In-Depth
Update: State of the Mainframe
Demand for mainframes dropped off late last year -- and was down in the first three months of 2012, too. Is it time for Big Iron boosters to start worrying? No -- it's all good.
IBM Corp.'s System z mainframe came up short again in the first quarter of 2012, posting revenues that were off by a whopping 20 percent, compared to the year-ago quarter.
Is it time for Big Iron boosters to start worrying? After all, although one bad quarter can be dismissed as an aberration, successive bad quarters suggest a trend.
Veteran industry watchers don't think so, however. For one thing, IBM itself predicted a slow-down in mainframe sales. For another, several other trends -- starting with the promotional momentum that Big Blue's putting behind Big Iron -- favor System z.
There's also the come-to-Leviathan factor. When Oracle Corp. introduced the second revision of its Exadata platform, complete with a combined OLTP and data warehousing (DW) pitch, it talked up the benefits of a single large system. According to industry veteran Dan Kusnetzky, a principal with the Kusnetzky Group, there are situations and contexts in which a single big system just make sense.
"Why do mainframe configurations find their way into the data centers of new customers? The answer appears to be the same now as it was in the 1970s and 1980s," writes Kusnetzky, in a recent report. "For some workloads, a single, very large system configuration works better and is easier to manage than a veritable herd of industry standard systems."
For the last half decade (and more), IBM has pushed precisely this point in its Big Iron marketing. According to Kusnetzky -- who in a former life was a highly respected system software analyst with International Data Corp. (IDC) -- it's a message that has especial salience for enterprise customers. "If one examines cost of ownership or return on investment studies, staff-related costs of administration and operations far outweigh the cost of systems and software. So, any platform that minimizes those expenses can actuallyn a lower overall costs," he writes.
Behind System z's Slump
System z sales were off last quarter compared to a blockbuster quarter a year ago. IBM CFO Mark Loughridge acknowledged as much in an earnings call last month with financial analysts:"[L]ast year, we had a very strong first quarter, driven by product cycles in mainframe and POWER, so our growth rate this quarter reflects that difficult [comparison], though we continued to take share in UNIX."
System z had such a strong year-ago quarter because it was riding plenty of pent-up demand for its next-gen zEnterprise mainframe. In the months prior to zEnterprise's release in July of 2010 -- with general availability following in late-September -- sales of IBM's z10 mainframe had seemingly evaporated. As a result, Big Iron posted one of weakest-ever performances in Q2 of 2010, prompting Big Blue to actually move up its announcement of zEnterprise by a few weeks.
The rest is, as they say, history. Literally. In the last three months of 2010, for example, System z set sales records. Big Blue's next-gen mainframe was a powerhouse performer through most of 2011, and in the first three months of 2011 -- the year-ago period Loughridge referenced in IBM's recent earnings call -- Big Blue moved more than $1 billion in Big Iron hardware. On a year-over-year basis, System z's performance in the first three months of 2011 was up by almost half (41 percent) from the same period in 2010. That's quite a performance, but in Q2, IBM did even better, generating $1.2 billion in mainframe sales, a 60 percent increase over Big Iron's Q2 performance in 2010 according to market watcher International Data Corp. (IDC).
By July of last year, however, Loughridge was warning that the System z troika might soon be slowing down. "The front end of the mainframe cycle [is] defined by new placements and new placement growth[;] as we go into the back half of the cycle ... [it's] more defined by ... less revenue, but more profit margin performance," he told analysts in an earnings call last summer. "As we go into the second half of the year, we do start to wrap on that substantial performance that we saw out of the mainframe [over the course of the] last year."
(Big) Iron-Clad Staying Power
System z's slump isn't unexpected, but Big Blue typically revises its mainframe CMOS every 24 to 30 months. Customers know this. Customers also know that there are price/performance incentives associated with holding out for new mainframe hardware.
Might we expect demand for zEnterprise hardware to start dropping off even more precipitously in the third and fourth quarters of this year, especially with new mainframe CMOS in the offing? Probably. Will this drop-off inaugurate a reversal of Big Iron's decade-long resurgence? Not at all. If anything, the mainframe has more going for it this time around -- i.e., during the zEnterprise boom-bust life cycle -- than it did two years ago, when demand for its z10 mainframe hardware first started to decline.
Historically speaking, there are several key differences. During the previous boom-bust cycle, IBM didn't have a zEnterprise BladeCenter Extension (zBX). It didn't have the ability to host non-mainframe workloads -- including Windows workloads -- alongside the mainframe, in a zBX sidecar. It likewise didn't offer its customers a means to manage mainframe, Linux, AIX, and (yes) Windows workloads from a single system, via its Unified Resource Manager (URM). True, IBM also didn't have a competing systems vision -- viz., PureSystems -- to take at least some of the sheen off of System z's hybrid computing bona-fides. After all, PureSystems -- like System z with URM -- aims to offer a one-stop shop for the provisioning, deployment, and management of heterogeneous workloads, much like System z claims to do with zBX and zManager.
Of course, PureSystems -- unlike System z -- doesn't have a completely integrated mainframe story: as veteran Big Iron watcher Alan Radding notes on his Dancing Dinosaur blog. "IBM's current approach for coupling PureSystems and System z is ... via Ethernet connectivity. It also entails the use of two distinct management tools: Flex System Manager (for PureSystems) and zManager for System z."
PureSystems complicates the one-stop, Big System message that IBM has carefully crafted for System z. By the same token, however, Big Blue continues to burnish zEnterprise's hybrid computing bona-fides. "IBM ... has enabled programmatic access to the zManager, expanded internal network communication between the zEnterprise and the zBX, and added support for virtual storage management," Radding writes. "Looking ahead, IBM already is planning zBX support for the next generation z and promises to more tightly integrate the zEnterprise with PureSystems. The zEnterprise, zBX, and hybrid computing apparently will be around for a while."
Big Blue is also tinkering with promoting System z in another cutting-edge context: that of virtual desktop infrastructure, or VDI. Last month, Radding drew attention to IBM's efforts with STASH, it's Smart Terminal Architecture with Secure Hosts. STASH -- which is actually a mutli-vendor effort (and which enjoys support from prominent defense contractors such as Raytheon Trusted Computer Solutions) -- could give IBM a compelling VDI option for System z.
"STASH challenges the traditional assumption that greater security and increased performance utilization comes with increased costs," Radding wrote. "What the zEnterprise brings to this endeavor is massive scalability, reliability, and security. Even the entry level z114 can easily handle dozens of management servers to support the desktops," he continued, citing an estimate by IBM's Jim Porell (CTO of IBM Federal) that zEnterprise-based STASH solutions could support between 14,000 and 28,000 desktops per server.
"[I]f VDI is a new z workload to your data center, which it almost surely would be, you could use the Solution Edition program to buy the z at a deep discount and pick up a zBX also at a discount, although not quite as deep. That should let you drive down the cost per desktop even more," Radding observed.