Red Hat Inc. moved into the crowded and hotly contested hyper-converged race today with a new open source software-based...
product that targets remote and branch offices.
Red Hat Hyperconverged Infrastructure integrates Red Hat's supported distributions of the open source Gluster distributed file system, enterprise Linux operating system, Kernel-based Virtual Machine (KVM) hypervisor and its Ansible deployment tool. The product offers management and orchestration capabilities through a RESTful API to the Red Hat CloudForms tool.
Ross Turk, Red Hat director of portfolio evangelism for storage, said some customers who deployed Red Hat Gluster Storage and Red Hat Virtualization together on their own asked for "a more productized, supported version."
He said Red Hat tested and validated Gluster and Red Hat Virtualization on the same servers across various deployment scenarios under a common single pane of management for the entire infrastructure.
Hyper-converged hardware options
The Red Hat Hyperconverged Infrastructure software is designed to run on commodity hardware. Red Hat has no hardware partners for its hyper-converged product, although Turk said there's "tons of potential for it." The company recommends that customers work with a Red Hat solutions architect to determine the proper sizing requirements.
Terri McCluresenior analyst, Enterprise Strategy Group
Turk said, in stand-alone Gluster deals, some Red Hat storage buyers prefer appliances, whereas others would rather select their own hardware. He said large organizations that operate at petabyte scale often prefer to choose their own hardware.
But David Floyer, CTO and co-founder of Wikibon, said Red Hat would need to partner with hardware suppliers or provide a short list of supported options to succeed in hyper-convergence. He said organizations do not want to put together the hardware and software themselves, and they do want one vendor to handle support.
"Every CIO or CTO I talk to is saying, 'I don't want to do anything which is undifferentiated,' changing the microcode or anything like that," Floyer said.
Stiff hyper-converged competition
Terri McClure, a senior analyst at Enterprise Strategy Group, said Red Hat's competition would include major server vendors Dell EMC, Cisco and Hewlett Packard Enterprise, as well as pure-play hyper-converged vendors Nutanix and Pivot3.
"There are other people that have a huge head start on them, so I don't see them toppling Dell EMC or any of the other vendors in the near future," McClure said of Red Hat. "But I do see the Red Hat loyalists out there -- and there are plenty of them -- welcoming this with open arms because they've been assembling these hyper-converged solutions on their own at the edge."
Turk said most hyper-converged vendors focus more on data center deployments than Red Hat will. He said many Red Hat data center customers choose to scale compute and storage infrastructure separately, and a hyper-converged product that would force them to scale compute and storage at the same time would not suit their needs. Conversely, remote or branch offices can benefit from having compute and storage on a single server to address space, power or staffing constraints.
"We have customers today who have computing needs beyond their data centers. They want to bring computing a little bit further out into their field offices," Turk said. "They need to be able to have their full software stack in a super minimal footprint."
Potential to boost Red Hat storage
Floyer predicted that hyper-converged infrastructure would give Red Hat its best chance to compete in the storage market. He said organizations would increasingly need edge-based hyper-converged systems -- with built in management, orchestration and automation -- for warehouses, electrical substations and other places to reduce application latency.
McClure said the Red Hat Hyperconverged Infrastructure product could also appeal to organizations that fear vendor lock-in. She predicted greater openness toward open source software as infrastructure becomes more software-defined.
"But they may face some headwinds relying on Gluster, because Gluster just is not as known a commodity as some of the things that they're competing against," McClure said. "They're going to have to prove themselves in the field."
Red Hat Hyperconverged Infrastructure pricing
Red Hat Hyperconverged Infrastructure has been in limited availability before this week's official product launch. The initial product combined Red Hat Gluster Storage version 3.2 and Red Hat Virtualization version 4.1. The starting list price is $12,000 for a three-node cluster with standard support.
Red Hat Hyperconverged Infrastructure customers can deploy three, six or nine physical servers. For small deployments, the servers must have at least two six-CPU cores, 64 GB of RAM and up to 48 TB of storage. Hosts for medium-sized deployments need at least two six-CPU cores, 128 GB of RAM and up to 64 TB of storage. And the nodes for large deployments require at least two eight-CPU cores, 256 GB of RAM and up to 80 TB of storage.
"They're being very careful with their rollout. They're making sure they can support the use cases, so they're targeting remote and branch office deployments where hyper-converged really makes sense. You don't have to maintain multiple layers of architecture," McClure said. "As they get more experience with it and get it out in the field and continue development on it, I expect them to bring it into the enterprise and lift some of the limits on cluster size."
Red Hat Hyperconverged Infrastructure deployments must have at least two network interface controllers, one for the back end to probe Gluster peers and set up the volume, and the other to create the virtual network bridge. For each network controller, Red Hat recommends dual interfaces in a redundant configuration.
Turk said the vendor is considering a hyper-converged option for Red Hat Ceph Storage. He said some customers already deploy Ceph, which supports block, file and object storage, and OpenStack cloud software in a hyper-converged architectural scenario, but Red Hat is not willing to discuss any future potential products.
Which workloads are suited to hyper-converged infrastructure?
Evaluating the benefits of hyper-converged infrastructure