Today marks the end of my third month at Concurrent and my first post for the Concurrent blog. It has been a hectic few months, with a lot of development meetings, positioning meetings, sales meetings, meetings about meetings, and even a trip to VMworld. In my contributions to this blog you can expect to find a combination of opinion pieces, technical and architectural pieces, case studies, and other random odds and ends. This first post is all about scale-up and scale-out storage, and the total cost of ownership for each. A lot has already been written on this topic, so I will try to cover new territory here.
One of the most overlooked costs of owning storage is the migration process from an old array to a new array. According to Wikibon, the cost of migration can be as high as 17% of the total cost of the storage system. This means that for every $100,000 spent on an array, including maintenance, you will spend $17,000 to migrate to a new system. In a paper by Hitachi Data Systems, the cost to migrate 1 TB of data is $15,000. No matter whose figures you use, the cost of data migration is very high.
Analyst firm IDC states that storage accounts for 60% of all large IT migrations. Migrations not only account for more than half of large projects, they also create a risk of data loss and application downtime. Because of this risk, it is typical for the old array and new array to be deployed at the same time for a few months. During this time, the new system is running while the old system is standing-by as a back-up. This means that you are paying to power, cool, and maintain two systems.
Scale-up arrays are the foundation that enterprise storage has been built on for decades, but they are approaching the end of their useful life. A scale-up solution is limited by the performance of the controllers sitting in front of the storage. For example, if a system has 32 Gbps of throughput and supports 150,000 IOPS, that is all that you will ever get out of the system. At some point, the capacity of the system will exceed the performance of the controllers. When this happens, it is time to get a new array and start the data migration process.
If you compare this to a scale-out array, one of the primary differences is readily apparent, scale-out systems can grow capacity and performance. With a scale-out array, each node adds not only capacity, but also network throughput, CPU, and RAM. Scale-out storage has been around for many years now, but one of the limitations has been the overall scale. Smaller scale-out systems are beneficial for small environments. However, when you reach the limitations of a small scale-out architecture, you have to begin the same costly and risky data migration as a scale-up system when it reaches its end of life.
A next generation scale-out storage system has the ability to not only add capacity and performance, but also scale to very high levels of capacity and performance. In the case of Concurrent’s Aquari storage system, we can scale to Exabytes of usable capacity with Tbps of available throughput. Yes, you read that correctly, Terabits of network throughput (not Tablespoons.) True scalability is the benefit of a scale-out architecture. Add capacity and performance when and where they are needed. This happens in the same cluster, using a single management console, no matter how much capacity is deployed.
Array migrations are a legacy of the traditional scale-up storage world. Next generation scale-out storage expandability, means cluster to cluster migrations are not required for capacity, performance, or hardware lifecycle management. Removing the need to migrate data to a new array saves money, time, and perhaps most important of all, reduces the risk of downtime and data loss.
And we have a winner….Scale-Out Storage.
There is no argument, scale-out is the future of storage. However, a scale-out system also needs a next generation architecture that can grow with your business. After all, the business that has less data today than yesterday is one that will soon be closing its doors.
Stay tuned for more information from this blog, some interesting, some technical, and even some rambling. In the coming weeks you can find more information on Ceph, the heart of Aquari, erasure coding vs. RAID, and probably some highlights of upcoming tradeshows and events.