Scale Computing today upgraded its HC3 hyper-converged platform with a higher-capacity node and software that includes nondisruptive rolling updates, virtual machine-level snapshots and thin cloning of virtual machines.
Hyper-converged systems include storage, servers and hypervisors in one box. Scale Computing HC3 systems began shipping in 2012. Unlike most other hyper-converged vendors such as Nutanix, SimpliVity and VMware, Scale does not support VMware in its appliances. It only supports KVM. Scale claims this brings the price down because it avoids the cost of VMware licensing.
The HC4000 is Scale's new high-end node, with 128 GB to 256 GB of RAM, 12-core dual-socket CPUs and 9.6 TB of storage per node. It is designed for approximately 50 virtual machines (VMs) per node. A three-node minimum cluster includes at least 384 GB of RAM, 36 CPU cores and 28.8 TB of raw storage. A 16-node cluster scales to more than 2 TB of RAM with 192 server cores and 154 TB of raw capacity, and can handle about 800 VMs.
The HC4000 has twice as much RAM, storage and CPU cores as the HC2000, Scale's next-highest capacity node. Scale also has an HC1000 entry-level node. The HC2000 and HC1000 support Gigabit Ethernet (GbE) and 10 GbE networking, while the HC4000 supports 10 GbE only.
"When you're pushing that many VMs, you need a bigger pipe," said Scale CTO Jason Collier.
Collier said customers requested the bigger node for workloads that include higher IOPS or large VMs.
Like the HC2000, the HC4000 uses SAS hard disk drives (the HC1000 uses SATA HDDs). Scale CEO Jeff Ready said the vendor is testing nodes with solid-state drives, and has all-flash and hybrid systems on its roadmap. Nutanix last week launched an all-flash version of its hyper-converged system.
The HC4000 is priced at $22,499, with a starting price of $67,497 for a three-node minimum.
Improved software eases upgrades
UltraEasy HyperCore Software version 5 allows for the upgrade of Scale Computing HC3 clusters with no downtime. The software automatically migrates VMs off a node during an upgrade, then migrates the VMs back to the original node after the upgrade completes. Collier said if a problem arises during the upgrade, the system will create a remote tunnel to the Scale support team, which can diagnose the problem. Scale support can roll back the system to where it was before the problem occurred.
Scale software also takes near-instant VM-level snapshots, which can be cloned to create new VMs. Customers can make thin clones using Scale's allocate-on-write technology to avoid deduplication of data. HyperCore can import and export VMs to a remote file server for backup and archive, and fully create a VM or virtual disk from a remote server.
Ready said Scale has more than 900 HC3 customers and more than 4,000 deployments. Its largest vertical markets are healthcare, education and manufacturing, and its customers are usually small and medium-sized businesses (SMBs) and mid-market companies.
"What defines our customer is what the IT department looks like. We're looking for organizations, departments or branch offices with one to eight IT guys," Ready said.
Its customers include NATO, which runs Scale Computing in its Special Operations headquarters in Belgium. Ready said the NATO office has three IT administrators for about 350 users.
Ready said approximately 80% of Scale customers have some virtual desktop infrastructure (VDI) -- a common use case for hyper-convergence -- but VDI is rarely the only workload on the HC3 systems. Ready said most customers use HC3 to replace VMware.
"We consider VMware our primary competition," he said. "About 75% of our customers are either replacing VMware with our box, or deploying us alongside VMware with the intent of moving over as their VMware contracts expire. We don't generally compete with Nutanix. If we do, it's usually us competing with VMware and if the customer goes with VMware, they might use Nutanix as the hardware stack behind it. We also compete with [Microsoft] Hyper-V, but usually VMware."
Gridstore recently added a hyper-converged system that runs Hyper-V and Nutanix supports Hyper-V, KVM and VMware, but hyper-converged systems still tend to be mostly VMware-centric.
Ben Woo, managing director of analyst firm Neuralytix, said the choice of hypervisor inside a hyper-converged system should have no effect on performance or ease of use.
"A hypervisor is a hypervisor," he said. "Windows and Linux run well on KVM, so that neither hurts nor helps Scale.
"Hyper-convergence is about the simplicity of the implementation and how many moving parts there are -- how can I get my whole infrastructure up and running?" he said. "Hyper-convergence is a data center in a box. It's reduced the moving parts to a minimum."
Woo said Scale Computing's biggest challenge is to get exposure in a market that has been dominated by early players Nutanix and SimpliVity, with VMware recently making a splash. Other large vendors such as EMC, Dell and HP are also planning hyper-converged systems.
"Scale needs to be able to get through to the SMBs, who are their targets," Woo said.
Scale Computing ships next converged storage platform
Hyper-converged storage options increase in the market