
EMC designs and manufacturers disk storage for medium to large-sized customers. For years the company sold enormous disk arrays that were partitioned into disks for dozens or hundreds of attached servers. These storage arrays often cost US$1 million or more. But with the advent of cloud-based storage, solid state storage devices (SSSD), and dropping prices, the company, like other vendors, has been forced to roll out new products to keep up with the competitors.
What the EMC disk arrays have traditionally delivered is high-availability and fault tolerance. “High-availability” means two or more disks operating in unison. When one is busy writing data, the other one can be used to read data. When both disks are busy, instead of slowing down the entire system, data can be written to cache and rewritten moments later to the second disk when the disk controller is freed up again.
“Fault tolerance” refers to the ability of computer storage to keep working when one of the disks has gone bad. In other words it keeps working even when one part of it has flown apart or perhaps has been taken down for maintenance.
To reduce the risk of losing data from these high-availability systems, these disks are configured in what is called “active-active” mode. This means data written to one is written to the other. If there is any latency in the write operation on one disk, disk caching ensures that data is not lost when one disk fails.
The proliferation of VMware machines means that computers and not just disks are deployed in active-active and failover configurations. No more do we have one server in one corner of the data center and another server located somewhere nearby. Instead one large server contains multiple VMware instances each running their own operating system and with their own memory separate from the others.
Making these technologies more nimble and dependable, VMs (or Virtual Machines) now include vMotion, which is a technology from VMware, the company which invented the idea of virtual machines. vMotion lets you move one running virtual machine from one physical server to another without any loss of operations. So they can move about in one data center or more from one data center to another.
The traditional way to protect applications from catastrophic failover has been to place servers and data storage in two different locations. This decreases the likelihood of a complete outage because if one location, say, catches on fire, the other still operates. However there is always the problem of latency with replicating data from one location to another when the data storage is running in active-passive mode. Worse is the problem of how to restore back to the original data center when it comes back online. Systems which have been out of sync for hours or days can take lots of time to restore to a state of equilibrium.
EMC says their new VPLEX data storage system mitigates these problems by configuring two data centers as if they were one. Paul Danahy, Senior Director of Global Product Sales, in a corporate video says that VPLEX, “Enables people to do what used to be done in a single data center over distance. VPLEX enables people to put data at two locations yet make it accessible simultaneously at both locations.”
With vMortion and VPLEX one entire data center can go offline and the application will continue running. How’s that for efficiency?