Hyper-Converged Infrastructure (HCI) is a tightly-coupled compute, network and storage solution that dispenses with the need for a traditional storage solution. In most cases, traditional storage environment can complicate management, maintenance, expansion and problem solving, as it requires communications involving multiple vendors. While reducing the complexity and increasing the operational efficiency, HCI bridges the gap by addressing the need for automation and simplification, and can also provide advanced storage management functions like backup, recovery, replication, deduplication and compression.
HCI can offer agility, simplicity, scalability, automation, high availability, self-healing, cloud readiness and security. HCI supports any workload at any scale, allowing you to manage your global cluster from a single pane of glass. It reduces not only the footprint in Data Center, but also the storage space required for data. By simplifying the infrastructure, upgrade and expansion can often be accomplished with zero downtime. While HCI solutions often require either three or four nodes to establish data redundancy and begin a cluster, almost all solutions will allow you to grow in single server increments. This enables a “pay-as- you-grow” solution that eases the financial burden for an organization with growth patterns that fluctuate or can be unpredictable.
Presidio's vendor certified expertise can recommend the right hyper-converged product for your business to support critical virtual workload and production applications after the assessment, and then design and integrate to your infrastructure. Below are some high-level overviews of current HCI market leaders: Nutanix, Cisco HyperFlex, SimpliVity and EMC VxRail.
The Nutanix HCI solution is hypervisor agnostic, supporting VMware vSphere, Microsoft Hyper-V and KVM hypervisors. The Nutanix management layer is also independent of the underlying hypervisor. Nutanix solutions support multiple hardware platforms, using server hardware from Supermicro, Dell and Lenovo. The Nutanix solution can start with as few as three nodes and expand up to 64 nodes with all flash or hybrid storage. While Nutanix offers inline deduplication and compression, these features can be disabled for better storage performance. Nutanix uses intelligent tiering within the storage system, using a RAIN disk configuration to distribute data across the cluster and avoid a single point of failure. A Nutanix cluster can also integrate with existing VMware infrastructure. Nutanix clusters can sustain a maximum 2 concurrent disks failures on different nodes and/or failures two concurrent nodes. Nutanix also supports drive encryption and works well with VMware VAAI, VM-aware storage and SRM. Nutanix also offers built-in replication, allowing both VM and File level restores. Backups can also be replicated to the cloud, using Amazon and Azure.
The HyperFlex Data Platform is a fully integrated HCI platform from Cisco, integrating Cisco UCS servers with the VMware Hypervisor and Cisco Nexus networking into a single stack. The HyperFlex environment currently supports the VMware vSphere hypervisor and is managed with a vCenter plug-in. HyperFlex can be deployed on either a dedicated or shared vCenter environment. To facilitate server and network management, each HyperFlex environment will also include a pair of Cisco Fabric Interconnects. This solution can begin with three nodes, and allows for growth with a mix of HyperFlex nodes for storage and compute or Cisco UCS B200 blades for adding just compute resources. This mixing of nodes within a cluster will allow for a high degree of flexibility in the effort to balance resources. The HX platform will provide always-on inline deduplication and compression, while also leveraging VMware’s VAAI support to provide improved native snapshots and cloning. The Cisco HyperFlex platform will offer cloud based monitoring and reporting with support for all components provided through Cisco’s Technical Assistance Center (TAC). More information about the Cisco HyperFlex Data Platform can be found here.
The SimpliVity HCI products are dependent on VMware vSphere and use a vCenter plug-in as their management interface. SimpliVity can be purchased as OmniCube (using Dell whitebox hardware) or OmniStack (using either Cisco or Lenovo server hardware). SimpliVity solutions are powered by an accelerator card that offloads the workload required to support all time inline deduplication and compression. This solution can start with two nodes and grow up to 32 nodes with hybrid storage. Data Encryption will be available with the release of 3.0.10 on Omnicube CN-3400 models, and it will be accomplished using SEDs for SSD Read Cache and the Data drives. Simplivity uses a mixture of RAID levels 1, 5 and 6 (RAID 1 & 5 on Flash + RAID 6 on Spindle disks). Drive parity and data replication are accomplished using RAID on local drives and RAIN to distribute the data across the nodes. Data protection is leveraged to accommodate up to 24 drive failures across as many as 4 nodes in CN-3400 and CN-5400 models. SimpliVity uses flash drives for reading, though it does not have a storage-tiering feature. A SimpliVity environment can be part of existing VMware infrastructure, and it supports VAAI. Although Simplivity does not work with VMware SRM, it works extremely well with automated Disaster Recovery orchestrator - EZDR. Offering built-in full backup policies, and a scheduler for both local and remote backups, SimpliVity can support restores at both the VM and file level. SimpliVity can also replicate to cloud, using either Amazon or Azure.
EMC’s VxRail HCI solution is fully integrated and validated with VMware, and works well with VCE and EMC products, while delivering improved scalability, greater performance and more features than VSPEX BLUE. VxRail is based 100% on VMware and supports all VMware components. Leveraging existing VMware licensing, appliance pricing can be lower for vSphere customers. This solution offers adaptive scalability with both hybrid and all flash storage. VxRail requires a minimum of four nodes, and can grow to 64 nodes, while using VxRail Manager and all tools in vSphere to manage the VxRail Appliance and vSphere environment. However, VxRail cannot integrate with existing vSphere infrastructure. It will need to be its own vCenter environment. VxRail also leverages VMware’s VSAN product for storage, which uses fewer CPU resources because VSAN is part of the hypervisor kernel. Deduplication and compression is only available on All-Flash storage configurations of VxRail, and erasure coding is enabled. A VxRail cluster can be built in as little as 15 minutes. VxRail appliance includes VMware hyper-converged software, Virtual RecoverPoint (15 VMs per appliance) and EMC Cloud Array (10 TB cloud storage license per appliance) which enables the integration of cloud tiering with encryption.
Follow us on Twitter @Presidio
Ramesh Rao, Data Center Solutions Architect, @NJRammy