The good and bad of Cisco's UCS servers

By now you've read numerous blogs, articles, and tweets about the Cisco Unified Computing System (UCS) announcement. While this event may not carry the same weight as the IBM PC, System/360, or VAX, it is probably the most significant server announcement in many years.

Cisco deserves a lot of credit for its industry chutzpah. John Chambers and Co. were willing to risk deep relationships with HP and IBM to enter the server market. In this way, Cisco is adding new innovation to an old market and shaking up the industry as well.

OK, so what about the UCS products? Here is my quick evaluation:

Pluses

1. Innovative packaging that requires less rack space, power, and cooling than a standard blade server.

2. Designed for tight integration with server virtualization and the network.

   a. Cisco Virtual switch (i.e. VN-Link) replaces VMware switch. This links virtual and physical networking policy and management.

   b. Cisco adds extra memory to its server platforms, which enables it to increase the ratio of virtual servers hosted on each physical server.

3. Cisco manages the entire UCS virtual data center with one management platform. Cisco management can be integrated with other management platforms from vendors like BMC.

4. The overall strength is in integrating and improving both storage and network I/O. In this regard, Cisco could have a significant performance advantage in large data center deployments.

Minuses

1. Extremely proprietary architecture. Heck, Cisco is implementing its own version of Ethernet (What is more standard than Ethernet, for heaven's sake?) to consolidate storage and network I/O. The "real" standards won't be in place for another year or two.

2. This is a brand new arena for Cisco where its market share is 0 percent. With Dell, HP, and IBM well established in this market, expect enterprise CIOs to proceed with extreme caution.

3. The advantages of this architecture are minimal in a mixed environment. Today, all enterprises have other servers, and heterogeneous server support is not a core feature of this announcement.

4. Systems management has always been a Cisco weakness. HP and IBM are much better positioned here.

Cisco is aiming for the clouds both figuratively and literally. It is betting that its highly integrated virtual UCS is the best fit for massive data centers and cloud computing. This is probably true as of now, but Dell, HP, and IBM can certainly respond with product enhancements and open standards to bridge this gap. In the meantime, Brocade and Juniper should benefit immensely as server providers look for Cisco alternatives. HP will likely buy Extreme Networks or Force 10 to bolster the high end of its ProCurve networking product offering.

Ten years from now, tech industry historians will remember at least two things about 2009: the economic mess and the Cisco UCS announcement. If nothing else, Cisco just made the industry much more exciting than it was last Friday.

[jpr85635]

Relative to today's economy, this article is missing some key information. The cost of the Cisco solution compared to other vendor products. While cost is not necessarily the final determining factor, it certainly does influence the final decision.

[meh130]

Not sure where you got this idea of the Cisco systems using a proprietary version of Ethernet. What Cisco calls "Data Center Ethernet" (DCE), Intel calls Enhanced Ethernet for Data Center (EEDC), and other vendors call "Converged Enhanced Ethernet" (CEE), is standards based:

802.1Qaz Enhanced Transmission Selection
802.1Qbb Priority-based Flow Control
Data Center Bridging (Managed by the IEEE Data Center Bridging Task Group). DCB is expected to leverage functionality provided by 802.1AB-2005 Link Layer Discovery Protocol
802.1Qau Congestion Notification

Yes, the standards are a work in progress at this point, but products from multiple Ethernet switch vendors already support some of these features in a pre-standards for, all major vendors are committed to supporting the final standards, and Ethernet switches supporting these features will be available from multiple vendors.

Today Cisco, Intel, Emulex, QLogic, Brocade, and Blade Network Technologies are just some of the vendors supporting DCE/CEE features. IBM is also a strong supporter of Enhanced Ethernet.

[wolfe135]

Sun Mircosystems have the same types of systems that are not proprietary are very simple and cost a hole lot less! Why would you even look at a Cisco system. Look at the Amber Road systems from Sun.

[RMB00]

I was expecting a more "consolidated" offering which would include firewalls and server load balancing - services which Cisco products have been providing for a long time.

As it stands, I assume that to firewall or load-balance one or more of these servers, you would still need separate, external Cisco firewalls or load balancers.

It seems to me that just selling "another" blade server offering isn't all that big of a deal - regardless of whether or not it may include some extra virtualization or higher density features which the other vendors will match in a few months.

They need to add some Cisco secret sauce - the firewalls, and load balancing, and routing, and other network services which is Cisco's core business. Basic ethernet switching doesn't cut it.

[dwreid]

Frankly I couldn't care less if the industry got more exciting. A mushroom cloud on the horizon is exciting. The question is "did it make it better"? If all we have here is yet one more blade center at Cisco's famously high price point then then the industry didn't get better or exciting. Spare me.

[rshimizu12]

I am waiting to see how Cisco's unified management is. Next thing Cisco is going to tell us that UCS runs on Cisco IOS.

I think one of the biggest things about UCS is the fact that hardware is configurable. The extension of the PCI bus has enabled expressing what we traditionally think of a server as an xml file. There is no standard way to doing this today, one would expect that standards might emerge from it. I think we will see CSCO and Intel both trying to move virtualization into hardware, one at a chip level and one at a systems level.

http://bit.ly/djzPz

[kwieberdink]

Seeing this thing first hand will show most people that this is not your typical blade server.. In my opinion Cisco has come to market with a blade server that is light years ahead of the competition. Shame on the other manufacturers for not having done this years ago. HP, IBM, Sun and Dell have been in this space a long time, doing the same thing year after year with no staggering changes in how to manage and provision servers.

True Stateless computing is actually possible with UCS and that's just one of the features that is different from the others. HP, IBM, Dell and SUN still have an underlying BIOS & hardware firmware that can't be moved around at will like UCS. BIOS and firmware can have real impacts on a server with an OS that is moved around from on physical to another. Some manufacturers can move the WWN and MAC addresses, but I haven't seen them apply the same EXACT bios and firmware.

I have to get back to work, but I would suggest you take a look at is and not assume it's just another blade server. From what I'm seeing, it's about time someone started to push the other server vendors and change the paradigm.