IoT
IoT
Data Management // Big Data Analytics
Commentary
1/15/2015
08:36 AM
Doug Henschen
Doug Henschen
Commentary
Connect Directly
Google+
LinkedIn
Twitter
RSS
100%
0%

IBM Mainframe Makeover: Mobile, Big Data Reality Check

IBM z13 refresh packs performance gains, but can it really join the mobile, big data, and cloud age? Here's a look at aspirations versus realities.

8 Quiet Firsts In Tech In 2014
8 Quiet Firsts In Tech In 2014
(Click image for larger view and slideshow.)

Yes, many of the largest organizations -- big banks, big insurance companies, big retailers, and big government agencies -- still run IBM mainframes. So when IBM announces a new generation of the platform, as it did on Tuesday (on its usual two-year cycle) with the release of the z13, hardware unit financial results always perk up -- which is news IBM could use these days.

Advances in chips and memory always bring new heights of performance, and it's no different with the z13. IBM says the new machine can process 2.5 billion transactions per day -- "the equivalent of 100 Cyber Mondays every day of the year." But IBM is also doing its best to give the z13 a very modern spin, touting mobile, cloud, and big data analytics performance.

How real are these claims? Here's a closer look at realities versus aspirations.

[Want more on IBM's hardware business? Read IBM Sells Chip Business, Declines Continue.]

Hardware constantly evolves, but 50-year-old mainframes have required periodic makeovers. Fifteen years ago, for example, IBM took advantage of virtualization to enable the z series to run Linux- and Java-based applications. That has opened the door to massive workload consolidation, and mainframes now run all types of Linux-based apps, such as mission-critical SAP apps and Oracle databases, while still running Unix apps; iSeries (formerly AS/400) apps; and classic CICS (Customer Information Control System), IMS, and DB2 workloads on the z/OS mainframe operating system.

That consolidation story still has strong appeal, and it helps the mainframe to win 80 to 100 new customers per year, according to IBM. (IBM could offer no details on how many customers get off mainframes each year, but a spokesperson says it's a short list of smaller companies.) Some new customers run Linux exclusively, as is the case with Brazilian credit union Sicoob, which first started using a z-series machine four years ago and is now among IBM's top-ten mainframe customers.

Mainframe as mobile platform?
IBM's z13 announcement put a big emphasis on meeting modern mobile demands, but surely it's not suggesting that companies should build mobile apps to run on mainframe, is it? Well, yes and no. Smartphones and tablets are in no way different from PCs, in that they're access devices calling on backend systems that run on mainframes. But because they're so handy and accessible, these new devices are generating a tsunami of requests for information. People now check account balances 12 times a month where they might have done so twice a month on a PC.

Mobile traffic is maxing out backend system calls, so z13 performance gains are needed. As for mobile apps running on mainframe? It's early days.
Mobile traffic is maxing out backend system calls, so z13 performance gains are needed. As for mobile apps running on mainframe? It's early days.

The z13 better addresses mobile workloads simply by offering more processing power. Another answer here is z/OS Connect, a connector technology that supports the delivery of information from CISC, IMS, and DB2 in a RESTful manner with simple API calls. Billed as a linchpin of Web, cloud, and mobile enablement for mainframe, z/OS Connect was introduced last year. It is compatible with z12 servers.

IBM also hopes that mobile apps might be developed to run on the mainframe. Here, IBM has ported its Mobile First (formerly Worklight) mobile app development portfolio to run on zSeries, and IBM says there are also plans to support IBM's Cloudant NoSQL database service and third-party products like MongoDB on mainframe. NoSQL platforms, often running in the cloud, are where the lion's share of mobile development and application delivery is happening these days. It remains to be seen whether third-party vendors and customers will bring this work onto the IBM mainframe in IBM's cloud.

Analytics and big data on mainframe?
In 2011 IBM introduced the IBM DB2 Analytics Accelerator (IDAA), a coprocessor that essentially puts a Netezza-based data warehouse alongside mainframes handling transactional workloads. The benefit is that you can do analysis of the transactional data on the mainframe without moving that information or degrading the performance of the transactional system.

With z13, IBM is putting more of its software onto the IDAA, including SPSS for analytics and Cognos BI for reporting. This opens up support for real-time fraud detection, real-time upselling and cross-selling, and other opportunities driven by predictive analytics. One major insurance company using this capability is shifting to doing fraud-detection analysis against 100% of the claims it processes on the mainframe. Previously the company used red-flag rules and transaction thresholds to kick out roughly 10% of its transactions into a separate analytical fraud-detection app, but now it can analyze every claim in real time.

