Posts Tagged ‘AIX’

IBM POWER8 CAPI for Efficient Top Performance

August 21, 2014

IBM’s Power Systems Power8 Coherent Accelerator Processor Interface (CAPI) is not for every IT shop running Power Systems. However, for those that aim to attach devices to their POWER8 systems over the PCIe interface and want fast, efficient performance CAPI will be unbeatable.  Steve Fields, IBM Distinguished Engineer and Director of Power Systems Design introduces it here. Some of it gets pretty geeky but slides #12-17 make the key points.

DancingDinosaur first covered CAPI here, in April, shortly after its introduction. At that point it looked like CAPI would be a game changer and nothing since suggests otherwise. As we described it then, CAPI sits directly on the POWER8 board and works with the same memory addresses that the processor uses. Pointers de-reference the same as the host application. CAPI, in effect, removes OS and device driver overhead by presenting an efficient, robust, durable and, most importantly, a direct interface. In the process, it offloads complexity.

In short, CAPI provides:

  • SMP Coherence Protocol transported over PCI Express interface
  • Provides isolation and filtering through the support unit in the processor (“CAPP”)
  • Manages caching and address translation through the standard POWER Service Layer in the accelerator device
  • Enables accelerator Functional Units to operate as part of the application at the user (direct) level, just like a CPU

What you end up with is a coherent connected accelerator for just a fraction of the development effort otherwise required. As such, CAPI enables more efficient accelerator development. It can reduce the typical seven-step I/O model flow (1-Device Driver Call, 2-Copy or Pin Source Data, 3-MMIO Notify Accelerator, 4-Acceleration, 5-Poll/Int Completion, 6-Copy or Unpin Result Data, 7-Return From Device Driver Completion) to just three steps (1-shared memory/notify accelerator, 2-acceleration, and 3-shared memory completion). The result is an easier, more natural programming model with traditional thread-level programming and no need to restructure the application to accommodate long latency I/O.  Finally it enables apps otherwise not possible, such as those requiring pointer chasing (e.g. Java garbage-collection).

Other advantages include an open ecosystem for accelerators built using Field Programmable Gate Arrays (FPGA). The number and size of FPGAs can be based on application requirements, and FPGAs can attach to other components, such as private DRAM, flash memory, or a high-speed network.

Driving the need for CAPI is the insatiable demand for performance.  For that, acceleration is required, which is complicated and resource-intensive to build. So IBM created CAPI, not just for pure compute but for any network-attached or storage-attached I/O. In the end it eliminates the overhead of the I/O subsystem, allowing the focus to be on the workload.

In one example IBM reported it was able to attach an IBM Flash appliance to POWER8 via the CAPI interface. As a result it could generate Read/Write commands from applications and eliminate 97% of code path length, a savings of 20-30 cores per 1M IOPS. In another test IBM reported being able to leverage CAPI to integrate flash into a server; the memory-like semantics allowed the flash to replace DRAM for many in-memory workloads. The result: 5x cost savings plus large density and energy improvements. Furthermore, by eliminating the I/O subsystem overhead from high IOPS flash access, it freed the CPU to focus on the application workload.

Finally, in a Monte Carlo simulation of 1 million iterations, a POWER8 core with FPGA and CAPI ran a full execution of the Heston pricing model for a single security 250x faster than the POWER8 core alone. It also proved easier to code, reducing the lines of C code to write by 40x compared to non-CAPI FPGA.

IBM is just getting started with CAPI. Coming up next will be CAPI working with Linux, mainly for use with analytics. Once Linux comes into the picture, expect more PCIe card vendors to deliver products that leverage CAPI. AIX too comes into the picture down the road.

Plan to attend IBM Enterprise2014 in Las Vegas, Oct. 6-19. Here is one intriguing CAPI presentation that will be there: Light up performance of your LAMP apps with a stack optimized for Power, by Alise Spence, Andi Gutmans, and Antonio Rosales. It will discuss how to leverage CAPI with POWER8 to create what they call a “killer stack” that brings together continuous delivery with exceptional performance at a competitive price. Other CAPI sessions also are in the works for Enterprise2014.

DancingDinosaur (Alan Radding) definitely is attending IBM Enterprise2014. You can follow DancingDinosaur on Twitter, @mainframeblog, or check out Technologywriter.com. Upcoming posts will look more closely at Enterprise2014 and explore some session content.

