This article is more than 1 year old
Scale Computing's SCRIBE vows to lob virty storage kit into the bin
Physical arrays too. Watch out EMC, we're tearing through those storage layers
Start-up storage shindig Scale Computing has crafted SCRIBE, an object store directly accessible by a hypervisor that replaces physical storage arrays and virtual ones (VSAs) like VMware's own vSAN.
SCRIBE, says Scale, is an architecture for expanding hyper-convergence to a complete, distributed software-defined environment. It creates a storage layer, which a hypervisor uses to access storage objects and deliver them to applications.
According to the company, the complexity of "storage protocols, file systems, and the overhead of VSAs are completely eliminated" thanks to SCRIBE. Existing physical storage arrays' capacity can be pooled with this object store, allowing for migration for customers using such arrays.
CEO Jeff Ready said: "Protocols, file systems and VSAs represent the old way of thinking. They are at the core of other attempts to create next-generation storage, but these are the very obstacles that prevent the true implementation of a software-defined environment.
"SCRIBE eliminates these,” Ready continued, “and instead makes storage directly available to applications, with performance, redundancy, and locality parameters set on a per-object basis. SCRIBE abstracts all of the complexity of both the storage and the infrastructure overall, delivering a platform that eliminates the complexity of traditional VMware environments and radically simplifies IT for our customers.”
To understand where SCRIBE is coming from, a look at Scale Computing's existing HC3 converged system is a good idea.
HC3 system background
Scale Computing has converged virtualised compute and storage system, HC3, was launched a year ago, and revenues for it have grown 65 per cent quarter-over-quarter since then. It's for small and medium-sized organisations and converges storage (1), servers (2) and virtualisation (3), hence Hyper-Converged 3, or HC3 for short.
HC3 nodes have 4-core CPUs and 1GbitE or 10GbitE connectivity. The storage is based on numbers of 7,200rpm SATA disk drives in 500GB, 1TB or 2TB capacities to build 2TB to 24TB systems. There can be eight nodes in a cluster and up to 100 VMs in a cluster.
An HC3x product uses 6-core CPUs, still limited to an 8-node cluster, with 10K or 15K SAS disk drives with 600GB, 900GB or 1.2TB capacities used to create 2.4TB to 14.4TB systems. This is the high-performance system. No SSD availability is mentioned, which seems an obvious capability to add in a future release.
HC3 systems do not require a NAS or SAN storage layer because they contain server virtualisation software on every node with access to a single pool of storage. That pool spans multiple HC3 nodes, yet looks like local storage to each node's hypervisor. The storage is easily managed as local, direct attached storage for your applications; no more storage subsystem, SAN, or storage protocols to manage.
The idea is to converge the many components of local and converged storage and virtualised compute into clusters of single boxes running Scale Computing's ICOS software. Storage and compute capacity is increased by adding nodes.
Storage features include:
• VM failover/restart upon node failure
• iSCSI, CIFS, NFS storage access
• Thin provisioning, space-efficient snapshots
• Cluster-to-cluster asynchronous replication
The boxes run a built-in hypervisor and thus the traditional VM-with-apps-inside scheme. VMs can be migrated between cluster nodes. Scale Computing states: "The hypervisor is seamlessly integrated with HC3. You do not need to purchase virtualisation software or licenses from any vendor, and you do not need any prior knowledge or training in virtualisation to use the HC3."
Virtual machine creation is done through the HC3 Manager's web browser interface. Here's a video of Scale's product manager Craig Theriac explaining and demo'ing VM creation, and mentioning the use of an ISO disk image file:
HC3 uses the KVM hypervisor, El Reg's storage desk understands:
HC3 is based on Scale’s ICOS – an advanced file system which is based on the foundations of IBM’s GPFS. ICOS combined with KVM, enables HC3 to deliver a highly available architecture that stripes and mirrors data across each node in the cluster. If a drive or node fails, the system continues to run.
Scale warns that customers using VMware face an EMC trap: "It’s quite probable that EMC could create a hyperconverged version of VMware that only runs on EMC storage gear. All other vendors could be locked out, severely limiting choices for vendors, resellers and especially users."
In our humble opinion, that seems rather paranoid.
HC3 looks like a Nutanix and Simplivity competitor. Possibly SCRIBE could run on HC3 boxes, or it could be made available as a hypervisor plug-in; Scale isn't saying yet. We'll have to wait and see.
A technology preview demonstration of SCRIBE took place at a San Francisco hotel at the same time as VMworld. The product is in beta test. Read Scale Computing's blog to get more background info. ®