The benefits here are twofold: Claims that turn out to be legitimate won't be delayed, risking penalties. And fraudulent claims that didn't set rules or value thresholds won't fall through the cracks or require follow-up investigations. In another real-time use case, IBM says a chain of Eastern European gas stations and convenience stores is using real-time analysis for upselling at checkout time. When customers fill up their gas tanks, real-time checks of the loyalty program trigger discount offers that have increased total sales.

IBM says z13 can easily support distributed systems that have become synonymous with big data analytics, but here, too, IBM's ambition will confront market realities. IBM has ported its own BigInsights Hadoop distribution to run on the mainframe, for example, but there hasn't been a stampede of third-party players in the big data community following suit. Veristorm, for one, has a Hadoop distribution ported to run on System z, but most distributed big data platforms -- Hadoop and NoSQL databases -- were designed specifically to run on commodity x86 servers. The whole idea is cheap compute. IBM says that with virtualization, one System z chip can handle the equivalent of 15, 20, or even 25 x86 chips. But if the software was designed to run on x86 and nobody has bothered to port it, it's a theoretical argument.

Cloud on mainframe?
The list of software designed to run on System z and IBM Power servers is long, but what's compatible with x86 is much longer. This is a big reason IBM bought (x86-centric) Softlayer -- to be at the center of its cloud story. IBM's cloud didn't catch fire until Softlayer entered the picture. IBM can talk about support for Linux, KVM virtualization, and OpenStack, and ambitions to see all sorts of third-party software running on mainframe. But the software that can run System z today and what IBM hopes for are two different things.

Where big companies already have big-iron workloads, the System z has already proven its appeal in workload consolidation, private cloud, and hybrid cloud scenarios. In fact, I suspect the success of System z has had a lot to do with the hardships of the IBM Power line (caught in a closing vise between System z and x86). But where the core of cloud deployment and big data activity is concerned, it will be an uphill climb for IBM to persuade those communities to run on System z, no matter how trivial porting software might be.

Employers see a talent shortage. Job hunters see a broken hiring process. In the rush to complete projects, the industry risks rushing to an IT talent failure. Get the Talent Shortage Debate issue of InformationWeek today.

Doug Henschen is Executive Editor of InformationWeek, where he covers the intersection of enterprise applications with information management, business intelligence, big data and analytics. He previously served as editor in chief of Intelligent Enterprise, editor in chief of ... View Full Bio
Comment  | 
Print  | 
More Insights
Comments
Newest First  |  Oldest First  |  Threaded View
TerryB
50%
50%
TerryB,
User Rank: Ninja
1/21/2015 | 3:04:27 PM
Re: Not x86 virtualization
I did not know you could run i5/os (what you are calling iSeries) on a Z system partition. But if mainframe is POWER, I guess that makes sense for consolidation reasons. A top end i5 server now (which shares same hardware as AIX os) is pretty much a mainframe now with the scale it offers. I have been working on AS400/iSeries/i5 servers since 1988 when they hit market to replace the old System 38. The growth of these servers has been amazing over the years.

My current job started with helping them implement ERP on a AS400 back in 1998 to replace LoB software they ran on a Wang (yes a Wang, only time I actually saw one in my career) and on Windows based Paradox software. That first machine in 1997 was 1000 lbs and cost about $100,000. I don't remember exact specs but memory was maybe a few MB, storage was maybe a GB and the CPW (performance measure) was, say, 100. The one we bought in 2003 was $50K, 200 lbs with everything else about doubled. The one we use now, bought 2009, is smallest model they have. 4U rack mounted for about $20K, 4GB memory, 460GB storage, CPW about 4300.

The new one I got quote on is less than $20K, 32GB memory, 849GB storage, CPW 9875, again smallest they make. And people wonder why IBM hardware revenue is down. This is still finest business server in available. And by business, I don't mean Facebook. Think actually making a product and selling it.