New zEnterprise Business Class Entry Model—zBC12

July 23, 2013

IBM introduced its new zEnterprise Business Class machine, the equivalent of the z114 for the zEC12, the zEnterprise BC12 (zBC12).  It offers significantly more power than its predecessor but the $75,000 base price hasn’t changed.

The company has been hinting at the arrival of this machine for months (and DancingDinosaur has been passing along those hints as quickly as they came). Of particular interest is that the System z Solution Edition pricing applies to the zBC12. Solution Edition pricing should make the machine quite competitive with x86-based systems, especially when running multiple Linux instances.

IBM isn’t being coy about its intentions to discount this machine. The initial announcement touted a new Linux-only based version of the zBC12, the Enterprise Linux Server (ELS). The ELS includes hardware, the z/VM Hypervisor, and three years of maintenance at a deeply discounted price. Besides over 3,000 Linux applications it includes two new capabilities, ELS for Analytics and Cloud-Ready for Linux on System z, each acting as an onramp for analytics or cloud computing.

DancingDinosaur has been a big fan of the Solution Edition program as the only way to get serious discounts on a mainframe. The big caveat is the constraints IBM puts on the use of the discounted machine. Each Solution Edition program is negotiated so just make sure you fully understand the constraints and all the fine print so you can live with it for several years. Of course, a zBC12 can be used for anything you would use a mainframe although enterprise Linux serving seems  an ideal use.

Besides its faster processor the zBC12 also offers up 156 capacity settings on each model  to choose just the right capacity setting for your needs along with a new pay-as-you-grow approach. When it is integrated with the IBM DB2 Analytics Accelerator, the zBC12 can perform business analytics workloads with10x better price performance and 14 percent lower total cost of acquisition than the closest competitor, according to IBM.

Out of the box the zBC12 specs look good:

  • 4.2 GHz processor designed to deliver up to a 36% performance increase per core to help boost software performance for business-critical workloads
  • Up to six general purpose processors designed to deliver up to 58% more capacity compared to the z114, which had five general purpose processors
  • Up to a 2x increase in available memory (496 GB) compared to the z114 for improved performance of memory-demanding workloads such as DB2, IBM WebSphere, and Linux on System z

The zBC12 comes in two models, the H06 and H13. Both are air cooled, single frame, or support 30 LPARS. The H06 has one processor drawer for 9 processor units. These can be divided between SAPs, CPs, IFLs/ICFs, zIIPs and zAAPs, and 1 IFP.  The Model H113 has two processor drawers to handle 18 processor units. It allows the same mix of processor types but in larger quantities and 2 dedicated spares. There are configurations where the H06 requires the second processor drawer. The entry processing level is 50 MIPS, up from 26 MIPS with the z114 with no change in the base price.

As far as other pricing, the zBC12 follows essentially an extension of the z114 stack pricing with a 27% price/performance improvement over the z114 for specialty engine pricing, which translates in 36% greater performance for the money.  Pricing for maintenance remains the same. Software keeps with the same pricing curve with a 5% discount applied. The price of Flash Express for the zBC12 remains at $125,000.

IBM has provided a straightforward upgrade path from the z10 or zEnterprise to the zBC12 as well as from the zBC12 to the zEC12. It also can be connected to the zBX (Model 003) to seamlessly manage workloads across a hybrid computing environment consisting of multiple architectures (Linux, AIX, and Intel/Windows).

The announcement of the zBC12 was accompanied by a slew of other new z announcements, including the new IBM zEnterprise Analytics System 9710,and native JSON support to bridge the gap between mobile devices and enterprise data and services along with conversion between JSON  and the new CICS Transaction Server Feature Pack for Mobile Extensions V1.0 and DB2 11 for z/OS (ESP).  Plus there is the new z/VM v6.3 and enhancements to the z/OS Management Facility.

As DancingDinosaur noted last week, expect z sales to get a boost in the next quarter or two as organizations choose the new zBC12. With its improved price/performance and low entry pricing and the Solution Edition deal for the zBC12 ELS the z should see a nice bounce.

IBM Introduces POWER7+ and More

October 4, 2012

IBM’s Systems and Technology Group (STG) introduced a slew of new products and enhancements, both hardware and software, for the System z and Power. The System z announcements, which DancingDinosaur will take up in subsequent posts, focused mainly on software enhancements, such as new revs of CICS and Omegamon, for the zEC12. The Power announcements covered new capabilities as well as new machines. And all the announcements in one way or another address IBM’s current big themes: Cloud, Analytics, and Security.

