Not all storage vendors see hyper-convergence as the cure for all storage ills.
Sanbolic CEO Momchil Michailov considers hyper-convergence a cure for some ills, but says it falls short for many use cases. Michailov says convergence is good, but hyper-convergence not so good for enterprise storage. That’s because the hyper-converged approach is tied to one hypervisor and a totally virtualized infrastructure.
Michailov claims hyper-convergence is fine for VDI and remote offices – two popular early use cases – but will never be able to scale into an enterprise storage system.
“There’s only so much you can stuff in one server, and only so many servers you can manage before it becomes a ludicrous proposition,” he said. “Hyper-convergence is 100 percent dependent on virtual workloads and requires that customers run 100 percent virtual shops. I don’t know anybody who’s 100 percent virtual. If you’re a homogenous hypervisor shop, providing customers with a locked down single hypervisor workload isn’t going to fly.”
Sanbolic sells software that can aggregate and manage storage and data services on a SAN, solid-state drives (SSDs), hard drives or server-side flash. That’s different than hyper-converged systems, which combine storage, networking and hypervisors in one box. Most hyper-converged systems are bought with the software and hardware in one package.
Michailov said Sanbolic has customers running multiple storage hypervisors, and he expects a lot more to go in that direction. “The customers we go after are going to have multiple types of hypervisors, and they are not 100 percent virtualized. They have physical infrastructure as well,” he said. “How do you create orchestration across that? We work with OpenStack and CloudStack. We use a share-all architecture, and that means we can have Hyper-V, Xen, KVM and VMware accessing the exact same data and exact same storage at the same time.”
Sanbolic in May revamped and renamed its host-based storage platform to support Linux, XEN, KVM and OpenStack along with its prior support for Microsoft Windows and VMware hypervisors. It changed the product name from Melio to Sanbolic Scale-Out platform while making it a better fit for a wider array of enterprises.
Like Melio, Sanbolic Scale-Out Platform runs on physical, virtual or cloud server instances to turn heterogeneous hardware into shared storage. The software provides storage services such as dynamic provisioning, quality of service across RAID levels, snapshots, and cloning. It supports flash and spinning disk storage.
Sanbolic automatically detects storage and servers and builds clusters that can grow to 144 CPU cores, 2.3 TB of RAM and 2,048 nodes.
“Instead of buying an EMC or NetApp array, we give you that capability on internal hard drives,” Michailov said.
Sanbolic is priced per core, beginning at $1,200 and decreasing as customers scale cores.
David Floyer, chief technology officer at Marlborough, Mass.-based research and analysis firm Wikibon, said the additional platform support is critical for Sanbolic. “There was a very big hole in their ability to go to market anywhere other than the Microsoft ecosystem,” he said. “That was very limiting. If they want to compete in this market, it is essential that they expand the platform.”
Wikibon places Sanbolic in the Server SAN category, which it defines as a combination of “compute and pooled storage resources comprising more than one storage device directly attached to multiple servers.”
Floyer said Sanbolic has a mature product and more flexibility than VMware’s Virtual SAN (VSAN) hyper-converged software. “Some might want a broader range of physical server and hyper-visor SAN support [than VSAN delivers,” he said.