Yesterday, Storage Switzerland was at EMC for a day full of briefings from various groups within EMC. First up on the agenda within the Emerging Technology Division was ScaleIO, a company that EMC acquired earlier this year. ScaleIO allows EMC to provide a storage solution to customers interested in a converged infrastructure solution. The integration of ScaleIO into the EMC product portfolio is a good example of EMC not resting on their laurels. They are a company that pushes itself internally more so than being pushed by outside competition.
ScaleIO is a software application that aggregates the internal storage inside a variety of server platforms to create a shared pool of storage capacity. The software can run on a bare metal Linux or Windows server or as a virtual machine in a VMWare, Hyper-V or Xen hypervisor platform. With any combination of deployment, a nodal architecture can be created which enables the other VMs running in this infrastructure or external bare metal systems to access the shared storage pool. The result is a converged infrastructure which doesn’t require an external SAN.
The servers that run the software can be any x86 or ARM server class system from a variety of vendors. Internal server storage may include any combination of hard disk drive (HDD), flash drive (SSD) or PCIe SSD resources. The drives can be grouped by type into pools and assigned out to VMs according to the class of service required by the underlying business application.
Volumes are then created on these pools and are presented to the connecting servers or virtual machines as block devices; the solution does not provide file or NAS capabilities. As data is written to the volumes, it is segmented and each segment is distributed across all the nodes in the cluster. At the same time, a second copy of each segment is created but the system has the intelligence to make sure that no two segments are on the same node. This establishes a policy based protection scheme. Any time a node or drive fails, the segments that were on it are recreated and auto-load balanced on another available node.
From a storage software features standpoint, ScaleIO seems to hit all the needed check boxes. The latest release added writeable snapshots, multi-tenancy, encryption, multi-tiering and quality of service. While auto-tiering of data is not available, that is typically not a requirement for Scale IOs target end users – Managed Service Providers (MSPs) and Web Scale clients. For example, MSPs typically want to hard set customer data based on subscription levels that tie to a specific service level. Where there is a need, however, to service hot data into a cache pool, ScaleIO can be front-ended with EMC’s XtremCache technology.
As the name implies, the solution offers incredible scalability. It can start with as little as three nodes and scale to thousands. In fact, EMC recently leveraged a public cloud compute provider to create a 1,000 node cluster in the cloud. They were able to achieve 1 million IOPS with that 1,000 node cluster through the use of very slow hard drives. Interestingly, EMC claims they could have continued adding nodes but instead decided to limit the proof-of-concept to a 1,000 node test.
This solution may appeal to WebScale and Service Provider Infrastructures that need a cost effective and flexible infrastructure. It allows these environments to scale storage at the same pace and within the same infrastructure as they do compute. Eliminating the need for a separate storage network and storage system should eliminate a lot of complexity as well as drive down costs.
Storage Swiss Take
It will be interesting to see how these solutions work their way into more traditional data centers. Something like a ScaleIO could appeal to the very small data center just getting started with virtualization as well as to the very large data center. It seems, however, that ScaleIO is a natural fit for online application and service providers who need a cost effective and scaleable storage solution.