Of the new Power announcements, Power7+ certainly is the star.  Other capabilities, such as elastic capacity on demand and dynamic Power system pools, may prove more important in the long run. Another new announcement, the EXP30 Ultra SSD I/O Drawer, may turn out quite useful as organizations appreciate the possibilities of SSD and ramp up usage.

Power7+, with 2 billion transistors, promises to deliver 40% more performance, especially for Java workloads, compared to Power7.  Combined with other enhancements Power announced, it looks particularly good for data and even real-time analytics workloads.  The new processor boasts 4.4 GHz speeds, a 10MB L3 cache per core (8 cores =80 MB), and a random number generator along with improved single precision floating point performance and an enhanced GX system bus. IBM invested the additional transistors primarily in the cache. All of this will aid performance.

The enhanced chip also brings an active memory expansion accelerator and an on-chip encryption accelerator for AIX. Previously this was handled in software; now it is done in hardware for better performance and efficiency.  Power7+ also can handle 20 VMs per core, double the number of Power7 VMs. This allows system administrators to make VM partitions, especially development partitions, quite small (just 5% of the core).  With energy enhancements, it also delivers 5x more performance per watt. New power gating also allows the chip to be configured in a variety of ways. The upshot: more flexibility.

Elastic capacity on demand (CoD) and Power System Pools work hand in hand.  Depending on the server model, you can create what amounts to a huge pool of shared system resources, either permanent or temporary. IBM has offered versions of CoD for years, but they typically entailed elaborate set up and cumbersome activation routines to make the capacity available.  Again, depending on the model IBM is promising more flexible CoD and easier activation, referring to it as instant elasticity.  If it works as described, you should be able to turn multiple Power servers into a massive shared resource. Combine these capabilities to create a private cloud based on these new servers and you could end up with a rapidly expandable private cloud. Usually, it would take a hybrid cloud for that kind of expansion, and even that is not necessarily simple to set up. The payback: greater agility.

There are, however, limitations to elastic CoD and Power Systems Pool. An initial quantity of CoD credits are offered only with new Power 795 and Power 780 (a Power7+ machine). There also is a limit of 10 Power 795 and/or 780 servers in each pool.

Enterprises are just starting to familiarize themselves with SSD, what it can do for them, and how best to deploy. The EXP30 Ultra SSD I/O Drawer, scheduled for general release in November, should make it easier to include SSD in an enterprise infrastructure strategy using the GX++ bus. The 1U drawer can hold up to 30 SSD drives (387 GB) in that small footprint.  That’s a lot of resource in a tight space: 11.6 TB of capacity, 480,000 read IOPS, and 4.5 GB/s of aggregate bandwidth. IBM reports that it can cut batch window processing by up to 50% and reduce the number of HDD by up to 10x. Plus, you can still attach up to 48 HDD downstream for another 43 TB. The result: great scalability and efficiency.

And this just touches on some of what IBM packed into the Oct. 3 announcement. DancingDinosaur will look at other pieces of the Power announcement, from enhancements of PowerVM to PowerSC for security and compliance as well as look at the enhancements made to zEC12 software.

HP-UX and AIX : The Difference is POWER7

July 10, 2012

HP’s enterprise-class UNIX operating system, HP-UX, faces a stark future compared to IBM’s AIX. The difference comes down to the vitality of the underlying platforms. IBM runs AIX on the POWER platform, now at POWER7 and evolving to POWER8 and even POWER9 (although the naming may change)—a dynamic platform if ever there was one. Meanwhile, HP-UX has been effectively stranded on the withering Itanium platform. Oracle has stopped development for Itanium, and Intel, HP’s partner in Itanium, has been, at best, lackluster in its support.

It not clear whether HP-UX is a better UNIX than AIX, but in an industry driven by ever increasing demands for speed, throughput, cost-efficiency, and energy efficiency, the underlying platform matters. HP-UX customers surely will outgrow their Itanium-based systems without a platform boost.

“There’s no question that [our] Business Critical Server business has been hurt by this,” said HP CEO Meg Whitman in the transcript of an interview with the Wall Street Journal’s All Things D column. The business, which had been growing 10% a year before Oracle spurned further support of Itanium now is declining by 20-30% a year (Ouch!).  So Whitman is counting on two things: 1) winning its lawsuit against Oracle, which is still making its way through the courts and 2) porting HP-UX to an advanced x86 platform, namely Xeon. “Ultimately we’ve got to build UNIX on a Xeon chip, and so we will do that,” she told All Things D. All spring long there had been hints that this was imminent, but an official HP announcement never materialized.

