Hyper-converged infrastructure has become a hot topic in IT.
All the major suppliers now support hyper-converged systems, with many recent software-defined storage startups re-positioning themselves as a hyper-converged play.
For small to medium-sized enterprises (SMEs), hyper convergence represents a decision point: whether to follow the traditional server, networking and storage architecture or to deploy a single hyper-converged system. Which, if either, is best?
Hyper-converged appliances combine server (compute), networking and storage resources with server virtualisation to create a unified hardware platform that incorporates all the above traditional disciplines.
Where previously, individual hardware products would have been sourced, deployed and configured to work together, hyper-converged offerings are delivered as a cluster of servers. Features of networking and storage are moved into software, using commodity hard drives and SSDs.
In hyper-converged systems, scalability and resilience is delivered through the deployment of multiple servers/nodes (typically a minimum of three or four) that distribute storage and virtual servers across the hardware.
A single node can fail (or be taken out for maintenance, upgrade or replacement) while the remaining nodes continue to run the infrastructure. Users can scale by simply adding a new node and making use of the additional capacity added to the cluster.
Hyper-converged systems represent an expanding market, as more suppliers and startups deliver products into this space, and some of the more prominent suppliers are listed below.
Dell EMC has a range of products under the VCE VxRail brand. This incorporates VMware’s Virtual SAN platform with additional software from VCE.
An entry-level system requires a minimum of four nodes, with the VxRail Appliance 60 providing a single Intel Xeon 6-core E5-2603 1.6Ghz processor, 64GB DRAM, up to 10TB of storage capacity (with 200GB SSD cache) and 1GbE networking. Dell VxRail appliances scale to 20-core processor and all-flash models with 10Gbps Ethernet.
HPE offers a hyper-converged system for SMEs based on the Proliant DL380 Gen 9 server and storage using the StorVirtual VSA.
The Hyper-Converged 380 system starts with two nodes, with dual Intel Xeon E5 processors (6-18 cores), from 128GB to 1,536GB DRAM per node, 1/10GbE networking and up to 40TB of storage capacity. Virtualisation is provided by VMware vSphere, integrated with HPE’s CloudSystem 9 software.
Nutanix is probably the most well-known hyper-convergence company, and recently debuted on the NASDAQ through an initial public offering (IPO).
Although initially very much enterprise focused, earlier in 2016 Nutanix announced Xpress, a hardware platform aimed specifically at SMEs.
The SX-1065 node ships with dual Intel Broadwell processors (up to 20 2.4Ghz cores), up to 512MB of DRAM and a choice of SSD and HDD storage. Network connectivity is provided through either 1GbE or 10GbE networking.
Xpress comes with the Nutanix AHV (Acropolis Hypervisor) and can support VMware and Hyper-V virtualisation.
Pivot3 is a relative newcomer to the hyper-converged market and recently acquired NexGen storage. Earlier in 2016 the company announced Edge Project, a hyper-converged platform aimed at SMEs and remote office requirements.
The platform is based on up to six nodes, using Intel Xeon E3-1270 processors, up to 64TB of DRAM per node, 32TB of SATA HDD storage and 1GbE networking.
Scale Computing offers a range of hyper-converged products that are specifically targeted at SMEs.
The HC3 platform comes in three hardware product ranges, from the entry level HC1000, through the HC3000 to the high-end HC4000. Each model range offers a hybrid and HDD-only system, scaling from a single 6-core processor, 64GB of DRAM and four 1TB HDDs (HC1000), to 16-core 3.4Ghz processors, 384GB of DRAM, 800GB SSD and six 2TB HDD (HC4150-SSD).
HC3 uses the open source KVM hypervisor and an internally developed distributed storage layer that removes the dependency on traditional commercial hypervisor software.
Simplivity offers a range of systems under the OmniCube brand. Hardware platforms are available from Dell, Lenovo, Cisco or white-box servers.
The CN1400 range, targeted at SMEs, comes in single Intel Haswell or Broadwell offerings, with up to 711GB of usable DRAM, a hybrid storage configuration of 400GB SSDs and 1TB HDDs and 10GbE and 1GbE networking. OmniCube integrates tightly with VMware’s vSphere hypervisor and vRealize Automation suite.
