As a technical architect at AHEAD, one of my responsibilities is to be a subject matter expert on converged infrastructure (CI) and hyperconverged infrastructure (HCI). Over the last month, I’ve been evaluating various new and/or popular products in the marketplace and have been relaying my findings in the AHEAD blog. Most recently, I reviewed Cisco HyperFlex and Nutanix Acropolis 4.6 release and for this edition, I’ll be focusing on VCE’s VxBlock and Vblock. If you like what you read, stay tuned for my next post around VxRail–the newest hyperconverged appliance from EMC, VCE, and VMware.
Company Intro and CI History
VCE was founded in 2009 through a partnership between Cisco, EMC, and VMware (owned by EMC). The initial goal of the company was to provide what is now known as reference architectures, by combining Cisco networking and compute hardware in combination with EMC storage and VMware software. This first iteration of VCE’s offering involved shipping the components to the customer’s location and assembling the gear onsite. Although this reference architecture approach provided some benefits, it was nowhere near as impactful to the industry as VCE’s current converged infrastructure offering.
VCE first established their now dominant foothold on the converged infrastructure space with the release of vBlock 300 and 700 series in 2011. By then, Cisco UCS Blade Servers, released in 2009 and integrated into VCE’s products, were gaining significant market share both within and outside of Vblocks. VCE also refined their manufacturing processes and met ISO 9001 requirements, which included assembling, testing, and applying logical configurations to Vblocks at the factory.
VCE is now considered a market leader in the converged infrastructure space with $2 billion in revenue. VCE was originally founded as a joint venture and operated that way until the end of 2014. As of January 2015, VCE is now an EMC business unit. This shift in ownership has enabled VCE and EMC to enter the hyperconverged market, where VCE is bringing products to market in an attempt to compete with companies like Nutanix and SimpliVity by introducing two brand new product lines: VxRail and VxRack (again, I will be reviewing VxRail in my next post). More recently, Dell’s acquisition of EMC will make the CI/HCI market even more exciting.
Architecture and Features
- The Vblock platform fits the needs of medium to large data centers. Although VCE has the VB200 models that target relatively small compute (four to twelve ESX hosts) and small storage (2+ TB), they have not shipped at volume or achieved significant market momentum, primarily due to the cost of the components when compared to reference architecture and hyperconverged solutions
- The VB340 scales up to 128 Blades and a VNX8000 (up to 1500 drives.) The Vblock340 is the workhorse Vblock, supporting medium to large deployments that don’t require AFA (all-flash array) performance. We are expecting a new version of this—the VB350—to be released in the next few months.
- The VB540 scales to 192 blades and six (soon to be eight) XtremIO X-Bricks, each of which is either 10TB, 20TB, or 40TB in raw capacity. This is the optimal platform for workloads that require high performance and/or are good deduplication candidates (VDI and environment clones, such as QA/dev/test/prod).
- The VB740 scales to 512 blades and can be built with a VMAX100, 200, or 400K that can scale to 5000+ drives. This platform is suited for the largest workloads and/or for workloads that require the highest resiliency.
- Note – VCE is becoming less rigid when it comes to their models; you can add several “tech extensions” to most models, such as adding XtremIO to a VB340 or adding C-series servers, or a VNX to any model. This means that customers scale and stretch beyond the base maximums defined by their Vblock model.
- VCE has also introduced the VxBlock product line that includes the VxBlock 340, VxBlock 540,and VxBlock 740. As part of the dissolution of the joint venture, VCE has created the VxBlock product line to enable it to replace some Cisco components with VMware. This line is almost identical to the Vblock, except that it supports NSX out of the factory but does not support Cisco 1000v.
- Base Vblock scalability is calculated based on the combination of network, storage, and compute elements. A VB340 doesn’t scale past 192 blades because a typical workload will max out a VNX8000 before it needs more than 192 blades.
- Overall, the Vblock product family offers a large number of options, which allows customers to scale the solutions when needed. Furthermore, the use of Cisco UCS blade infrastructure enables for high density compute resources while minimizing network and fabric infrastructure.
- The primary limiting factor of Vblocks are that they can only contain a subset of Cisco or EMC hardware.