Of course Oracle wants the HP customers running Oracle on HP-UX with Itanium to jump to its Sun platform.  IBM, however, has been wooing and winning those same customers to its System z or POWER platforms. Oracle runs on both the z and POWER platforms.  Running Oracle on Linux on System z yields substantial savings on Oracle licensing. But IBM wants to do even better by migrating the Oracle shops to DB2 as well, with incentives and tools to ease the transition.

What HP customers also get when they move to POWER or to the z is a platform in both cases with a real platform future, unlike either Itanium or Sun’s server platforms. DancingDinosaur has long extolled the zEnterprise and hybrid computing, but POWER is dynamic in its own right and when you look at the role it now plays in IBM’s new PureSystems, another IBM hybrid platform, POWER becomes all that more attractive.

From the start HP with HP-UX and Itanium was bound to have to settle for compromises given the different parties—HP, Intel, Oracle—involved. With POWER7, IBM system developers got exactly what they wanted, no compromises. “We gave the silicon designers a bunch of requirements and they gave us our wish list,” says Ian Robinson, IBM’s PowerVM virtualization and cloud product line manager. As a result POWER7, which runs AIX, Linux, and System i on the same box, got a slew of capabilities, including more memory bandwidth and better ways to divide cores.

POWER7, which amazed the IT world with its stunning Watson victory at Jeopardy, also is turning out to be an ideal virtualization and cloud machine. The rate of virtualization and cloud adoption by POWER7 shops is running something north of 90%, notes Robinson. The adoption of PowerVM, the POWER7 hypervisor built in at both the motherboard and firmware levels is close to 100%. And now POWER7 is a key component of IBM’s PureFlex initiative, a major IBM strategic direction.

Meanwhile, Whitman is fighting a costly court battle in the hope of coercing grudging support for the Itanium platform from Oracle. The trial began in June and mud has been flying ever since. Even if HP wins the case, don’t expect the story to end soon. Using appeals and delay tactics Oracle could put off the final outcome so long that Itanium will have shriveled to nothing while POWER7 continues along IBM’s ambitious roadmap.

Addressing zBX Concerns

June 3, 2012

DancingDinosaur has been looking into user interest in the zBX almost since its introduction well over a year ago. Over that time adoption has grown from an initial handful to more than 125 devices this spring. While the devices aren’t exactly flying out the door, IBM has evolved it in ways that generally boosts its appeal.

For example, IBM improved its efforts at publicizing use cases. It is not intuitively obvious what a z data center would want to do with the device. IBM also has taken some baby steps in lowering the overall cost, mainly with a promotional deal for free blades—DancingDinosaur described it here—and by incorporating zBX discounts in System z Solution Edition packages. In fact a few of the Solution Edition deals offer what should become popular zBX use cases, specifically for the Enterprise Linux Server, GDPS, and SAP. Finally, the z114 provides a low cost first step to the zBX.

The introduction of Windows on x blades for the zBX also should boost the appeal of the device. DancingDinosaur spoke with several  recent zBX adopters who cited the availability of Windows as a key factor.

On the other hand, the elimination of the Smart Analytics Optimizer as a blade for the zBX has diminished the interest of others. The replacement, Netezza, entails different adoption and deployment case. It is not automatic to jump from the Smart Analytics Optimizer as a zBX blade to the Netezza appliance.

DancingDinosaur has spent the past few months speaking with mainframe data center managers about their concerns with the zBX. Below is a sample of their responses.

Cost, of course, is always an issue. This is how one IT manager put it: Even though the zBX represents a significant technological breakthrough, it is not clear to us whether it currently represents a sufficient financial breakthrough as to justify its introduction here at this time. He is waiting for either a compelling deal or a compelling use case.

Another issue is the alternative of continuing to run AIX, Windows, and Linux on their usual platforms. As one manager noted:  We opted to run AIX and Windows server applications on boards inserted into rack mounted servers. If we wanted blades we would just as likely use  a stand-alone Blade Center server, which costs a lot less than a zBX.

Some IT managers simply didn’t want to ignite a  platform war among the IT staff. Moving to a zBX implies consolidating at least some Windows, Linux, and AIX workloads on blades in the zBX. To undertake what might turn into a bitter political fight they need an indisputable business case and a top executive willing to lead the battle.

