The Cloud Needs Fast, Scalable Storage

While cloud storage providers need scalable, inexpensive but reliable slow storage, cloud compute providers typically need fast, scalable storage. In this case you have potentially thousands of users, asking potentially hundreds of users to request data from potentially one massive storage array and it can often become the bottleneck.

George Crump

June 24, 2009

2 Min Read
NetworkComputing logo in a gray background | NetworkComputing

While cloud storage providers need scalable, inexpensive but reliable slow storage, cloud compute providers typically need fast, scalable storage. In this case you have potentially thousands of users, asking potentially hundreds of users to request data from potentially one massive storage array and it can often become the bottleneck.

As we discuss in our article Searching for High Performance Storage, cloud compute is a classic case of a multi-tenant environment. It can generate enough Storage I/O requests to sustain typically very high drive counts. What is needed is to make sure that the file system in the case of NAS or the stripping pattern in the case of block storage can allocate across a high drive count without having to resort to a series of small file systems or short stroking drives. Both of these work-arounds add complexity and cost, both of which are considered criminal in most cloud providers.

The challenge with adding drives is that at some point you might saturate the controllers in the storage system so you either need to buy a much higher end storage system with controllers that can handle the future workload or have a storage system that can add to the processing power of the controller. In the world of the cloud where they bill in a pay as you grow model, paying upfront for this kind of storage I/O processing power is difficult to justify and that is why manufacturers that have the ability to add to their storage processing power are having such success in this market. These storage manufacturers essentially have a growth model that matches the cloud compute provider.

The classic use case is a provider that is offering a software application that can do CRM, Accounting or some other business function, but it is not limited to these. One of the most frequent types of companies that are pushing storage IO to its envelope are companies that offer online gaming. For example if 1,000 users needs to simultaneously check their "weapons cache" that is a lot of storage IO requests. Moving this to a more responsive platform can increase gamer response time and thereby increase customer satisfaction. Satisfied customers tend to come back and play more often and renew their subscription.

From a user perspective, one of the interesting aspects about cloud services is the relatively short term nature of the relationship. If you don't like the service, other than data transfer (which could be big) you don't have a huge investment in physical hardware or software. For cloud compute companies, customer retention has to be priority number one.  As a result cloud compute companies need to be attentive and possibly prepared to invest more dollars in their storage architectures than do cloud storage companies. In their world the basics (data protection) are important but response time is critical and may make the difference between an organization that renews and one that moves on.

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like


More Insights