Storage for SMEs is offered by all of the major suppliers.
Dell EMC has a range of SME storage offerings, including Unity, VNX and VNXe, as well as the Dell SC series products. NetApp has entry-level FAS 2000 (2500 and 2600 series) that deliver block and file access. HPE has SME-friendly MSA 2000 and StoreVirtual systems.
There are also good entry-level systems available from QSAN, DotHill, Tegile, Tintri and others.
Traditional storage vs hyper-converged
With all the above in mind, how should IT organisations go about choosing whether to stay with dedicated hardware or make a move to hyper-converged infrastructure?
Making this decision is a question of looking at the benefits of hyper-converged or dedicated SAN storage and applying that to an IT organisation’s requirements. Hyper-convergence is sold on the following benefits:
- Ease of use – Certainly from a storage perspective, there is very little to do with hyper-converged systems. Storage is exposed as a single unified layer of capacity based on the resources available on each node, usually as a single data store in the hypervisor layer. Administrators don’t need to have in-depth knowledge of Fibre Channel, iSCSI or other specialist storage protocols.
- Reduced TCO – With no dedicated storage hardware to manage, less time is spent on storage management in hyper-converged systems. There’s no firmware to maintain and upgrade, no mapping of LUNs and volumes and no migration work to perform when new hardware is deployed.
- Reduced capital cost – Hyper-convergence removes the need to run a dedicated storage network. Nodes in a hyper-converged cluster communicate with clients and with each other over Ethernet (typically 10GbE). This means existing networks can be reused, with no need for Fibre Channel switches.
- Additional features – Many HCI systems also incorporate other features, such as integrated backup and WAN acceleration.
- Service-based delivery – Hyper-converged hardware provides SMEs the ability to start moving to a more service-based IT model and represents an opportunity to position for a move to public cloud.
Hyper-converged systems also offer a range of operational benefits. The integrated, pre-tested/validated nature of the configurations means SMEs only have to think about which supplier to choose; there is little or no certification and testing to be performed.
Hardware refresh, in most cases, is simply a case of deploying new hardware and migrating workloads from the nodes to be decommissioned.
By comparison, dedicated SAN storage has a number of benefits:
- Workload isolation – Having a dedicated storage appliance means all the behind-the-scenes storage tasks are managed by a dedicated processor, memory and backplane resources. This includes the overhead of data optimisation (dedupe and compression), as well as hypervisor extensions such as VMware vStorage application programming interfaces (APIs) for Array Integration (VAAI), which were initially intended to take some pressure off the hypervisor for repetitive tasks. There’s no competition for resources between storage and virtual machines.
- Sharing with non-virtualised systems – SMEs may have platforms that aren’t virtualised and so may need some shared storage. Many of the gains of hyper-convergence for SMEs come from moving all compute resources into hyper-converged hardware, so maintaining two systems may not be cost effective.
- Multi-protocol support – Many SMEs buy storage for block and file requirements. Shared storage can provide support for both block and file requirements. Unfortunately, hyper-converged systems generally only deliver storage for internally hosted VMs. Providing file-share space, would for example require purchasing additional hardware.
- No lock-in – One obvious benefit of choosing separate components is the lack of lock-in. While hyper-converged infrastructure has some great benefits, some features, such as integrated backup, can tie the customer into their hyper-converged platform.
There are clear benefits to choosing hyper converged over traditional storage and servers. Making the choice to move to hyper converged provides significant operational and simplification benefits. There’s still a place for shared storage but, for SMEs, those cases are becoming less and less.
Read more about hyper-converged storage
- In the first of a two-part survey, we look at the hyper-converged infrastructure market and the startups providing VM-native servers and storage, and datacentre-in-a-box products.
- In the concluding part of a two-part survey, Computer Weekly looks at the offerings of the big players – Cisco, Dell, Fujitsu, HPE, VMware – in the hyper-converged infrastructure market.