Even managers who are holding off on the zBX decision see advantages to the zBX and hybrid computing. Three in particular come up:

  1. Mainframe reliability and availability—they have Windows applications that suffer from downtime.  Running within the zBX under the zManager enables these applications to easily fail over to another virtualized server, thereby avoiding troublesome and costly downtime. Of course, they could assemble a failover architecture on their own but it would unlikely be as bullet-proof as what they get out of the box with the zEnterprise.  This alone might justify the investment.
  2. Staffing versatility—today most data center staff have platform-specific skills.  The zManager enables an administrator to attend to the full range of platforms through a single tool.  As IBM adds management expertise and automation it has developed for PureSystems to the zManager it should help to reduce the labor overhead associated with supporting the different environments. Again, the savings on staffing can go a long way toward justifying the zBX investment.
  3. Simplified backup and recovery—it is not a sure thing to backup and reliably recover a mix of virtualized and non-virtualized Windows systems plus Linux applications running on x-based servers. GDPS, available at a Solution Edition discount,  now can backup and recover the zEnterprise, zBX, and all the blades it contains fast and with confidence.

There are more advantages to zEnterprise-zBX hybrid computing than the three noted above. Watch for an upcoming Independent Assessment white paper that lays it all out. Similarly, concerns revolving around cost can be addressed—IBM has signaled in numerous ways that it is willing to deal. The immediate issue comes down to pulling together that first business case and making it a compelling one.

IBM zBX Blades Cheap (Free!)

December 8, 2011

T’is the season of discounts, and it apparently applies to the zEnterprise as much as it does to holiday gifts. This deal, however, does not appear to end with the holiday. DancingDinosaur supports anything IBM does to lower enterprise data center costs.

Here’s the deal: migrate competitive workloads to the zEnterprise Blade Center Extension (zBX) and you can receive up to six zBX Power or x86 blades for free. So, replace 2, 4, even 6 HP UX or Linux Systems or x86 servers (remember, there are now Windows blades) and receive an equal number of free Power or x86 blades to run in the zBX. The deal focuses on HP, but IBM staff says it applies to Oracle/Sun systems too. Given a $5000 cost for low end x86 blades, that could amount to a $30,000 discount on top of whatever other discounts IBM will throw in. For high end, more richly configured blade replacements it could come to much more.

Of course, you need a zEnterprise (z196 or z114) and to buy a new zBX to cash in on this. But, if you took a deeply discounted System z under the IBM Solution Edition program you could get in at a bargain price. And if your choice was a z114, IBM also is offering the DS8800 storage system for the z114 at an attractive entry price.

The deals are being offered under IBM’s Freedom by Design  program. IBMer Paulo Carvao details the blade offer here in a presentation titled System z: Delivering on the Promise of Smarter Computing. Check out slide #15.

Even without free blades, the z makes an attractive consolidation play. According to IBM you can consolidate an average of 30 distributed servers or more on a single z114 core, or hundreds in a single footprint. In effect, you can deliver a virtual Linux server for approximately $500 per year, which works out to be as little as $1.45 per day per virtual server. If you are consolidating Oracle servers, the savings in Oracle licensing costs alone would cover a big chunk of the investment.

A dearth of zBX blade performance data, however, has slowed zBX adoption for some. A little performance data, however, has started to trickle in. For instance, some recent results came from an Italian company that moved its SAP workload to POWER7 blades on a zBX. It was able to boost bill processing from 60K per hour to 430K per hour, better than a 7x increase.

In general, IBM blade performance in the zBX should be the same as the performance in its standard blade centers. Actually, it might be a little better since the consolidated zEnterprise-zBX combination can cut down the number of network hops in some situations. And IBM insists zBX blades are priced competitively like its standard IBM blades. And then there are the free blades with a competitive upgrade, with which no one will quibble.

While on the subject of zEnterprise deals, the regular prices of specialty engines continues to be a MIPS bargain, delivering more MIPS for the money than earlier versions. One customer used the increased MIPS from zEnterprise specialty engines to reduce the number of cores the company bought, which resulted in a substantially lower acquisition cost with no reduction in overall MIPS. With the right workloads, this is a very effective cost saving strategy.

IBM System z for Social Business

November 30, 2011

Most IT people do not think of the System z for social business. Probably they think more in terms of x86 systems running Linux or Windows. Some might think Power Systems and AIX. The System z, however, has a good story to tell when it comes to social business, and with the zEnterprise and zBX machines that story only gets better.

Social business today generally refers to some form of collaboration, information sharing, or interaction through social networking or social media. Blogging and micro-blogging, like Twitter, Facebook, Foursquare, LinkedIn, and such are what usually come to mind. Here’s a link to the Society for New Communications Research, which covers this topic in depth.

