SAN FRANCISCO -- Scale Computing today unveiled the HC3, a scale-out converged storage system combining compute,...
server virtualization and capacity in one box.
Launched at VMworld 2012 here, HC3 runs in clusters of three or more nodes. Each node includes Red Hat Inc.'s KVM hypervisor built in; no additional virtual machine licenses are required. With the HC3, Scale is targeting midsize companies with small IT staffs, according to CEO Jeff Ready.
The HC3 converged storage system is the same form factor as Scale's M-Series scale-out multiprotocol storage system with compute built in so no discrete servers are needed. HC3 systems support iSCSI, CIFS and NFS storage, as well as Gigabit Ethernet and 10 GbE networking. Customers must use separate switches, but future systems will support built-in networking, Ready said.
HC3 is available in three-node, four-node, six-node and eight-node clusters. A customer who buys a three-node cluster can add a single node to the cluster. After that, upgrades are done in paired clusters. Each node has 32 gigabytes of RAM, so a three-node cluster has 96 GB of RAM and an eight-node cluster has 256 GB. Each node includes a quad-core Intel CPU, two GbE network ports and one 10 GbE network port. Each node also includes up to four hard drives, either 15,000 rpm SAS or 7,200 rpm SATA drives. Usable capacity ranges from 12 terabytes on the three-node system to 32 TB on the eight-node system.
A three-node system handles as many as 30 virtual workloads, and an eight-node cluster supports 100 virtual workloads. Pricing begins at $25,499 for three nodes, $33,999 for four nodes, $50,999 for six nodes and $67,999 for eight nodes.
The HC3 is being launched without solid-state storage, a rarity these days. That is done to keep costs down, and Scale is counting on HC3 driving enough performance from its built-in compute to get enough speed with SAS and even SATA drives to get away with not using flash, Ready said.
Adding an HC3 node scales computing power and storage capacity. The vendor may eventually add nodes focused more on capacity or compute for scaling one or the other, but today customers cannot scale capacity or compute independently, Ready said.
Ready said the HC3 alleviates the need to install virtual machines (VMs) and applications on servers, connect servers to SANs or networked-attached storage (NAS) filers, then manage the storage separately. Virtual machines and applications install natively on the HC3 clusters. For high availability, customers can migrate VMs across nodes of a cluster to maximize application uptime; failing nodes will automatically move VMs to another cluster.
HC3 is generally available today. Scale had about 20 beta customers using a beta version of the system, Ready said.
Arun Taneja, president of Hopkinton, Mass.-based analyst firm Taneja Group Inc., said HC3 is among a new category of systems he calls hyper-converged. These systems include storage, compute and virtualization in a DAS-type implementation. They simplify managing storage for VMs, a process that confounds many administrators.
"Storage companies are having this 'aha' moment as they realize there are so many compute cycles available that they can’t use them all in storage anymore," Taneja said. "They're all trying to figure out how to use those cycles. These three guys are saying, 'With virtualization, I can bring compute, storage and networking together in such a concentrated way that I can get maximum performance. '"