I guess when I say unified I mean unified in management, not unified in faceplates. HP has many different interfaces to manage the contents of a single blade chassis – let alone any storage that falls underneath it. HP blades are just servers. HP switches are just switches. The fact that they are in the same chassis is coincidental and simply a matter of location, not management or functionality. In the case of the UCS, the fabric interconnect houses the management console as well as the profile of the server, and the server blade hardware interacts with the management logic in the switch to acquire it’s outwardly facing properties such as MAC address and BIOS settings. If a particular blade is removed, it’s replacement takes on the identity of the prior blade all the way down to those properties normally associated with the hardware itself. This is increasingly important as virtual environments have increasingly complex network structures where standardization is key.

From this, I see the UCS memory, CPU, and network unify into a dynamic, centrally managed entity. I would extend this paradigm identification to the chassis servers as well that integrate into the UCS framework to be centrally managed with the blades. I fully suspect that Dell and HP could do the same thing but the fact is they haven’t and they treat their blade systems as if they were typical server/switch facilities, just in a smaller package.

In terms of workloads, yes, UCS is aimed at the x86 market for sure. I guess Cisco just has a bad habit of developing to what the market is asking for. While I have nothing but respect for the robust nature of UNIX, iSeries, Mainframe etc. the market is generally moving away (dated but makes my point: http://www.computerworld.com/s/article/9177445/Unix_server_sales_sink_IDC_cites_a_perfect_storm). X86 hardware augmented by improved platforms such as VMware provide excellent reliability for all but the most highly availability demanding workloads at a fraction of the costs. So, no, the UCS is not a replacement for a mainframe, but the UCS is not trying to be.

In terms of oversubscription, yes, they can be oversubscribed, but the 10G converged networking platform is a major boon to virtual environments that have historically had to dedicate 4 to 7 connections to get the job done and even then the links were not balanced to ensure they actually benefitted fully from all that connectivity. New backplanes in the Nexus line are evolving very quickly which again, could certainly be oversubscribed, but what an evolution. Beyond the bandwidth we must also recognize the flexibility that the virtual interface card (VIC) has brought as well, where a single controller card can emulate over a hundred different cards to provide logical traffic isolation. Yes, great companies like Xsigo do this and more with switching and cards but Cisco has integrated this technology into the UCS as a whole, into the single management interface, and into the concept of the server profile.

Ah yes, margins.

I get the sense you believe Cisco sells at list price. Uhhhh, no…

I have to concede that the UCS sale has a lower margin than a Catalyst 6500 sale. However, the switching market has long been commoditized and servers are the very icon of IT commoditization. Speaking from experience I have seen many true HP blade diehards admit that they see a benefit in the UCS management capabilities and have made the switch. If you purchase a UCS, you are purchasing the fabric interconnects and also the fabric extenders, so when Cisco sells servers, they sell two other major elements of their portfolio and therefore increases the revenue to the company from a “server” purchase. I will happily also concede the per sale margin on the UCS may be lower than other larger Cisco product sales, but the net is that UCS is bringing Cisco a bigger footprint in the data center and they are not losing money doing it. To continue to grow as a company with such immense market penetration in the route switch space they can either innovate their products are or widen the offering and between Nexus and UCS. Cisco is doing both.

No cost? It’s true, somewhere there is a cost. But I did get my cell phone for free if someone else was going to give me the same contract at the same price without a phone? So if Cisco gives me a great management interface with the UCS at the same price for another manufacturer’s blade center that does not (or cannot) provide said interface, I’ll call it free.

I assure you, I do not believe that the UCS is the silver bullet for server/network integration, but I really like where Cisco went with it – especially for the virtualization workspace. If the UCS spurs the likes of HP and Dell to innovate and unify their products, then all the better. But I strongly believe that the UCS is different than anything else in the server/switching space, and it is helping Cisco continue to grow and thrive as a company.

- David

Great article, you bring up a lot of different points that must be addressed in the IT workforce.

CIO Talk Radio has a show coming up on January 25th at 9AM Central, 10AM Eastern, 7AM Pacific called “IT Workforce Skills, Competency, and Engagement: The Road Ahead” with guests Pablo A. Vegas, VP and Chief Information Officer at American Electric Power (AEP), Maurice Tayeh – SVP and CIO at the Shaw Group Inc. and Art Lofton, CIO and VP of IT Solutions at Northrop Grumman Aerospace Systems.

It would be great for you to listen in and see if this show can answer any of your questions about what is needed in the IT workforce from the understanding of the key IT executives. Our website is http://www.ciotalkradio.com... I will follow up with you after the show on some of the key points if you are interested.

Does Cisco UCS differentiate itself in the market place?

No, the same virtual fabric interconnect of servers and FoE storage is being provided by HP and, after the acquisition of BNT, IBM. Either of these providers are better suited to discuss “unified computing” as they own storage sub-systems (which Cisco does not, EMC partner) and system management and provision tools (which Cisco does not, BMC partner). Cisco is a switch provider which has white-boxed some x86 servers. Even Dell, the laggard of the industry, will be able to duplicate this architecture with Force 10.

Also, UCS is only unified if all your workloads are x86, not realistic in large companies with mission critical Unix, mainframe, etc. Even if you had all x86 workloads, you can easily oversubscribe those switches. In a real world scenario of 4:1 Network oversubscription, (2) 6140′s and (1) UCS Manager would be required per rack.

UCS will increase Cisco’s margins?

How in the world do you think UCS will raise Cisco’s margins? They are accustomed to having little, if any, competition in the switching space which has allowed them to charge ludicrous 60% gross margins on their commodity switches. Even if Cisco retains their networking customers, they will retain them at much lower margins as HP and Juniper are offering the same performance and QoS at 50% the TCO or less. Cisco took on the rest of the IT industry, with the exception of EMC and BMC, when they released UCS. This will turn networking margins into x86 server margins. Competition makes margins go down, not up.

Cisco UCS Manager is a “no cost” option?

Do you think you are getting your cell phone at “no cost” when you sign a two year contract? No, the price of the phone is baked into the service cost. The Cisco UCS Manager cost is baked into the switch and SmartNet costs. I assure you that BMC is not giving them the software for free and they are no giving it away at “no cost” because they like losing money. There is a difference between “included in the price” and “no cost.”

Interested in what you may have found about tape out with Avamar. I am also looking at Data Domain and tape out options.

Thanks,
Dan Dwyer

Preston Deguise has a great article here – http://nsrd.info/blog/2011/05/07/there-is-no-such-thing-as-a-backup-failure/ – addressing the latter and the importance of resolving your backup issues as they represent an inability to recover your data. Perhaps it’s worth mentioning?

We find most customers interested in array based snapshot backups are primarily focussed on improving their recovery times where traditionally this may have taken upwards of a day to recover a tape full of data and they can instead rely on the ability to mount the snapshot or revert their data to a known good state. I’m a massive champion of snapshot technology and it’s great to see others including David Chapa – http://www.navigatingthebarscene.com/2011/05/notorious-b-i-g-data-infamous-b-i-g-backup/ – giving us all a firm shove into it’s adoption .