We’re seeing an enormous growth in unstructured data, especially files that are seldom used but must still be saved. The question is where to put them. An object-based architecture is a widely talked about option, but the question then becomes how do you scale that modular storage system or upgrade it over the decades that some files will be kept?
In this ChalkTalk Video, George Crump, Lead Analyst at Storage Switzerland and NEC’s Director of Advanced Storage, Gideon Senderov discuss ways to address the issues with and solutions for scaling storage when data has to be kept for a very long time:
Most companies need to maintain these kinds of data sets up to 25 years, or even longer. Scale out systems need to be upgraded as new nodes become available, nodes with higher capacities, greater overall density and better performance. With the typical hardware refresh cycle between 3 and 5 years, “future proofing” the storage infrastructure can be a daunting problem.
Data migration is the term nobody wants to hear. The unspeakable truth is that once a data set gets into the hundreds of TBs, migration is all but impossible. It becomes the “immovable object”, from the Chinese proverb, making data-in-place upgrades the most feasible way to maintain the storage infrastructure.
Another issue is scaling the system but maintaining storage performance as it grows. Some “tightly coupled” scale-out architectures make you add more generic nodes that include both processing power and capacity, which means you have to increase both (and pay for both) when you may only need one. Many also require you to buy identical nodes, so you can’t take advantage of the improvements offered in future generations of hardware.
NEC’s HYDRAStor was designed for these exact situations, with front-end processing nodes that scale throughput and back-end storage nodes that comprise an object store to scale capacity. You can configure the initial system as you need it today and increase capacity and/or throughput by adding just those types of nodes. And, you can keep older generation nodes running as long as desired.
Eventually, you’ll want to replace older nodes with newer, faster hardware, that’s also more efficient (which reduces operational costs). When you want to retire them, the system will assimilate the new nodes’ capacity and redistribute data from the old nodes, automatically – true data in place migration. For more information, watch the above ChalkTalk with George Crump from Storage Switzerland and NEC’s Director of Advanced Storage, Gideon Senderov.
NEC is a client of Storage Switzerland