Thinking about traditional commercial storage in large data age: comparison between central storage and distributed storage

Today and our department's eldest brother 1*1, everyone face-to-face communication to the new project team ideas. But also i to the traditional commercial storage in the current Big Data cloud computing development in full swing of background, how to improve, fusion exchange my views.

Traditional commercial storage, born in the mainframe, minicomputer era. At that time, all of the hard drive concentrated on a few racks, mainframe through fiber, iSCSI and other high-speed connection to the storage system. Storage system needs to ensure high concurrency, high IO read and write speed, data redundancy, such as backup protection.

Now x86 servers, especially Google, are assembled into large clusters, Distributed file Systems (GFS,HDFS), and various memory based systems (memcached, GemFire, ramcloud) through inexpensive hardware. They are built with relatively inexpensive hardware to build high concurrency, high read and write speed, high reliability (although each node has a high failure rate, but the cluster can quickly failover this failure node and make the overall service unaffected), high scalability: this its nature.

It is clear that traditional commercial storage needs to adapt to this trend of computing development and respond.

Open-stack is favored by major storage vendors, OpenStack Cinder can easily access commercial storage, that is to say, the storage system used by OpenStack is still the major commercial storage systems. But this does not solve the disadvantages of traditional commercial storage or central storage. It is also the root knot of a large environment that is not adapted to the current cloud computing. The way out, or there is the integration of existing products, is really the use of software Defined Storage (software definition Storage), or Datastorage as a Service.

What I see now is just storage vendors passively adding APIs to passively integrate into the cloud environment. We should need to transform, truly implement software to define storage, and truly build distributed storage systems. Hopefully, these helm can grasp the direction, and hope that after this transformation, we can still guide the times, and then change the world.

See more highlights of this column: http://www.bianceng.cn/database/storage/