Scale Computing: New Twists To Scale-Out Storage For The Mid-Market

Posted by David Hill
January 27, 2012

Startup Scale Computing delivers scale-out, unified storage for the mid-market, meaning users can access SAN/NAS resources from the same, scalable pool of disk storage. Scale Computing is by no means alone in doing that, but the company goes beyond just delivering storage in a box to delivering a data center in a storage box. And that is very interesting.

Scale provides enterprise-class storage based on IBM's General Purpose File System (GPFS) at SMB prices for smaller IT organizations made up of IT generalists and not specialists. These organizations can be either SMBs or departments, such as a developer group or research team, in a larger organization.

Scale's architecture builds on the concept of self-contained yet unified storage nodes, where each includes four disk drives and the associated processing power required to integrate into the larger storage architecture. The nodes can be clustered, allowing them to act together as if they are one. This means that capacity and performance can scale as more nodes are added. It's a unified approach because iSCSI, CIFS and NFS protocols can be accommodated in one storage pool. Scale supports valuable storage management capabilities, including thin provisioning, snapshots and replication.

IBM's GPFS gives Scale built-in features such as high-availability/reliability, which translates into the elimination of production outages and enables non-disruptive maintenance and capacity upgrades, improved performance and added functionality such as seamless capacity expansion and an extensive management and monitoring infrastructure to simplify file system administration.

One problem that mid-market customers might encounter with a powerful clustered file system is having far more capabilities available than they would ever need. After all, most users of storage software use only a fraction of the available functions, although some may use a particular function or feature that few others use. Although ease of use is welcome by all, IBM GPFS clients typically have specialized knowledge, whereas Scale's users are typically generalists. Scale's value-add is in making GPFS easy to use for mid-market generalists.

Putting a Data Center in a Storage Array Box
Each of the company's storage nodes is powered by an Intel general-purpose CPU running Redhat Linux-based software. The storage system demands on the CPU use only a fraction of its overall capacity, so Scale lets you run a KVM server hypervisor with guest operating systems and associated business applications. As such, Scale's storage node CPUs can be used for both storage processing and general-purpose computing.

Downscaling GPFS is a useful way of providing key enterprise capabilities for the SMB market while also providing computing resources. Scale Computing takes innovative advantage of the unused CPU capacity in each storage node to provide general-purpose computing (including server virtualization) in the storage box. All in all, Scale Computing is giving mid-market storage customers and enterprise vendors like IBM a lot to think about.

At the time of publication, Scale Computing was not a client of David Hill and the Mesabi Group.

Related Reading

More storage-networking-management Insights