- Cisco and EMC both have very large portfolios.
- Vblocks are becoming less and less constrained as VCE continues to expand the technology extensions that can be added to Vblocks. Notable additions include storage extensions (additional VNX and XtremIO) and C-Series servers.
- During the first two to three years of VCE’s market presence, many people thought that the product was very inflexible with few configuration options or rigid exception processes. Product, organizational growth, manufacturing, and support maturity, in addition to thousands of real-world deployments, have enabled VCE to be tremendously more flexible in deployment and support of many configurations today.
Data Protection, Replication, and High Availability
- Vblocks can be equipped with EMC Avamar, DataDomain, and EMC data protection products, providing single call support that includes the data center hardware and the backup solution
- Vblock storage can be replicated using RecoverPoint appliances, which is supported fully and directly by VCE.
Highly Available Solutions
- Vblock storage can be stretched with EMC VPLEX, providing true active-active data centers for scenarios where data centers are sufficiently close to each other.
- Compute resources are provided primarily by Cisco UCS blade servers with Cisco Fabric Interconnects. Chassis supports eight half-width blades, and a single Fabric Interconnect pair can support up to 16 chassis or 128 blades.
- Blades are primarily B200s, which support up to 36 cores on two sockets and 768GB of RAM, although other models are supported for special use cases like SAP HANA.
- Vblock 100s, 200s, and C-Series technology extensions use C-Series servers (notably the C240 M4). In the VB200s the servers are not managed by Cisco UCS, so you don’t gain the operational advantages of UCS.
- C-Series extensions are connected to VB340s, 540s, and 740s to fabric interconnects, meaning that server management is centralized along with the blades in the Vblocks. Workloads that demand specific PCI configuration (Heavy GPU VDI, for example) are great use cases for C-Series extensions.
- 200s and 300s use the VNX series array, which ranges from 125 to 1500 drives, depending on the model. This is an “old generation” array that is very established in the enterprise; it can’t provide the performance benefits of AFAs and hybrid arrays, but it’s a solid platform for most workloads certified for enterprise applications.
- VB540s are based on the XtremIO AFA platform—EMC has had terrific success with it. Orders of magnitude improvements in performance over the VNX line, along with terrific dedup ratios for dedup friendly workloads, make this the default array by many customers who are considering purchasing a Vblock.
- VB740s are based on EMC’s newest VMAX; the VMAX3 can range up to 5760 drives and is considerably denser than in prior models. Although EMC has faced criticism for rushing this platform to the market before all of its enterprise features were ready, this is still the primary choice for enterprise applications that require the highest uptime.
- The base Vblock model determines the storage fabric options, which span from unified storage over Cisco Nexus 5Ks to MDS 9148, MDS 9148s, and MDS 9706s. Additional storage switches can be added for special use cases, such as to provide connectivity to iseries servers.
- Network services are powered by Nexus 5548/5596K or 9396 switches, providing enterprise grade capabilities. Note that Vblock does not allow Layer 3 configurations on these, but instead requires that layer 3 services be provided by northbound switches outside of the Vblock.
- Nexus 9K switches have the benefit of supporting 40GB and are capable of running Cisco ACI, although the decision of whether or not to run ACI vs. NX-OS has to be made before workloads are added to the Vblock.
- VCE can implement ACI or VMware NSX in the factory if requested by a customer on any Vblock model. In general, software-defined networking is a very new concept, and with the marketplace dynamics being unsure, we have not seen any SDN factory deployments to date.
- An additional pair of Cisco 3064 Nexus Switches provides out of band management.
- Vblocks officially support VMware vSphere and VCE can provide single call support from the hardware all the way up through this hypervisor.
- Customers can choose to install alternate hypervisors such as KVM or Hyper-V, but this isn’t as efficient from a VCE support perspective.
- Vblocks are managed through VCE’s Vision Intelligent Operations software. The solution enables an organization to review the health of all major Vblock hardware components.
- Vision 3.0 allows for multi-Vblock management from a single window.
- VCE Vision does not directly manage the Vblock components, and is aptly named for its ability to only provide visibility. It does, however, provide links to all element managers, enabling administrators to quickly navigate to the management screen in question.
