The market for converged infrastructure is growing rapidly as IT departments small and large decide that IT complexity is bad. Options from Hewlett-Packard, Dell, IBM, Cisco Systems, VCE and NetApp attempt to reduce acquisition and operational costs, consolidate networking, increase manageability, decrease time to deployment, cut storage costs and reduce power consumption.
Converged platforms get their name from combining storage, networking and compute resources into one easy-to-manage, easy-to-acquire unit. They all traditionally offer a single pane of glass for management, or at least a single management model, and support for converged solutions usually comes through a single vendor.
Beyond that, though, the similarities end. Some options may use converged networking techniques such as Fibre Channel over Ethernet or iSCSI to access storage, while others use traditional, separate Fibre Channel switches. Some may have preset quantities of storage and compute while others let you customize heavily, designing a system that meets your needs exactly. What follows is a high-level overview of the major converged infrastructure offerings from Cisco, VCE and NetApp . We look at the architecture, packaging, support and other details you may want to know before deciding what to put on your data center floor.
Cisco UCS ages gracefully
Introduced in 2009, Cisco Unified Computing System (UCS) is the granddad of converged infrastructure solutions. As a newcomer to the server space, Cisco thought long and hard about the problems IT faces. From the beginning, that meant putting Fibre Channel and IP networking on the same back-end infrastructure, and rethinking how to better manage the hardware itself.
UCS servers include B-Series blades and C-Series rackmount systems, both of which are managed through the built-in Cisco UCS Manager software. UCS Manager abstracts all the vital data from the hardware and stores it as a profile. This makes it easy to deploy or replace server hardware, since UCS Manager can be used to copy or automatically reassign addresses and configuration data to new or replacement blades.
In addition, UCS is quite scalable, and the blades come in the same sizes as physical servers. For example one-, two- and four-CPU models have increasing numbers of memory sockets as the CPU count rises. There’s also a low-end UCE E-Series that fits into Cisco’s Integrated Services Router chassis, for remote and branch office setups. The UCS products are fairly operating system-agnostic, with a traditional hardware compatibility list, so it is likely they’d fit into your data center setup with minimal IT staff training.
Despite all the benefits, there are a few drawbacks to UCS. Unlike some competitors that deliver what’s essentially a private cloud in a box, UCS is a server infrastructure solution, though it often serves as a basis for cloud offerings from other vendors. The initial purchase price may be a bit steep since you’ll need to invest in Cisco Nexus switch infrastructure for converged SAN and IP networking. As you expand the environment, those costs will be amortized across all the hardware and be more in line with other vendors’ solutions. Finally, though the support tends to be great, Cisco’s hardware warranty lasts just three years, compared to five years from most other vendors.
VCE Vblock reference architecture
VCE is a company formed of VMware (V), Cisco (C) and EMC (E) technology, with each member contributing technology to the Vblock platform. VMware provides its vSphere and vCloud software suites for virtualization and management. Cisco contributes the Unified Computing System for compute capacity. EMC includes its NX and VMAX storage, as well as the Ionix Unified Infrastructure Manager software that is the single pane of glass for management.
Interestingly, VCE doesn’t actually sell Vblocks; they are a reference architecture that value-added resellers use to build actual products. These reference architectures are heavily validated and tested, and the end products are delivered preconfigured and pretested. As of this writing there are just two Vblock models, the 300 and the 700. That might seem like a drawback, but for IT shops that want converged infrastructure that just works, this is a good thing. A Vblock is turnkey. Its performance is a well-known quantity and you know the configuration will work as advertised.
VCE supports Vblocks with a single group that transcends all three companies, and it maintains a lab for testing and troubleshooting. Because of this model, plus the limited number of configurations, VCE can put out regular releases of firmware and software upgrades, or what some VCE folks call “known good save points.” That saves IT staff a lot of time testing firmware, checking compatibility lists and troubleshooting problems from upgrades.
The lack of a real low end in the Vblock product lineup is a problem for many IT shops, since it is unlikely you’d have a Vblock in a remote office. Others complain about how Vblock can’t be customized and about its inability to connect to existing infrastructure, but the product isn’t geared toward that market. You have to purchase through VCE's partners, which may add cost and complexity to the acquisition process. You’re also limited to virtual servers that use VMware hypervisors, so if your IT dreams include Hyper-V or OpenStack, you are out of luck. Likewise, if you’re hoping for a turnkey private or hybrid cloud, Vblock is not for you. This is a server infrastructure solution.
NetApp FlexPod's single pane of glass
NetApp’s answer to VCE is the FlexPod, which is strikingly similar to the Vblock but less tightly controlled. Like the Vblock, it is based on Cisco’s Unified Computing System B-series blades and VMware software. It is not a product you can buy directly from NetApp, but rather a reference architecture built out by partners.
Unlike Vblock, the FlexPod uses NetApp’s FAS storage arrays. You can choose whichever FAS array you’d like, since they all run NetApp’s Data OnTAP operating system and have a common interface. The FlexPod also supports Hyper-V 2008 (but not yet Hyper-V 2012), as well as bare-metal operating systems. It doesn’t have a unified management interface, so customers use the native management interfaces for the storage, networking and compute components. Because NetApp has chosen components with robust APIs, it’s possible to buy third-party management tools that provide a single pane of glass.
FlexPod can scale to fit your needs, which is nice for shops that would like to run a single converged infrastructure platform in multiple locations. Customers purchase FlexPod through a NetApp partner, which might lead to additional cost and complexity during the acquisition process. Customers also need to maintain separate support contracts with Cisco, VMware and NetApp. NetApp offer a cooperative support model, but the workflow diagrams for that suggest that problems won’t be resolved quickly.
Whether you just want to replace complicated server infrastructure with a simpler setup or host your own cloud, there is likely these are a few of the many converged infrastructure choices that can meet your business needs. In Part 2 of this series, we’ll look at three additional converged infrastructure choices from Dell, IBM and HP.
About the author:
Bob Plankers is a virtualization and cloud architect at a major Midwestern university. He is also the author of The Lone Sysadmin blog. Let us know what you think. Write to us at [email protected]
What converged infrastructure means for IT managers' workflow