This cloud/scale internet business stuff can use cheap x86, whole different ballgame. But as I commented on another article in Dark Reading, still waiting for somebody to tell me story about an IBM mainframe getting pawned that wasn't inside job. My i5 is not going to taken by ransomware, ever.
D. Henschen
50%
50%
D. Henschen,
User Rank: Author
1/15/2015 | 3:25:31 PM
Not x86 virtualization
It wasn't x86 virtualization that was introduced 15 years ago -- it was RISC (IBM Power) emulation for migration of Unix apps to Linux and iSeries emulation for formerly AS/400 apps. To my knowlege, mainframe can't run x86 software, despite IBM's contant comparisons of mainframe compute-per-processor power to x86. The latter will always lose by that measure, but distributed systems run on hundreds if not thousands of cheap, easily replaceable x86 boxes.  
Charlie Babcock
50%
50%
Charlie Babcock,
User Rank: Author
1/15/2015 | 3:08:17 PM
Mainframe still relevant...
The mainframe is basically a set of concepts that remain in play today. Combine lots of processing power and memory with short data connection channels between them so that large jobs can be undertaken. The cloud tries to do much the same thing, only it wants to use the cheapest commodity parts, and get those parts to scale out infinitely.Until the cloud came along, no one could do it at scale quite as well as the mainframe. That's why it remains entrenched in legacy systems. "Fifteen years ago, for example, IBM took advantage of virtualization to enable the z series to run Linux- and Java-based applications." Well, yes, it took advantage of x86 virtualization. But IBM invented virtualization 50 years ago with its VMS operating system, which could run guests underneath it.
David Wagner
50%
50%
David Wagner,
User Rank: Strategist
1/15/2015 | 3:02:48 PM
Re: Cost?
Thanks, Doug. You'd think the "C" part owes them considering how much of their business they have already sold to C.
D. Henschen
50%
50%
D. Henschen,
User Rank: Author
1/15/2015 | 2:57:28 PM
Re: Cost?
In the case of Sicoob, for example, it's a credit union. IBM has tons of ready-to-customize systems for banks, insurance companies, retailers, government agencies, etc. These apps were born on the mainframe. IBM was making big inroads in BRIC countries with these apps until China put a crimp on the "C" part of that growth story.
David Wagner
50%
50%
David Wagner,
User Rank: Strategist
1/15/2015 | 2:36:07 PM
Re: Cost?
@Doug- I think that makes total sense. Facebook and the other big companies you mention certainly have the cash to switch if they want, and they're not. 

Is there any indication what is drawing the new customers then?
D. Henschen
50%
50%
D. Henschen,
User Rank: Author
1/15/2015 | 2:30:15 PM
Re: Cost?
IBM doesn't divulge pricing. It talks about it in terms of MIPS (millions of instructions per second). You're right about consolidation of massive workloads being the best economic use of mainframe processing power, but MIPS at scale aren't exactly cheap. Oracle claims IBM has been jacking up rates. Customer Sears has made a business out of its Metascale big-data spinoff, which specializes helping companies move workloads like data transformation OFF of mainframes and into lower-cost, commodity Hadoop clusters, where companies can harness lower-cost processing cycles on massive clusters running on commodity x86 servers.

The bottom line is this: if the likes of Google, Facebook, Yahoo and Amazon could run more cheaply on IBM Mainframes, that's what they would use, but they don't. Those guys invented big data platforms on commodity hardware specifically so they could scale affordably. And I don't think they would migrate to mainframes if they could start from scratch assuming their current scale.
David Wagner
50%
50%
David Wagner,
User Rank: Strategist
1/15/2015 | 2:14:09 PM
Cost?
Hi Doug. Interesting article. Unless I missed it, the one thing I missed was the associated cost with mainframe versus non-mainframe options. It seems possible that for big enterprises mainframe solutions might be cheaper. I always feel like the problem with the mainframe is that it is always like paying for everything with a 50 dollar bill and no one has change. Mainframes seem cheaper and better as long as you are fully using it. Is that perception still correct?
6 Tools to Protect Big Data
6 Tools to Protect Big Data
Most IT teams have their conventional databases covered in terms of security and business continuity. But as we enter the era of big data, Hadoop, and NoSQL, protection schemes need to evolve. In fact, big data could drive the next big security strategy shift.
Register for InformationWeek Newsletters
White Papers
Current Issue
Top IT Trends to Watch in Financial Services
IT pros at banks, investment houses, insurance companies, and other financial services organizations are focused on a range of issues, from peer-to-peer lending to cybersecurity to performance, agility, and compliance. It all matters.
Video
Slideshows
Twitter Feed
InformationWeek Radio
Sponsored Live Streaming Video
Everything You've Been Told About Mobility Is Wrong
Attend this video symposium with Sean Wisdom, Global Director of Mobility Solutions, and learn about how you can harness powerful new products to mobilize your business potential.