- Vblock hardware and software do not provide any automation or orchestration tools.
- However, Vblocks are great candidates for automation and orchestration through the addition of products like UCS Director or VMware vRealize Suite.
Monitoring and Alerting
- VCE Vision aggregates alerts from the different element managers and provides a single SNMP manager that can then be piped to any SNMP compatible monitoring tool.
- VCE Vision can be integrated into VMware vCenter, allowing VMware administrators to continue using vCenter as their primary management portal.
Based on all of the capabilities, it’s not surprising that VCE’s Vblock is one of the dominant solutions in the converged infrastructure space. It has combined the most dominant enterprise-grade components that customers have already been purchasing into a single solution, and has done so while reducing complexity and time to deployment, providing true single vendor support, and increasing availability by producing fully pre-tested and preconfigured packages. VCE also leverages the large EMC product portfolio, providing enough solution breadth to cover most data center requirements. In addition, it has also established the release certification matrix, a firmware and software bible that details the steps required to ensure that a Vblock is kept up to date—a critical differentiator when comparing Vblocks to reference architectures. Vblock customers know that they will have a documented, certified path that ensures that their Cisco networking and blades, EMC storage, and VMware software will continue working together throughout the Vblock’s lifecycle.
Vblocks also have several weaknesses. For the most part, they are not a fit for small environments, partly because the smallest model, the VB100, was never very successful until it was discontinued. By definition, Vblocks can only contain specific hardware components, so customers that want to leverage hardware from market disruptors like Tintri or Rubrik will not be able to do so on a Vblock.
My recommendation to organizations is to always consider converged infrastructure when developing infrastructure strategy. It’s not a fit for all scenarios, but it will most likely be cheaper, faster, and easier to manage. And if you have a higher risk tolerance, tighter initial budgets and/or web-scale needs, definitely consider products in the HCI infrastructure space.
In the past year, HCI has definitely become an accepted category for enterprise data centers to the point where products like Vblock are competing head-on with HCI vendors like Nutanix. It typically isn’t clear whether an organization should go for a CI or an HCI solution, or rather, its very easy to build arguments for either option, but difficult to argue against them. I’ve been lucky enough to have guided dozens of customers through the decision matrix and can say that there are financial, technical, and organizational sentiment criteria that have to be studied prior to making this type of decision If you’re interested in finding out more about the decision matrix, feel free to reach out if you find yourself unable to make these decisions for your organization alone.
In this post, I’ve mentioned both converged infrastructure and hyperconverged infrastructure. These terms are not fully defined or understood in the marketplace, so I’ve provided definitions that we use at AHEAD when working with clients. For additional context, I’ve also provided a definition of reference architecture.
- Converged Infrastructure – A manufactured solution that combines enterprise grade compute, storage, and network products. These solutions combine equipment from one or more vendors and provide ongoing single-call support and lifecycle management. These solutions tend to provide maximum flexibility and support for enterprise applications. They are typically sold by established infrastructure vendors such as VCE, HP, IBM, and Hitachi, and perform significant intra-component testing and validation along with certification with major software applications (SAP, EPIC, etc.).
- Hyperconverged Infrastructure – A manufactured solution that combines non-enterprise compute, storage, and network resources. Critical services such as data integrity and high availability capabilities are managed through software features. These solutions tend to be node-based following a web-scale approach and typically do not have unique components (all nodes are equal). These solutions scale computing and are typically sold by startups and market disruptors such as SimpliVity and Nutanix. Established firms like VCE are entering this market with new product lines.
- Reference Architecture – A reference guide that describes a pre-designed and pre-tested combination of compute, storage, and network products made by different manufacturers. This arguably provides the least value to customers, because reference architectures do not provide single call support, do not pre-assemble the solution, and do not provide ongoing lifecycle management in the form of guidance on how to ensure that future updates and expansion do not invalidate the original reference architecture. Reference architectures are commonly used by established manufacturers such as Cisco, EMC, NetApp and IBM in order to provide a vision of a complete data center solution.
AHEAD specializes in both CI and HCI solutions and we’d be happy to meet with you if you’re interested in learning more.