The Ominous Return of Closed Architectures
A Contributed Article by Lloyd Carney, CEO of Xsigo Systems
In the world of IT, “open” is practically synonymous with “good.” The terms “open source,” “open standards,” and “open systems” all imply efficiency and cost effectiveness.
But there is a trend towards closed system architectures that reverse this cycle. And it’s a trend you should keep a wary eye on.
For two examples, look at HP and Cisco. Both have server offerings that incorporate closed elements that are not now interoperable with other vendor’s products.
Start with Cisco’s UCS. You’ll need Cisco’s servers, their own CNA, and Cisco’s Fabric Interconnects as well as Cisco switches to achieve the promised flexibility.
Cisco has been clear that UCS is not a fully open architecture. Network World reported in this June, 2009 article, “Cisco will not license UCS technologies and has no plans at present to have the platform support multivendor servers. ‘Would Cisco license IOS?’ she asked rhetorically,” quoting Soni Jiandani, Cisco’s vice president of marketing for the server access virtualization group.
Next look at HP’s Blade Matrix. To get virtual I/O here, you’ll need HP’s Virtual Connect, a closed server I/O solution that works only with HP blades, not HP rack mount servers, and not other vendor’s servers.
In both systems, the server and virtual I/O are tied together. You may ask why they chose this approach since it’s not really necessary to weld virtual I/O to the server design, any more than it’s necessary to lock server virtualization software to any particular server. Xsigo virtual I/O achieves full I/O flexibility without being tied to a server.
Yet that’s what they chose. Whatever the motivation (and vendors will always claim there are benefits to be gained), bear in mind there are costs as well, costs that go beyond the obvious vendor lock-in.
Why open systems matter
The open systems revolution arose as a rebellion against mainframes (the ultimate incarnation of closed computing) and the closed “super mini” architectures. In that era, no one vendor had the full range of products needed to satisfy the emerging needs of a corporation. And each vendor introduced unique capabilities and features (sound familiar?).
The result was multiple architectures that could not communicate with each other: IBM mainframes running SNA over token ring, DEC super minis running DECnet Phase 5, Apple machines running Appletalk, and Prime computers running Primenet.
People soon recognized the impracticality of this approach, and the open systems movement began. In this revolution, portability and flexibility became paramount. Even IBM, the king of mainframes, became an advocate.
Now, with this trend seemingly reversing, it’s worth a look at the costs of a closed system and why we should be cautious about going back.
Pricing leverage
Vendor lock-in increases switching costs. You cannot change one component without changing others at the same time. If your server I/O, for example, is locked to one server type or switch type, that vendor becomes harder to displace. When it’s time to add gear, your vendors won’t compete on an equal footing.
The costs of a closed system may not appear on your first purchase. It’s the second or third purchase where, perhaps, software license fees that were at first heavily discounted suddenly become expensive. (On the servers themselves, vendors will need to maintain competitive list prices, so temporary price relief is more likely to appear in the form of discounted licenses or even in “free” gear.)
Best-of-breed technology
Vendors advance the state-of-the-art in non-linear ways. There are jumps that occur. For example, the first blade servers were launched early 2001 (by a vendor that HP acquired). It was 2003 before blades became widely available elsewhere. Do you want to be locked to one vendor’s platform when another has elements you can benefit from now?
The fact is, it is very difficult for any vendor to innovate across the entire data center. The closed architectures from Cisco and HP will eventually require you to use sub-optimal solutions until they choose to innovate or update their portfolios. But one thing they will not do is become more open. When the closed players (IBM, DEC , Prime, et al) decided to go open, they created unique versions of UNIX that were really disguised proprietary closed solutions.
Best-for-your-job technology
While maintaining a consistent server pool may have value, sometimes you just need the best tool for the job. If a commodity server is appropriate for a particular task, or a certain high-end design best suits your needs, you’ll want the flexibility to make that call on technical or business merits alone.
Mergers and consolidation
Change is a fact of business life, and mergers and acquisitions often become drivers of data center growth. With an open architecture, you have flexibility to re-deploy resources into jointly managed pools. That’s not possible when elements are bolted to specific vendor’s hardware.
Personnel changes
Personal preferences also factor into buying decisions. When management shifts occur, brand preferences can shift as well. An open architecture is far more adaptable than one that is rooted to a closed design.
The trend towards closed systems is of course a part of the larger trend towards increasingly monolithic suppliers. HP and Cisco were formerly partners. Now Cisco’s launched servers and HP’s acquired 3Com. And there is talk of more convergence to come, which will no doubt further advance this trend.
While we’re not yet back to the closed world of the mainframe and super minis, remember why we worked so hard to escape that world. And remember that vendor choice is the most powerful tool you have to ensure a competitive advantage for your data center.
Ten years ago, as we happily threw new servers at every task, how many people foresaw that server virtualization and power management would become critical to data center operations? I guarantee you that no one vendor has all the products for the emerging needs of your data center. They can’t because vision and foresight are uniquely creative processes. Which is why it’s important to keep your options open.
--------------------------------------
About the Author
Lloyd Carney, CEO, Xsigo Systems
Lloyd Carney's 25-year career encompasses a proven track record of leading and growing technology companies. Prior to joining Xsigo, Mr. Carney was general manager of IBM's Netcool Division, which acquired Micromuse where Carney had been chairman and CEO. IBM's Netcool Division provides IT and telecom infrastructure management tools to a variety of customers in enterprise computing, transportation, and wireless networking. Prior to Micromuse, Carney was COO at Juniper Networks, where he oversaw the sales, marketing, engineering, manufacturing, and customer service organizations. He also has headed up three divisions at Nortel Networks, including the Core IP Division, the Wireless Internet Division and the Enterprise Data Division. Mr. Carney currently sits on the Board of Directors at Cypress Semiconductor and at BigBand Networks.