Part of Dell's server business is building custom machines for very large customers who want fewer bells and whistles but higher computing density and lower power consumption. The Dell PowerEdge C6220 is a server the company has brought out of this custom arena and into its regular sales channels. It is the second in a new series of servers that combines internal storage and two or four two-socket "nodes" (aka sleds or blades) wrapped up in a highly efficient 2RU package.
The PowerEdge C6220 is being marketed as a high-density server solution primarily targeting HPC (high-performance computing) and virtual server clusters. In a major change from the previous model (C6100), the C6220 lets you mix and match node configurations. With support for heterogeneous nodes, this platform becomes a great fit for branch offices, allowing you to configure the nodes at the right size for your branch office database, VDI (virtual desktop infrastructure), Active Directory, and more -- as opposed to racking up a bunch of monolithic servers or partially filling an expensive blade server chassis.
The Dell C6220 is targeted at folks who need to pay close attention to power consumption, heat, rack space, and even the number of Ethernet ports they consume in the data center. It's lighter than four equivalent machines, and it consumes significantly less power and pumps out much less heat. It is, simply put, a competent four-headed server that's light on your data center consumables. It offers the density of blades, but without the frills.
Sysgen was identical to every other Dell server I've prepped and configured in the past -- except I had to find a USB hub so that I could boot from the flash drive and still have both USB mouse and keyboard. (There are only two USB ports on the back of each node.) In another slight difference from the rest of the PowerEdge line of servers, there isn't a separate startup hotkey to configure iDRAC, Dell's Integrated Remote Access Controller; instead, IPMI configuration is done from the regular system BIOS menus.
Dell is careful to call these server boards "nodes" and not "blades," for a very good reason. Unlike server blades in a blade chassis, each server in the C6220 is completely separate from the other nodes. The nodes share a common distribution infrastructure that divvies up the DAS (direct attached storage), but they don't share network switches, a common management infrastructure, or any such frills that one might find on Dell's M1000e blade chassis. The only chassis-based information I found while digging around the IPMI interface is basic presence or not for the power supplies. While the nodes share a chassis, you manage them as if they were completely separate servers.
Notice that I said DAS, not NAS. In the first available chassis options, which have a passive backplane for either 3.5- or 2.5-inch drives, the drive slots connect directly to the SAS/SATA lanes from each motherboard. This limits you to a maximum of three 3.5-inch drives or six 2.5-inch drives per node in the four-node C6220 -- and six 3.5-inch drives or eight 2.5-inch drives in the two-node C6220. Note that the eight-drive limitation for 2.5-inch drives in the two-node chassis is a hard drive controller issue. Although 12 drive slots are available per node, Dell doesn't sell any controllers with more than eight ports.
Come the end of May, however, Dell will offer a third chassis option -- an active backplane with a SAS expander -- that supports much more flexible drive allocations. The expander not only allows the C6220 to talk to more drives than SAS/SATA ports (thereby supporting 12 drives with an eight-port controller), but can be configured to allow any server node to talk to any connected drive. In short, with the new active drive backplane option, you should even be able to assign all the drives to a single node and make the rest PXE boot from the primary node. Unfortunately, the active backplane will work only for the 2.5-inch drives.
HPC, private cloud ... and VDI?
These days I see the world through VDI-colored glasses, and the C6220 with the dual wide nodes, along with their full-length, full-height PCI-e v3 slot, could physically accommodate a GPU and thereby support Microsoft's RemoteFX; think full-motion video, 3D graphics, and Windows Aero over RDP. The only hitch is the lack of a GPU power cable. Interestingly, it seemed I could fit an ATI FirePro card onto the node, and the PCI-e riser card seemed to have a 12-volt feed. Perhaps if Microsoft would certify a half-height GPU that fully utilizes the new power capabilities of PCI-e v3, then the C6220 could be a mean VDI machine. But without official support, I'd be afraid of potentially starving the system if I put a power-hungry GPU into the node.
What is a bit puzzling are the two SATA connectors and the Mini-SAS connectors on each node right next to the mezzanine card connectors. I'm thinking an as-yet-unannounced mezzanine option supporting diskless boot for VMware ESXi. I've also been hearing rumblings about a 10GbE NIC option. The current 256GB RAM limit is also due to shoot to a new maximum of 512GB once the new 32GB DIMMs are certified. Think Beowulf compute cluster in one Dell 2RU.
Dell's IP KVM comes by way of Avocent, and the IP KVM functions are in Java. So why only Internet Explorer? Considering that Avocent bought Cyclades, which had a perfectly good Java KVM client that was nearly platform agnostic, I find it a little annoying that I can't KVM from Chrome, Firefox, or a Mac. I'm also wistful about the onscreen virtual keyboard I have on my Supermicro IP KVM implementation, but considering Dell's macro set -- the drop-down list of frequently used keystroke combos such as Ctrl-Alt-Del, F1, Del, and Printscreen -- is pretty comprehensive, I'll call this a wash.
You could spin the Dell PowerEdge C6220 as Dell playing catch-up to the wildly successful Supermicro Twin family, but Dell has done a better job with the same class of machine. Plus, the C6220 is backed by Dell's formidable engineering and support services. (I just want to say that the idea behind the Dell service tag is brilliant, especially for those of us with lots of machines of varying ages. The ability to look up the shipped config even years after purchase is a godsend.) With support for heterogenous nodes and (soon) asymmetrical drive allocations sweetening the deal, the C6220 is a great compact server solution for the branch office, co-location presence, VDI, or workgroup number crunching.
Dell PowerEdge C6220 at a glance
|$36,248 as tested for 4-node system with 4 3.5-inch 2TB SATA drives. Each node with 2-socket Intel Xeon E5-2600 server with 16DDR3 DIMM slots, onboard BMC, onboard SATA controller, PCI Express x8 Gen3 custom mezzanine slot, PCI Express x16 Gen3 low-profile slot.|
This story, "Review: Dell simplifies the blade server," was originally published at InfoWorld.com. Follow the latest developments in computer hardware,servers, and the data center at InfoWorld.com. For the latest business technology news, follow InfoWorld.com on Twitter.