesson to takeaway here is to ask your vendor about this next time they are pitching their solutions to you. And this serves as a cautionary tale for vendors too.

Note: As a result of this, the vendor had to give us a 50% discount on our next purchases because, literally, we can use only 50% of the promised disk space.

Leave a comment

Posted by on December 4, 2015 in Operations, vmware



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: