Beyond x86 Servers: Should You Take a RISC?

by Drew Robb

As x86 servers increasingly dominate the landscape, is there room for RISC? Absolutely, says one prominent analyst.

Previously, we showcased the argument (and supporting stats) as to why the vast leaps in performance of x86 servers are signaling the beginning of the end for RISC/UNIX. Now, we will look at why you should consider purchasing a RISC/UNIX system.

It seems hard to imagine a future with only one kind of chip, yet incessant yapping continues about "industry standard" systems -- i.e. Wintel (with perhaps a minor concession to AMD and Linux) with the implication that anything else is inferior and sub-standard.

Dan Olds, an analyst with Gabriel Consulting Group, is firmly opposed to such a view. He uses a truck analogy to clarify: Once upon a time there was only one kind of vehicle –- a big truck, and that's what everyone drove because that's all they had. When different, smaller, less-expensive cars were introduced, people understandably flocked to them. But that doesn't mean they no longer needed trucks. No matter how big and roomy the trunk of the family car is, it's not like the bed of a dump truck. No matter how powerful the engine is, it's never going to become the mainstay of inter-state freight. Thus, Olds doesn't believe the future is going to be systems based only on x86 processors.

"While x86 systems will sell more units and dollar volume, they will not take over the world entirely," said Olds. "There will still be plenty of market for RISC-based UNIX systems and mainframes."

To his mind, the value proposition of non-x86 systems will continue to be performance, high availability and security. Vendors of non-x86 systems, he said, can do things that are much more difficult to accomplish in x86 systems. Non-x86 vendors own all of the parts that make up their solutions -- the hardware (including processors, assist processors and interconnects) and the software (OS, firmware, often middleware, and even apps in some cases), and they all have storage, too. This enables the OEMs to put together highly integrated and optimized systems that provide value to customers.

This ability to design and manufacture integrated solutions to match up to business challenges is what RISC systems do best, said Olds, and what we'll see more of in the future.

When pointed to the stats, Olds said he sees a different interpretation to steady decline due to the inevitable onslaught of x86. He offers an explanation for at least part of the drop in UNIX sales in recent quarters. Projects using UNIX systems are typically larger than those using x86 boxes -- not always, but usually. This means they require larger investments not only in hardware, but also in services, software, business disruption, and so on. In fact, in those types of deals, the actual hardware forms a relatively small part of the deal. When the economy went south, many decided to delay, postpone and even cancel any big cash outlay. This hit the bigger projects first -– easier to postpone those and a bigger payoff for doing it.

"I think that this is responsible for a significant portion of the UNIX vs. x86 drop off," said Olds. "Smaller projects didn't get delayed as much."

He doesn't deny that x86 is growing faster than UNIX –- but that doesn't mean UNIX is going away, just as it doesn't mean mainframes have gone away. It also depends on whether we're talking revenue or shipments. x86 is ahead on both counts, of course. But Olds doesn't think UNIX sales are fewer than the number of machines being replaced with x86.

"If you took the amount of capacity that is shipping out every year in new UNIX and mainframes, that number is bigger than the amount of systems that are being decommissioned every year," said Olds. "The overall pie is still growing, and each platform finding its place in the data center."

Regardless of the advances of x86, UNIX and mainframe systems continue to run the lion's share of important workloads that require the most scalability, availability and predictable performance.

"UNIX is still king of the hill when it comes to mission-critical applications," said Olds.

He backs this up with some recent survey results on UNIX users that his company carried out: 68 percent of survey respondents said their overall Unix usage was increasing; 72 percent disagreed with the statement "Unix usage is declining at the midrange/high-end" while 10 percent agreed and 19 percent weren't sure; and 91 percent agreed with the statement, "Unix platforms are strategic in our organization."

Olds sees a lengthy continuation for UNIX, then, due to what he calls the IT pecking order. Users will change system hardware before they change operating system, and they will change hardware and OS to keep the same database. Then, the applications that sit on top of databases are most important of all. Customers will change anything under them to avoid having to change the apps and all of the business logic they've built on top.

"Customers are loath to change things that are working well and need to see a huge advantage to take the risk when we're talking about something like an enterprise database or set of important applications," said Olds.

Drew Robb is a freelance writer specializing in technology and engineering. Currently living in California, he is originally from Scotland, where he received a degree in geology and geography from the University of Strathclyde. He is the author of Server Disk Management in a Windows Environment (CRC Press).

Follow ServerWatch on Twitter

This article was originally published on Friday Mar 19th 2010
Mobile Site | Full Site