Article

Product Review – Hitachi Data Systems’ Unified Compute Platform (UCP)

As a specialist technical architect at AHEAD, my primary responsibility is to be a subject matter expert on converged (CI) and hyperconverged (HCI) infrastructure. As I review and evaluate the different products in the marketplace I will be writing a series of blogs, focusing on a single product per post. To kick off this series I will start with a review of Hitachi Data System’s Unified Compute Platform (UCP).

Company Intro & CI History

Hitachi's Converged Infrastructure Product: UCP

Hitachi Data Systems officially came to being in 1989 as a partnership between Hitachi, Electronic Data Systems (EDS), and National Advanced Systems (NAS). This partnership first started as a mainframe manufacturer, but by 2000 HDS moved away from mainframes and focused instead on storage. HDS had success in the enterprise and mid-market storage space with leading edge features such as the ability to virtualize third party storage arrays.

HDS entered the US blade market in 2007 but has struggled to gain market share. This entry has, however, enabled HDS to combine its own storage and compute platforms into a single converged infrastructure product. HDS launched its truly CI (HDS also has a reference architecture line called UCP Select) UCP line in 2012 called UCP Pro, which has since been rebranded as UCP for their main product, and UPC 4000e for their smaller converged infrastructure solution.

Architecture and Features

  • Product Range

    The UCP 4000e fits the small to mid-size organization (or enterprise satellite offices) and supports between two to 16 blades and 360 drives. Released in December of 2014, this model is new to the market.

    The UCP line covers medium and large organizations and is capable of supporting up to 128 blades and either VSP or HUS VM arrays, capable of scaling to 2000+ drives.

  • Scalability Interdependence

    Current UCP architecture uses fixed-port Brocade and Cisco network and storage fabric. This means that you can end up with a large number of small-ish switches which increases complexity as you scale. Perhaps this is why the solution doesn’t currently scale past 128 blades.

    As you fill cabinets and add new ones, you add more network and storage fabric switches.

  • Flexibility

    UCP can be configured with Cisco and/or Brocade switches, and it is expected to support Cisco blades as early as Q3 of 2015. *Even though HDS currently advertises Cisco UCS as an option for UCP, UCS blades won’t be available until later this year for their CI product.

    UCP’s virtual storage option enables customers that have made recent storage investments to repurpose that storage for their UCP implementation, this is unique in the CI space, where other solutions require a new storage array purchase within their CI product.

Data Protection, Replication, High Availability

  • Data protection

    UCP does not offer any integrated data protection products made or supported by HDS. Instead, UCP customers can purchase Commvault as part of a UCP solution or use another solution altogether.

  • Replication Solutions

    Both local and remote asynchronous replication features are available and powered by the HDS storage arrays and include all common industry features.

  • Highly Available Solutions

    The enterprise-grade VSP G1000 can support true active-active stretch clusters across datacenters for critical applications.

Primary Components

  • Compute

    Compute is provided by the HDS Compute Blade 500 system. This enclosure supports eight half-width blade slots similar to Cisco UCS and has a few very interesting, but not always useful, features.

    LPAR is a physical virtualization technology created by HDS (originally created by IBM on their mainframes), which allows customers to provide physical virtualization, enabling multiple operating system environments (OSE) to run on a single physical blade, without a hypervisor.

    Hitachi Multi-blade allows you to combine up to four full-width blades to create an eight socket and 192 DIMM slot monster which could be a fit for HANA deployments.

    As mentioned earlier, HDS UCP will be offering Cisco UCS blades instead of HDS blades. I think that this is an excellent idea because I suspect that it will be very difficult otherwise for UCP sales to enter the US market which is dominated by Cisco and HP blades.

  • Storage Array

    UCP can be built with HDS VSP or HUS VM arrays. The VSP G1000 is a truly enterprise array that can compete head to head with EMC’s VMAX or IBM’s DS8870.

    The HUS VM is a mid-market solution that runs the same code as the VSP and that, over time, is receiving many of the features built into the VSP line.

    Both arrays are unique in the marketplace in that third party arrays can be connected to them. Even more unique, HDS fully supports this configuration in their UCP solutions, although it does delay the otherwise fast deployment time.

    One could also argue that allowing for the virtualization of third party arrays makes this converged solution a bit less predictable than an all-Hitachi UCP solution or a VCE solution, since HDS cannot test the client’s exact configuration in their labs.

  • Storage Fabric

    UCP uses Brocade 6510 48 port switches to aggregate connections from Brocade 5460 embedded switches (within the HDS blade chassis). These switches are solid and Brocade itself is a market leader in storage fabric. However, this decision makes this platform less efficient at scaling, and is likely one of the reasons why it doesn’t scale past 128 blades.

  • Network

    UCP can be built with a pair of Cisco Nexus 5548 switches at the core and a pair Cisco Nexus 2232 for each rack, along with a pair Cisco Nexus 3048 management switches. Alternatively, Brocade switches can be selected which include a pair of VDX 6720 at the core, a pair of 6746 for each rack, and a pair of FCX 648 switches for management. Note that the Brocade network option limits scalability to 64 blades.

Hypervisor Compatibility

HDS fully supports three hypervisors that include VMware vSphere, KVM, and Microsoft Hyper-V. Although fully supported, Hitachi only provides direct support for vSphere. KVM and Microsoft support is relayed to Microsoft or your linux vendor of choice.

Management Features

  • Centralized Visibility

    UCP is centrally managed through UCP Director, an HDS software solution that enables you to monitor all components of the UCP solution.

  • Centralized Management

    UCP Director is possibly one of the most capable CI automation tools in the market because it not only lets you see components, but it lets you make changes without having to log into individual element managers.

    UCP Director is available as a plugin to VMware vCenter or SCVMM.

    UCP Director allows you to manage both Hyper-V and VMware on the same UCP system, along with bare-metal servers.

  • Automation

    UCP Director provides much of the functionality that you would expect from UCS Director in that you can automate tasks that includes firmware updates, zoning, vSphere cluster creation, and VLAN provisioning on the UCP switches.

  • Monitoring and Alerting

    Although hardware alerts are centralized in UCP Director, SNMP traps are not! If you want to pass alerts upstream to another monitoring tool you have to individually configure each element manager to do so.

  • Integration

    UCP Director truly integrates the subcomponents into a single product, both from the perspective of visibility as well as management. It does this cleanly both for Microsoft Hyper-V and VMware vSphere.

UCP Director: Tightly Integrated into VMware vCenter

Summary

Like many of the non-market leaders in the CI space HDS has chosen to include significant levels of automation within their CI product. This could be an attraction for clients that are looking to automate their operations, although it might also be seen as unnecessary for environments that want to standardize another automation/orchestration tool like vRealize, UCS Director, Puppet, or Chef.  Although I haven’t experienced it myself, I’ve been told that HDS can deliver UCP systems to the customer and have them up and running 40 days after a client submits an order.

The aspects of UCP that concern me are the lack of a full ecosystem of ancillary and approved products (think of all the EMC add-ons that you can add to a Vblock), the lack of scalability for enterprise customers, the lack of a proven track record (why is it that they haven’t penetrated the CI market?), and the fact that it uses Hitachi blades. The toughest part about the blades is that they appear to compete with every other manufacturer from a features and dependability perspective but are still unpopular in the US, which affects the desirability of the entire UCP solution. If HDS is able to penetrate the blade market, either independently, or via UCP, this stigma would quickly go away, making UCP a true CI contender.

At AHEAD, we are constantly testing the latest converged infrastructure technologies and solutions in our Lab and Briefing Center. Learn more by requesting a briefing today.

Schedule an AHEAD Lab Briefing