Integrated Management, Commodity Servers, and Lock-In
Conventional wisdom says that clouds are built on inexpensive commodity servers. Scale out, reliability through redundancy, cheap hardware, server interchangeability – through all these factors, the cloud drives the cost of servers down. So I find it interesting that the leading server vendors have been investing in high-end server platforms targeting large cloud deployments. Cisco has garnered lots of attention from its move into the enterprise server market, with its Unified Computing System platform. With its BladeSystem Matrix, HP intends to extend its lead. And IBM’s iDataPlex promises mainframe-like features on an x86 platform. The capabilities of these platforms are very impressive.
All of these systems require large up-front capital; they aren’t what you’d call inexpensive to purchase. Is this a last-gasp attempt of hardware vendors to preserve margins and attempt to stave off commoditization in the cloud era? The answer may surprise you.
In the course of my work, I spend a lot of time talking with enterprises around the world – enterprises in a variety of industries, with different philosophies to technology adoption, and in very different financial conditions. Almost all of these companies are either deploying private clouds, or planning to deploy them. These customers also include the world’s largest public cloud providers. I routinely ask them their opinion about these high-end new server platforms.
The majority plan to use inexpensive servers to power their clouds. But a significant number have either begun deploying high-end servers in their clouds, or plan to, based on their early evaluation. Why? It turns out that, for many companies, the vendors’ promises of substantially lower operating expenses are compelling. A significant number of companies are willing to trade off higher capital expense for lower management and administration costs.
Of course, to realize these benefits, the customers must standardize on one server vendors’ platform. A lot of the benefits vanish when an enterprise has to manage incompatible systems from multiple vendors. Instead of a cross-platform management framework running in software, important layers of the management system are incorporated into the hardware. It has benefits, but it also has costs and risks, such as being wedded to one server vendor.
Back in the days before cloud, enterprises typically had policies to procure at least 20% of their server capacity from a second vendor. This emphatically showed their primary vendor that, if the customer didn’t get a great price, the customer could switch to another vendor. It had the desired effect: server prices dropped and dropped and dropped. It only worked because – and I apologize to all my friends in the commodity server business for saying this – servers used to be boringly interchangeable.
That changes as these innovative server platforms become more popular. If these servers live up to their promise, customers will benefit greatly from enhanced efficiency and flexibility. While conceding a significant amount of power (including economic power) to the vendor. The customers I’ve spoken with are fully aware of these tradeoffs, and trying to balance them. This is going to be interesting to watch.
