Scale computing are starting to make inroads within the SMB storage market and with good reason too so I thought I’d put together a little scale computing review to help you decide if this storage is for you. They position themselves as true scale out storage in that when adding additional capacity to your array you also add network, cpu and memory capacity too. They provide iscsi/nfs/cifs services along with snapshots/replication within the box without any extra cost.
When you add storage to a traditional array such as the VNXe by EMC you essentially add another tray of disks and plug them into the back of the VNXe. This means that the additional storage is still using the same storage processors, same memory, same cache same everything which will be pumped harder because all that additional storage is going through the same channels to get out to your targets.
Scale storage has a few different variations the entry level being the S1 cluster. When you purchase scale storage you buy 3 nodes and additional nodes can be purchased separately. Scale storage also differs in other ways in that the S1 cluster has 2 network adapters one for the LAN connection and another for backend duties. M class clusters has 4 network adapters with a bonded pair for each LAN and backend connections.
The LAN connection sensibly enough plugs directly into your production network and must be routable to your DC’s etc and the internet. This is where the scale storage breaks away from traditional storage in that with a comparable device such as a VNXe your storage processors which connection to your ISCSI/ NFS targets are isolated entirely from the LAN and you have another Network port for management duties only that you connection into your production network.
The backend network adapter or bond is used only by the scale cluster for inter-device traffic and striping of the data across the nodes so they should be isolated either on a seperate switch or vlan.
The clusters use virtual IP’s that you need to use in conjunction with a round robin dns entry to provide HA with the failure of one node in the cluster meaning the virtual IP simply hops onto another node and the target is unaware of any issues.
Scale storage fits itself in a great position of the market in between that of a readynas device and the base models of VNXe’s and other vendors SMB offerings but with the benefit of true scale out expansion and no need to pay those added extra licences for things such as replication that are chargable on many other vendors products.
As of this blog post 20/4/12 they are also have a great deal on a cluster and Veeam bundle saving a whole stack of cash so SMB’s can save even more on their new virtualization projects.
Scale are also putting through beta a storage array with upgraded processor and memory capability which also has the KVM hypervisor on the array essentially providing a total virtualization solution within a the cluster, should the price point come out similar then it will be a tempting array indeed.
Author: Dale Scriven