BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
Backup vendor Idealstor is entering the hyper-converged market with a system aimed at small and medium-sized businesses...
that starts at $15,000 for two nodes.
Hyper-converged systems combine storage, compute, networking and server virtualization in one box. The model is becoming a popular option for departments, remote offices and companies implementing virtual desktops.
The Idealstor Hyper Converged Systems (HCS) platform runs the vendor's software on x86 servers and direct-attached internal hard disk drives (HDDs). Idealstor software runs as a virtual storage appliance on the hardware, which is built on Intel servers and commodity HDDs and solid-state drives (SSDs).
The three HCS configurations include high availability and they support VMware and Microsoft Hyper-V hypervisors, but Idealstor CTO Nandan Arora said the main focus is the price. Typically, hyper-converged systems run into the six-figure price range.
"We worked backwards," Arora said. "We looked at a price and said this is our goal for how much it should cost, and we went back and engineered it based on that price."
Idealstor's HCS models start with two-node clusters that scale to 64 nodes. The HCS1000 is a 1U box with 64 GB of RAM, dual Intel Xeon processors with eight cores, 400 GB of SSD capacity, from 2 TB to 8 TB of hard disk drive (HDD) capacity and eight Gigabyte Ethernet (GbE) ports, starting at $7,498 per node. The HCS2000 is a 2U system with 128 GB of RAM, four Xeon processors with 24 cores, 1.6 TB of SSDs, between 2 TB and 16 TB of HDD capacity and eight GbE ports. The HCS2000e uses the same chassis as the HCS2000 but includes 256 GB of RAM, 3.2 TB of SSDs, and eight 10 GbE ports.
Jeff Katosenior storage analyst, Taneja Group
Arora said the vendor is considering an all-flash version in future releases.
Most hyper-converged systems, such as those sold by Nutanix, SimpliVity and VMware, require a minimum of three nodes for high availability. Idealstor HCS works with two nodes that require no switches. A two-node HCS2000e cluster can run as two separate boxes that don't share a backplane, and can be located in different parts of a building up to approximately 330 feet apart. The HCS2000e uses synchronous mirroring to protect data, making sure data is written to two nodes for high availability. Write operations don't complete until they are acknowledged on both HCS virtual storage appliances, making data copies on both nodes active.
Still, the HCS lacks sophisticated data management services such as deduplication, compression, replication, snapshots and thin provisioning.
"Everyone else is focusing on the enterprise," Arora said. "Our goal is to take on SMBs and remote sites. Hyper-convergence is hot these days, but most SMBs have not heard the term. We wanted to build a system where you know what kind of performance to expect, and there's no complexity in putting it together."
Idealstor's closest hyper-converged competitor is Scale Computing, which also plays in the SMB space and has three models. The low-end HC1000 is priced at $22,499 for a three-node configuration with 96 GB of RAM, three Intel quad CPU cores and a minimum of 6 TB (24 TB maximum). Scale's largest capacity system is the HC4000, which includes 384 GB of RAM, six hex core Intel CPUs, and from 14.4 TB to 28.8 TB of raw storage starting at $67,497 for three nodes. Scale includes features such as virtual machine-level snapshots and virtual machine thin cloning that Idealstor's HCS lacks.
"Idealstor is going after the cost dimension, and not trying so much to be feature-rich," said Jeff Kato, senior storage analyst at the Taneja Group. "It's good enough for the smaller businesses Idealstor has contact with."
Kato said he expects the more-established hyper-converged players such as Nutanix and SimpliVity to scale up farther into the enterprise and down into SMBs. Both of those vendors now support KVM hypervisors, which drives costs down because customers don't have to buy VMware licenses.
"This is another sign that hyper-convergence is going mainstream," Kato said. "This is typically what happens when markets mature. People will nibble at the edges and try to carve out a space. Idealstor comes from a low-cost backup background. Its customers are saying, 'Hey, can we move hyper-convergence to a lower price point?' And just like we're seeing with the all-flash guys, it's a badge of honor to get to a lower price point."
What's the difference between converged and hyper-converged systems?
Guide to the hyper-converged landscape
How hyper-converged storage can relieve server strain