Thursday, 18 November 2010

How To Get High Performance Cloud Storage

One of the challenges with cloud storage is the connection between you and the storage. For almost everyone it is going to be slower than what is available within the data center. This performance difference does not mean a more limited use of cloud storage, it means that greater intelligence is needed to load data into the cloud. With that intelligence cloud storage could be leveraged for even the most demanding of applications.

In almost all use cases, but especially cloud storage as part of a primary storage solution, it is going to require some sort of local presence to cache the active data sets. This local presence can come in the form of a stand alone appliance, a virtual appliance or can be integrated into the storage system itself. The goal of the local presence is to store the active data subset on local high speed storage and then as the data ages push it out to the cloud storage service but do so transparently.

This hybrid type of deployment does mean that the data set does have to be something that can be segregated by access dates. It also means that the ideal data set is one where it has a short create and edit cycle, then is rarely accessed in the future. A file server is an obvious example but messaging and group collaboration tools are as well.
No matter what the local data set is you are always going to need to copy data to the cloud. Most of these hybrid type of solutions will want to copy all new or modified data the moment that the change occurs, this provides a level of redundancy from a data protection perspective but means that the WAN bandwidth utilization is upfront as well. Most of these hybrid type of devices can trickle data to your cloud provider so bandwidth can be throttled back. More importantly most of them have some form of WAN optimization either compression or deduplication that reduces the data set before it is sent and after it lands. For example in the solution we are currently testing, while we have placed 77GB's of data in the cache device only 27GB's of that data has been actually transferred and stored thanks to compression and deduplication.

Even with this intelligent use of the available bandwidth there are some practical steps you will want to take. First you need to have a decent connection to the internet. When we began our testing we immediately found our connectivity to be a little lacking. We doubled our bandwidth for about 15% extra per month and it made the application significantly more usable.
Second you also want to select a data set that can be gradually migrated to the cloud, net new projects are ideal or data that can be easily isolated by age, migrating the oldest data sets one at a time. In our case if we used it for hosting our various projects and simply decided that all new projects would go on the cloud storage appliance. As a result we have seen almost no impact from having all of our data be on the cloud storage device and we have seen a performance improvement in local response time since our appliance is on high speed storage,
With these considerations in mind cloud storage can be a viable option for many applications and data sets. As the hybrid technology continues to improve and the cost of bandwidth continues to come down even more applications and data sets will be deemed cloud appropriate, but the time to develop a cloud storage strategy is now. We will be answering questions like this in more detail in tomorrow's live webinar "What's Your Cloud Strategy, Answering The Top Ten Questions".

Join Us:

1 comment:

cloud hosting india said...

Really excellent blog. A benefit of this type of hosting service is that a large number of software can be downloaded at the same time, without disrupting the browsing experience of the users.