Hewlett Packard Enterprise (HPE) announced the newest addition to their StorVirtual line up last month. The new HyperConverged 380 (HC 380) is dubbed the “VM Vending Machine” by HP and scales from 2 nodes to 16 nodes, with 4 nodes per physical chassis. It is based on the HPE ProLiant DL380 Gen 9 server. The architecture utilizes the VSA based StorVirtual software, the lineage of which traces itself to Lefthand, and the key technology being the HPE Data Fabric that enables the clustering.
It leverages the 24 Drive, ProLiant DL380 model. This keeps the drive count per node consistent with the HC 250 models. Hybrid and all flash configurations are possible- with suggestion to future NVMe SSD support in the HC 380. It also appears that any mixing of Microsoft and VMware nodes in the same cluster is not supported at this time (the HC 250 has two unique models, one for VMware and one for Microsoft). The ProLiant 380 also supports/plans to support 40GbE infrastructure, though nothing was suggested in the announcement that 40GbE would be supported on the HC 380. HPE is bundling OneView, StorVirtual VSA, Analytics via Cloud Optimizer into the offering, with option to purchase VMware directly for SPOC support that includes VMware.
HPE One view is now used to manage HPE ConvergedSystem, HPE BladeSystem, HPE ProLiant servers, HPE Synergy Composable Infrastructure and HPE 3PAR StoreServ storage. HPE is placing their bets that common and simplified management will entice customers to additional platforms in the HPE portfolio.
As awesome as the new HC 380 is, it is not much of an update from the older models. When compared to its competitors in this space (VxRail, SimpliVity, HyperFlex, etc.), the biggest thing in its favor is its massive storage capacity. Other than that, this is just a small hardware upgrade from their last model.
- HPE Hyper Converged offerings utilize StorVirtual, a VSA based implementation that requires dedicated resources that reserve CPU and RAM from the host. (2-5 vCPUs and 9-26 GB reserved RAM)
- Contrast this with the kernel based integration of VSAN which does not have VSAs anywhere; therefore there are no CPU or RAM reservations to configure, and VSAN is designed to consume less than 10% of host resources, even under stress.
- HPE’s Hyper Converged line up does not offer any all-flash options
- StorVirtual software lacks newer modern data services, like Erasure Coding , Qos, and dedupe/compression or tiering to the cloud (unlike something like SimpliVity or VxRail)
- StorVirtual requires complex set-up to create redundancy: Physical disks used by each VSA must be setup for RAID before creating storage volumes; Then to further ensure node redundancy across nodes network RAID must also be configured amongst VSAs; Only then do RAID settings apply to the entire storage volume - there is no ability to customize redundancy levels for individual VMs or virtual disks except to move them to another volume.
- Contrast this to VxRail utilizing VSAN redundancy that is tunable via storage policy. “Failures to Tolerate” setting works to detect drive, controller & node failures and is node & rack aware and is applied on a virtual disk basis.
- StorVirtual snapshots are done at the volume level, which can result in reserving capacity for those VMDKs in a volume that may not require being snapped.
- Contrast this to snaps in VxRail that are done on a per VM basis, consuming less space for only the VMs that utilize snapshots
- HPE asynchronous replication is called Remote Copy, and is a snapshot-based, asynchronous replication that can hit 30 minute RPO windows
- Contrast this to SimpliVity or VxRail’s WAN based dedupe and compression for Seconds of asynchronous replication.
- HPE Hyper Converged portfolio does support multiple hypervisors, and HC 380 does not allow mixing of hypervisor environments in the same cluster (unlike EMC’s ScaleIO).
- HC 380 uses a vCenter plugin via HP OneView for VMware vCenter, so while it is VMware capable, it adds another layer of management for environments
As far as the hardware enhancements goes, there has been a good effort from HPE to boost up their storage and compute density. It still seems to fall behind some of its competitors though. Here is a side-by-side comparison to some of its features compared to VCE’s VxRail product.
All in all, the HC 380 has a bit of a letdown for me. When HPE first announced their Hyper-Converged offering I was excited to see how HPE’s innovations would help further the technology in the Hyper-Converged market. But so far, it doesn’t seem like they are doing enough to different in such a new market but they need too. There is some of the stiff competition out there from other Hyper-Converged venders and they all came to play. Hopefully HPE can step up to the plate soon as well.