IBM Unveils Powerful New EC12 Mainframe
IBM Corp. this week announced its new zEnterprise EC12 mainframe -- its biggest and brawniest Big Iron system to date. Thanks to a new analytics-based health assessment capability, it might be the smartest-ever mainframe, too.
“It is architected for the cloud, and, of course, it is bigger, badder, and bolder, with 25 percent better performance than the previous zEnterprise system, and it also adds 50 percent greater capacity,” said Rod Adkins, senior vice president of IBM's Systems and Technology Group, during the EC12 launch event. Adkins described the EC12 as “the result of an investment of over $1 billion ... and the work of more than 2,500 ... engineers across 18 labs worldwide.”
Big, Bad, and Bold
First, the bad news: no instant gratification. The EC12 won't officially ship for a little over three weeks, but when it does, the system will come outfitted with a new 5.5-GHz CMOS, which Big Blue says can deliver up to 25 percent more processing power.
In terms of clockspeed, this marks a slight increase (approximately 6 percent) over IBM's flagship zEnterprise z196, which clocked in at 5.2-GHz; but Armonk likewise expects to push its latest revision of Big Iron CMOS up to 6.0 GHz. The EC12 boosts its maximum core count by a fraction -- five cores -- clocking in at 101 cores compared to the z196's 96.
The EC12 puts on muscle in plenty of other places, however. For example, Big Blue beefed up its Level-2 (L2), Level-3 (L3), and Level-4 (L4) caches, to 2MB, 48MB, and 384MB, respectively. That's one-third more L2 cache (1.5MB) and double the L3 (24 MB) and L4 (192MB) cache of the z196. Memory hasn’t changed: the EC12 ships with the same 3TB of RAIM as did its flagship predecessor, the z196.
The EC12 incorporates a new solid state disk (SSD) memory tier that IBM says can help improve its overall availability story. Add in a new Crypto Express 4S cryptographic card, an array of Java-specific enhancements, and improvements to Big Blue's Unified Resource Manager (URM) and you have a strong set of new features.
That's not all. The EC12 touts another mainframe first: an embedded analytic firmware stack -- dubbed zAware -- that can perform analytic health assessment on z/OS OPERLOG messages. zAware comes out of IBM's Haifa (Israel) lab. According to the company, it uses “self learning” to profile 90 days of system behavior. From there, it can analyze z/OS OPERLOG messages for anomalous events.
“This is true predictive analytics to help avoid downtime and outages,” explained Adkins, who added that “the industry average in terms of financial impact for an outage can be more than $500,000 so this is a very important feature.”
Shock and Awe?
Count industry veteran Joe Clabby, a principal with Clabby Analytics, is duly impressed.
“[T]he EC12 is very highly focused on workload optimization for existing traditional applications, as well as for next generation Java, SAP, and business analytics workloads,” he writes. The EC12 does include the requisite manageability features, but Clabby sees it as more of a performance release. “[L]ook more closely and you’ll notice that improving processor speed and data serving, combined with the new instructions and workload optimization facilities, have one primary focal point: to make certain that workloads critical to enterprise customers run faster.”
IBM's mainframe technology dividend is likewise alive and well with the EC12, notes Clabby.
“From a cost perspective, IBM is offering a slight $/MIPS [dollars per millions-of-instructions-per-second] cost reduction on the EC12 as compared to the $/MIPS cost of the z196 -- meaning that processing existing workloads on standard, general purpose z workloads is going down in price,” he writes, adding that IBM is likewise discounting specialty engines (with reductions of approximately 20 percent on IFLs, zIIP, zAAP, and so on) and plans to cut maintenance costs by between 2 and 20 percent.
Industry watcher Anne MacFarland, a principal with MacFarland Consulting, echoes Clabby's assessment of the EC12's performance bona fides. For starters, says MacFarlane, the EC12 is larger and (potentially) cooler than its flagship predecessor. “As usual, the physical capacities of the latest System z are larger than previous IBM mainframe solutions -- the largest zEC12 [viz., the Model HA1] is a 101-way system!” she writes. “A new radiator-based air cooled design is embedded in the zEC12. Despite the larger capacity and improved performance, the footprint and energy draw of the zEC12 remain the same as the previous generation system, which is very good news.”
According to MacFarlane, “water cooling ... can cut data center energy savings up to an additional 9 percent, and the optional HV DC power, in a new data center, can reduce the power input power by 7-12 percent.”
MacFarlane, too, applauds the economics of the new EC12, which she says “will give a 2 percent [to] 7 percent price-performance for MLC products at flat capacity upgrades from the z196 to the zEC12.” In addition, she explains, “software for IFLs will benefit from 25 percent more capacity at the same PVU rating. Hardware maintenance also delivers a price performance benefit on flat capacity and greater benefits with platform growth. Maintenance pricing for IFLs will see a 20 percent price performance improvement, on a per MIPS basis.”
Although the EC12 ships with a revamped zSeries Blade Extender (zBX) -- namely, the Model 003 -- existing customers aren't out of luck: if you buy a new EC12 system, MacFarlane points out, you're eligible for a “no-charge” upgrade to the Model 003.
Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.