New Scale Computing HC3 flash arrays hit the market this week as part of the hyper-converged vendor's effort to...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
reach into larger organizations.
Scale Computing, based in Indianapolis, launched the HC1150DF all-flash and the hybrid HC5150D virtualization arrays, in tandem with an upgraded HyperCore operating system that adds data deduplication, remote cluster monitoring and open APIs.
The Scale Computing HC3 flagship is geared for heavily virtualized environments. The HC1000 family targets small and midsize organizations, while the HC2000 Series is a midrange product line and the HC4000 Series arrays are high-end models.
Scale Computing launched in 2007 by selling hyper-converged storage for video surveillance systems. The primary focus in recent years has been selling its KVM-based HC3 systems to small and midsize businesses, although Scale has been taking steps to address a broader set of storage customers.
The denser flash options continue Scale's trend upmarket, said Terri McClure, a senior analyst at Enterprise Strategy Group.
"Unlike some of the other hyper-converged guys, Scale has been on a slow and steady growth rate. They are not spending four or five times their revenue on marketing and such. The all-flash and denser hybrid flash product gives them the ability to continue that slow, steady growth rate," McClure said.
HC3 flash arrays dynamically distribute shared storage
The all-flash HC1150DF uses the same 1U hardware chassis as the HC1150 hybrid model, but replaces near-line SAS disks with SAS-attached solid-state drives. A single HC1150DF node provides 7.68 TB of raw flash storage, or 3.84 TB of mirrored flash. List price is $18,500 per node.
Jason Collierco-founder, Scale Computing
The HC5150D is a hybrid high-capacity model aimed at data-intensive workloads such as data warehousing, computer-aided design engineering and medical imagery. Storage scales to 77.7 TB of raw capacity per HC5150D node, or 233 TB per three-node cluster. A single HC5150D accommodates nine 8 TB HDDs and three 1.92 TB SSDs in a 2U chassis.
Scale Computing sells a three-node HC5150 base configuration for $94,500, which includes 48 cores and 384 GB of RAM. Users can expand the HC5150 cluster to support up to 25 eight-node subgroups, or 200 nodes in a management group.
As with previous generations of hardware, the new dual-controller flash arrays can be plugged into any existing Scale Computing HC3 cluster. All HC3 arrays are available as a single controller for use cases such as edge computing or to run applications that don't require chassis-level failover.
The hybrid HC5150D storage density is expected to broaden HC3's appeal beyond its historical focus on SMBs. Jason Collier, Scale Computing co-founder, said Scale is being pulled into larger deals by enterprises seeking distributed flash storage.
"We are seeing ourselves being pulled further and further upmarket," Collier said. "There's a ton of customers we had to turn away because our node wasn't sized quite right for them. The all-flash array gives us a product for midsize customers that run a specific application with an all-flash requirement."
Scale's deduplication debuts in HyperCore 7.3
Scale Computing does not support VMware ESXi or Microsoft Hyper-V hypervisors. The HyperCore hypervisor is based on open source KVM and the Scale Computing Reliable Independent Block Engine (Scribe) object store. Scribe runs in the Linux kernel to create an object storage layer that is directly accessible to the HyperCore hypervisor.
Despite disavowing VMware, supporting it could be inevitable for Scale to realize its ambitions. Scale also has to jockey for position against other hyper-converged competitors claiming to offer low-cost alternatives to VMware.
"Part of Scale's attractiveness is that customers are not locked into the VMware licensing fees, but that's also where they'll run into friction. We always hear that people are tired of paying license fees for the VMware hypervisor, but we're not seeing any mass migration" away from VMware, McClure said.
Still, Collier insists VMware support is not in Scale's plans, for both marketing and technical reasons.
"Our mission in life is to displace VMware," Collier said. "Our architecture is different than anybody else's. We couldn't use VMware or Hyper-V with our architecture. We use KVM, KVM doesn't use us. Our entire software stack is bundled into the operating system."
HyperCore 7.3 embeds the first iteration of Scale's deduplication engine, which applies data reduction asynchronously after writes are committed to storage. Inline deduplication per virtual disk is on the HyperCore roadmap.
Scale's deduplication provides efficiency, provisioning and utilization metrics across logical and physical storage. Remote monitoring allows health and status checks on multiple Scale Computing HC3 clusters. HyperCore 7.3 marks the first time that Scale has opened its REST-based APIs to developers to write custom script to automate HC3 deployments.
Most of Scale's 2,500 customers run their entire infrastructure on HC3, Collier said. The vendor is "actively looking" for venture funding as it goes head to head more often with hyper-converged pioneer and market leader Nutanix, which has started to pursue the midmarket customers that are Scale's bread and butter. According to Crunchbase, Scale Computing's last funding in July 2015 was an $18 million Series E round led by ABS Capital Partners. In March Scale disclosed to the Securities and Exchange Commission a $2.51 million private placement from sales of unregistered securities.
"We see Nutanix in a lot of deals now. We used to see them a lot less. We can win deals if we get in the room. We just need to get in the room more," Collier said.
In another move aimed at larger enterprises, Scale is testing a pilot project involving four Scale Computing HC3 nodes built with NVM Express flash, using Remote Direct Memory Access across 100 GbE. Product details are expected later this year.
Performance, vendor lock-in among the hyper-converged hurdles
As networked storage declines, all-flash and hyper-converged systems win favor
VMware touts all-flash storage in VSAN 6.5 licensing tweaks