Is IT Really Ready for Unification?

Arthur Cole

It's hard not to be impressed by all the unified computing platforms hitting the channel from Cisco, Dell and others. After all, these are mammoth feats of engineering incorporating the latest technological advances that, in one move, provide a full data center infrastructure that instantly becomes the envy of the neighborhood.

And yet, I can't help but think that these platforms could prove to be a tough sell, at least in the near term, and could find themselves swimming against some pretty tough currents in the IT industry.

First the news: Cisco Systems announced a series of major enhancements to its Data Center 3.0 portfolio this week aimed at ramping up the capabilities of its Unified Computing System in the drive to unite processing, networking, storage and virtualization into a single, cohesive system. The package includes new two- and four-socket servers featuring the Intel Xeon 5600 and 7500 processors, as well as a new pair of Nexus 10 GbE fabric switches, all governed by a new management stack that aims to more closely align data loads with available resources to help conserve energy.

A key component of the system is the FEXlink architecture, already available on the Nexus 2000 platform, that allows the new Nexus 2248 and 2232 fabric extenders to access legacy networks ranging from 100 Mb to 10 Gb Ethernet, as well as Fibre Channel and Fibre Channel over Ethernet (FCoE), InfiniBand over Ethernet, iSCSI and NAS. Not only does it provide upwards of 160 Gb connectivity per blade, but it cuts network management points more than 90 percent and knocks out close to 80 percent of existing cabling.

Cisco has a lot riding on this strategy, highlighted most strongly by the number of bridges it has burned to get to this point. Once a strong partner with HP, the two are now on parallel tracks for many of the same features and components. Witness HP's integration of video and voice collaboration tools from Polycom into its own UC platform. Polycom's chief competitor is Tandberg, which Cisco recently acquired for a cool $3.4 billion.

As I said, these are impressive systems and will no doubt find eager buyers looking to either expand existing infrastructure in a hurry or build it from scratch -- growing legions of new cloud service providers come to mind. The main issue I see with existing infrastructure is 1) integration with legacy systems, and 2) the real threat of vendor lock-in. Once you've installed a Cisco or HP unified architecture, there's no real turning back, so you better hope their customer service can meet your needs. Plus there is the fact that there are probably a few users out there that may already have HP infrastructure hooked up to, say, a Tandberg system or two. HP says it will continue to support Tandberg products, but for how long?

The other option is to build a unified platform based on open standards, which would give you an integrated architecture but still allow enough flexibility to develop customized systems and services through third-party providers. That's the approach Dell is taking as it looks to transition from a pure hardware provider to a full data center solutions partner. The company is focusing its energies on infrastructure, application and data management through a mix of in-house offerings and those of partners like Kace and Perot Systems.

Now, to be sure, open platforms have their own sets of issues, namely hefty integration costs and more complicated management stacks. But at least you gain a higher degree of flexibility for your trouble, with the added advantage of having a broader set of solutions to target specific needs.

So the question comes down to this: Have you had enough of the cobbled-together data center that you're willing to sacrifice some of your autonomy for a more cohesive environment? Or do you prefer to maintain a more free-wheeling environment that can be tweaked at a moment's notice?

The collective answer will determine the future direction of the IT industry.

Add Comment      Leave a comment on this blog post
Apr 13, 2010 4:31 AM Omar Sultan Omar Sultan  says:


You bring up some good questions in your post--ones that everyone should consider.

From our perspective, a couple of thoughts on vendor lock-in and integration into legacy environments:

1) The UCS is built upon open standards--x86 processors, standard DIMMs and hard drives, northbound links that support standard 10GbE and Fibre Channel.  FCoE is a ratified standard, the underlying IEEE DCB standards are well on their way to ratification and we will support .bh and .bg once they are standardized.

2) With the exception of UCS Manager, every part of the UCS is available as an unbundled product.  For example, we offer the Nexus 5000 series rack switches, Nexus 2000 FEXs, Nexus 4000 FCoE capable blade switches, so, if you like your existing servers, but like our networking, you do not need to choose between the two.  As you note some customers want us to do the integration and choose UCS, while others want to do their own integration.

3) We certainly do not expect our customers to forklift their existing servers no matter how much they may like our UCS.  Our expectation is that data centers will remain heterogeneous server environments for a long time to come.  With that in mind, we have spent a great deal of effort ensuring our UCS easily integrates into existing management environments.  The system is directly manageable via SNMP or XML (the entire XML schema is published), the system supports existing standards such as PXI, and we have integrated our UCS Manager with enterprise management tools from BMC, CA, EMC, HP, IBM, Microsoft, Symantec, and VMware.


Omar Sultan


Apr 14, 2010 1:47 AM Ken Oestreich Ken Oestreich  says:


You raise valid points.  Unified computing platforms, aka converged infrastructure, hold great promise. Egenera delivers on that promise now.  The major benefits of converged infrastructure are simplifying data center management and lowering operating costs (by reducing or eliminating some hardware and cabling) while leveraging existing standards and investments. You're also correct about the risk of vendor lock-in, particularly with Cisco UCS and HP Matrix. These vendors have a clear hardware agenda.

I agree that using open standards is a very viable approach. This is the path Egenera has taken. However, I disagree that there is an inevitable trade off of integration for agility or that converged infrastructure means incurring 'hefty integration costs' and increasing complexity. Customers using Egenera PAN Manager, market-tested converged infrastructure software, have the opposite experience because it:

1. Runs on commodity, standard x86 blades, standard Ethernet and standard switches, yet uses converged networking.

2. Has embedded VM management but allows the use of third-party VM managers.

3. Provides a single console for provisioning physical or virtual servers, high availability failover, disaster recovery and rapid provisioning and configuration.

4. Does so for about 30% less than alternatives. Our pricing is transparent.

I might also add that from an integration cost, our approach has been to converge management as well.dispensing with being required to integrate multiple software management systems too.  In this integration area in particular, I welcome a closer look/comparison of what a true 'cohesive' environment really looks like.

~ Ken Oestreich, Egenera


Post a comment





(Maximum characters: 1200). You have 1200 characters left.



Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.