IBM, through its 1995 acquisition of Lotus Development Corp that brought it Notes, has been involved in social business for quite some time. At that time Lotus Notes was called groupware, but even then it did considerably more. Basically, it included what amounted to a Notes application development environment. Groupware was a vague concept back then. Still Notes, even then, went beyond groupware and shared documents.

Social business today is equally vague, although the market seems to be coalescing around the concept of information sharing and personal interaction, mainly for the purpose of some form of business collaboration. Lowe’s Home Centers, for example, uses IBM Connections, a Lotus Notes social business product, to enable its floor staff to quickly locate expert resources across its stores in response to customer inquiries. Lowe’s is a mainframe shop, but like many large enterprises, it supports a number of platforms. It deployed IBM Connections on AIX.

As it turns out, IBM offers several social business tools from its Lotus group. Dubbed the Lotus collaboration suite, the tools can be run on the System z or other IBM platforms. Find details on the Lotus social business tools for z here.  Of course, running the social business tools on the z instead of a distributed platform brings the advantages of the mainframe’s scalability, availability, manageability, and security.

The Lotus collaboration suite includes:

  • Lotus Domino and Lotus Notes—the original server and client components as an enterprise-class platform for critical business, collaboration, mail, and messaging applications
  • Lotus Quickr—consists of  team collaboration software for accessing and interacting with the people, information, and project materials and provides team spaces, content libraries, team discussion forums, wikis, and connectors to simplify the sharing and management of documents and information among a team
  • IBM Connections—social software for business by enabling the development and management of networks of resources and expertise
  • Lotus Sametime—the Notes platform for unified real-time communication and collaboration

For organizations looking to run these tools on the System z or zEnterprise the workloads could qualify for the System z Solution Edition discounted pricing with either the System z Solution Editions for Enterprise Linux or the IBM Enterprise Linux Server, both of which provide discounted hardware, software, middleware, and maintenance for Linux deployments on the z platform.

With any social business tools, the question of ROI becomes tricky. It is difficult to identify quantifiable measurements for collaboration or information sharing. Managers will have to look at business processes like new product development or customer service and even then the one-to-one correlation may not be there. For instance, if a Lowe’s floor person can get the right answer to a customer’s question on the spot while the customer is still there in the store, it might save a sale. Or, Lowe’s might have gotten the sale anyway. Go figure.

With the advent of the zEnterprise and hybrid computing, the System z social business story gets more interesting. It should be possible to deploy pieces of the IBM collaboration suite on different platforms and manage them as a single virtual platform under the z.

IBM Power Systems Winning Streak

November 22, 2011

With all the attention the zEnterprise has gotten in the past year, IBM Power Systems almost seem like an afterthought, at least in terms of the ink it generates. Back in August, however, IDC noted a rebound in the high end UNIX server market of which Power is a big part. Fueling the rebound apparently is the gradual easing of the recession, which has led organizations to begin revamping their systems, and the Power Systems group has been scoring wins all along.

Whether economic recovery is the driving reason or IBM simply is reaping the benefits of the 2011 refresh of the entire Power Systems product lineup, the Power Systems group experienced a 15% year-over-year share increase and expanding profit margins. It also has cashed in on the mistakes of HP and Oracle to the tune of over 250 competitive displacements, resulting in over $225M of business equally split between former HP and Oracle-Sun customers. In you are interested, click here for IBM’s 3Q11 details.

But the financials, good as they are, aren’t the most interesting Power story. A pair of presentations earlier this year by IBMers Steve Will and Patrick O’Rourke laid out some of the goodies the refreshed platform is delivering. For example, PowerVM can drive over 90% virtualization while VMControl delivers industrial strength automation required to take full advantage of that level of virtualization. At the same time, IBM’s EnergyScale technology can reduce Power Systems energy consumption up to 90%. Meanwhile, IBM also was upgrading to AIX 7, which the company dubbed the future of UNIX. (DancingDinosaur sees Linux as the future of UNIX, but that another post.)

The latest rev of the POWER7 processor offers 4, 6 or 8 cores per socket and up to four threads per core. With up to 4.25 GHz processor speed and an integrated eDRAM L3 cache these systems can fly. And the next rev of the processor already is on the roadmap.

Of course, the Power Systems poster child is Watson, the system built around a set of Power 750 servers that won the Jeopardy challenge. The IBM POWER7 processor is optimized to meet the demands of natural language processing, which is what Watson is all about. The processor can handle thousands of analytical tasks at once combined with massive parallelism in which multiple complex tasks execute simultaneously on individual processor threads. Specifically, Watson used multiple IBM Power 750 servers clustered together, each with four processor sockets containing eight POWER7 cores per socket and four threads per core. And you know the result: Watson won by a mile.

Although the number of Power Systems 750 servers, processor sockets, cores per socket, and threads per core used in Watson may not be your typical Power Systems configuration, IBM insists that the processor was not designed specifically for Watson but can handle a wide range of analytical tasks. IBM already has targeted healthcare, financial services, and call centers as primary use cases for Watson-like capabilities.

The interesting thing about the Power Systems win streak is that it does not include natural language processing workloads. The new wins look more like traditional Power Systems workloads than like Watson. For example, the University of Texas at Austin is attempting to predict river behavior in real time. The system combines river systems data with weather and sensor data to predict a river’s behavior more than 100x the normal speed. The combination of analytics and weather simulation on a Power 7 and generate 100  miles of river simulation in an hour, fast enough for people to get out of the way.

Staples, the office supply superstore, turned to Power Systems running IBM WebSphere Commerce 7 to optimize its website for high volume transactions. Staples saw page performance improved anywhere from 25- 55% with IBM POWER7 Systems, according to the company.

Power Systems can process an enormous number of concurrent transactions and data while analyzing information in real time. With Black Friday and Cyber Monday, two particularly intense retail shopping days almost upon us, Staples will want all the performance boost they get from the Power platform.

With the introduction of the zEnterprise and hybrid computing and with the recent announcement of x86 blades for the zBX to fill out the multiple zEnterprise hybrid computing platforms, it is easy to forget that Power blades and AIX also can play in this game.

New Workloads for the zEnterprise

July 18, 2011

Even before the introduction of the z196 a year ago, IBM had been steadily promoting new mainframe workloads. With the introduction of the zEnterprise, consisting of a z196 and an attached zBX, the hybrid mainframe became real and with it the possibility of running and managing truly new workloads through the z.

Obvious new workloads would be AIX workloads previously running on Power Systems servers, but these aren’t truly new to the organization, only new to the z. The adoption of the Smart Analytics Optimizer, as Florida Hospital plans to do for medical research analytics, is a truly new application for the hospital and for the z.

The introduction of the z114 opens up the potential for new workloads on the z. This would be due mainly to its lower cost, entry pricing starts at $75,000. This lowers the risk of testing new workloads on the z. For example, would an organization now be more willing to try BI against production data residing on the z as a new workload if they could get a discounted price? They could, of course, run BI on a slew of Intel servers for less, but they would give up the proximity of their data and the potential for near real-time BI.

As recently as this past March Marie Wieck, General Manager of IBM Application and Integration Middleware, made the new workloads case in a presentation titled New Workload and New Strategic Thinking for z. In that presentation she identifies five categories of new workloads she deemed strategic. They are:

  1. Business Intelligence (BI) and Analytics
  2. Virtualization and Optimization
  3. Risk Management and Compliance
  4. Business Process Management (BPM)
  5. Cloud computing

None of these, with the possible exception of Cloud, are new workloads. Organizations have been running these workloads for years, just on other platforms. But yes, they certainly are not the traditional System z workloads, which typically revolve around CICS transaction processing, OLTP, and production database management.

DancingDinosaur would like to suggest some areas of other new workloads for the z, especially if you can grab a deeply discounted z114 cheap through the Solution Edition program. And since they are new workloads, they should automatically qualify for the discount program.

The first would be Linux development and testing using a deeply discounted enterprise Linux Solution Edition for the z114. Developers could put up and take down servers at will, runs gads of test data through them, and the machine wouldn’t break a sweat.

Another should be SOA. Enterprise Linux combined with CICS access to production data should be a ripe area for new services-oriented, web-based workloads. You could even pull in smartphones and tablets as access devices.

Finally, there should be much that organizations could do in terms of new workloads using Java, WebSphere, SAP, and even Lotus on the z114. Here too, there will likely need to be Solution Edition discount programs available to reduce costs even more.

And then there are the x blades for the zBX and the imminent arrival of Windows on those x blades. That has the potential to open maybe the largest set yet of new workloads for the z.

The big obstacle to new workloads on the z is that these workloads already are running in some form on other platforms in the organization. So, what the z gains, the other platforms and the teams that support them lose. That’s a difficult political battle to fight, and the best way to win is to offer an unbeatable z business case. Even with the z114, IBM isn’t there yet.

To get there, IBM has to add the last piece missing from the new workloads picture painted above—a System z Solution Edition discount program that also includes a deeply discounted zBX. That could prove irresistible to organizations otherwise contemplating new system z workloads.

New IBM z114 Reduces Mainframe TCA

July 12, 2011

IBM introduced its newest mainframe in the zEnterprise family, the z114, the zEnterprise equivalent of a z10 BC. With the z114 IBM can now deliver a more compelling total cost of acquisition (TCA) case. The z114 comes with a $75,000 entry price, 25% less than the z10 BC entry price while delivering 18% more performance per core. At this price, it can begin to compete with commodity high end servers on a TCA basis, especially if it is bundled with discount programs like IBM’s System z Solution Editions and unpublicized deals from IBM Global Finance.

First, the specs, speeds, and feeds:  the z114 is available in two models; a single-drawer model, the M05, and a two-drawer model, the M10, which offers additional capacity for I/O and coupling expansion and/or more specialty engines. It comes with up to 10 configurable cores, which can be designated as general purpose or specialty engine (zIIP, zAAP, IFL, ICF) or used as spares. The M10 also allows two dedicated spares as well, a first for a low end mainframe.

Like the z196, the z114 uses a superscalar design that runs at 3.8 GHz, an improved cache structure, a new out-of-order execution sequence, and over 100 new hardware instructions that deliver better per-thread performance, especially for DB2, WebSphere and Linux workloads. For CPU intensive workloads, additional gains of up to 25% can be achieved via multiple compiler level improvements. None of this, however, is technically new to the zEnterprise. The base z114 starts at 26 MIPS but can scale to over 3100 MIPS across five central processors plus the additional capacity provided by its specialty engines.

The z114 mainly will be a consolidation play. IBM calculates that workloads from as many as 300 competitive servers can be consolidated onto a single z114. It will become the key offering as IBM pursues competitive wins against HP and Oracle/Sun, ripe targets these days. IBM figures the z114 can consolidate workloads from 40 Oracle server cores using just three processors running Linux. And compared to the Oracle servers IBM estimates the new z114 will cost 80% less. Similarly, IBM figures that a fully configured z114 running Linux on z can create and maintain a Linux virtual server for approximately $500 per year.

As a consolidation play, the zEnterprise System will get even more interesting later this year when x blades supporting Windows become available. Depending on the pricing, the z114 could become a Windows consolidation play too.

As part of the zEnterprise family, the z114 connects to the zBX where it can manage workloads running on POWER7-based blades as well as theIBM Smart Analytics Optimizer and IBM WebSphere DataPower for integrating web-based workloads. The zBX can handle up to 112 blades, which can be mixed and matched within the same zBX chassis. The maximum number of blades varies depending on the type of blades selected: 112 Power blades but only 28  System x HX5 or DataPower blades or 56 Smart Analytics Optimizer blades. In additionIBM also promises support for Microsoft Windows on select System x server blades soon.

To drive competitive TCA, IBM clearly is ready to make deals. For example, IBM also has lowered software costs to deliver the same capacity for 5%-18% less through a revised Advanced Workload License Charges (AWLC) pricing schedule. A new processor value unit (PVU) rating on IFLs can lower Linux costs as much as 48%. Some prices, however, may bump up when new pricing takes effect in January 2012.

The best deal is the System z Solution Edition Program which DancingDinosaur has written about here and here.  It bundles System z hardware, software, middleware, and three years of maintenance into a deeply discounted package price. Initial System Editions for the z114 will be WebSphere, Linux, and probably SAP. You also can expect what IBM refers to as snap-ons that provide discounted optional service components to the existing base system. Expect these initially for the IBM Enterprise Linux Server. Cloud and Smarter Planet offerings also may include snap-ons. Not all Solution Editions will be available for the z114; there also may be some new ones.

IBM Global Finance (IGF) can lower costs, starting with a six month payment deferral. You can acquire a z114 now but not begin paying for it until the next year. The group also is offering all IBM middleware products, mainly WebSphere Application Server and Tivoli, interest free (0%) for twelve months. Finally, IGF can lower TCA through leasing. Leasing could further reduce the cost of the z114 by up to 3.5% over three years.

In the end the z114 doesn’t bring any major technology breakthroughs to the zEnterprise that weren’t previously seen in the z196, except maybe the PCIe I/O subsystem. What it brings is a new pricing structure for lower TCA and better price/performance. Yes, the capabilities have been ratcheted down, but the growth path to the full zEnterprise remains if you need it.


Follow

Get every new post delivered to your Inbox.

Join 672 other followers

%d bloggers like this: