Why Citi Is Buying IBM's New Mainframe for Mobile Transactions

In an era where businesses are increasingly focused on cloud computing, Citigroup is turning instead to souped-up hardware in the form of IBM's newest z13 mainframe, which it says can help it better deal with the explosive growth in mobile and online transactions.

The mainframe, which launched Wednesday, comes with an impressive list of capabilities. It can process 2.5 billion transactions a day, encrypt mobile and online banking transactions, and allow those transactions to be monitored with real-time analytics, allowing users to spot potential fraud and find opportunities within transactions as they're happening.

The new mainframe also comes with an application programming interface that lets developers create apps for the system quickly — an important selling point for banks that need to add new features to their apps frequently for competitive and business reasons.

Citi was one of the mainframe's first purchasers, and believes it can help further its focus on high performance and security of mobile transactions.

"It's causing us to see a different blend of transaction mix and volume, and it's causing us to look at how we deliver that elegant, seamless experience to the customer in a very scalable and secure way," said Martin Kennedy, managing director for global enterprise platforms and storage at Citi, in an interview Tuesday. "This new machine and some of the new enhancements of this machine build on the hallmark characteristics of the mainframe being a secure and scalable platform."

The z13 will let Citi "scale in place," he said, rather than having to expand the footprint of the bank's data centers to accommodate growing mobile transaction volume.

It will also let the bank explore the notion of near-real-time analytics on data.

"We'll be able to do it in place as opposed to an old platform analytics engine, which is something that's attractive to us because the less moving parts in any part of the system, the more available it's going to be and the less complicated it's going to be," he said.

The machine will provide enhanced support for Java (a language often used in writing mobile applications) and WebSphere (IBM's web application server) workloads, Kennedy said.

"That enables us to improve throughputs and our response time and keep focusing on the customer experience as well as enable security and fraud detection," Kennedy said.

Kennedy says the bank is also exploring taking advantage of the z13's crypto coprocessors to encrypt every mobile transaction.

"We do a fair amount of encryption today with different endpoints," he said. "We'll continue to look at how we can leverage this technology to further encrypt data and the techniques we use to encrypt it. It's a process that typically requires a certain amount of processor overhead, volumes and throughput restrictions, and this machine works through all those limitations."

Citi is also interested in seeing how else it can use the mainframe's capabilities.

"We, like everyone else want to have the best piece of technology to host a particular function," he said. "They're adding more capabilities to this processor, and in particular it allows us to explore the notion of bringing a lot of this processing closer to the data."

Traditionally, to use analytics software it's necessary to feed data to the analytics engine.

"What IBM seems to be doing here is bringing the analytics right to the data, you don't have to worry about getting data off platforms," he said.

Fraud detection and customer analytics can be done in place, on the fly.

"Why move data around and make extra copies of data when you can do it in place?" he said. "And perhaps you can do it close to real time. The closer you can get to the data, the more you can analyze it in real time, that gives everybody a leg up on fraud and other sorts of things."

Is the Mainframe a Relic?
Of course, the tech field continues to debate whether the mainframe is going away in favor of cloud computing, with processing distributed among farms of standard component servers. That has already happened in some corners of the business world, including banking. But for core, mission-critical work, bankers and their vendors tend to prefer the solidity and security of big iron.

Mike Desens, vice president, systems at IBM, said he understands the debate.

"In many ways the mainframe is not well known by the general public," he said. "Yet we all know it's the heart and soul of businesses all over the world."

Kennedy said the mainframe has a long-term future.

"All banks host the core banking apps on a mainframe," he said. "When you need a fit-for-purpose analytics engine, the mainframe will have a place in every financial institution's hosting architecture."

About 96 of the top 100 banks in the world still use mainframes. The vast effort that would be required to shift massive volumes of work to new, smaller servers means that mainframes will stick around for a while.

"The mainframe is still a ubiquitous element in most large corporations," said Richard Fichera, vice president and principal analyst at Forrester Research.

The Focus on Mobile
IBM's focus on mobile and analytics in the new system was based on processing trends it was seeing in existing mainframes.

"We're seeing unprecedented volumes of mobile transactions," Desens said. Not only do consumers conduct banking transactions more frequently on mobile devices than on any other channel, mobile apps tend to generate additional, sometimes unnecessary streams of data, he said.

"Banks and other businesses that are mobilizing enterprises are developing rich apps, because this isn't just a new way of interacting with clients, it's a new way of doing business and trying to grow their businesses," Desens said. "With some mobile banking apps, every time you touch your smart phone, that sends a flood of transactions back into the systems of record, just to update information like account balance, mortgage balance and line of credit balances. If you touch your phone once and then five minutes later tap it again, they don't need to update your mortgage balance for you — it hasn't changed in five minutes. Yet they're aspiring to provide one view of the customer in an efficient and effective manner."

To enhance the mainframe's ability to handle mobile transactions, IBM has introduced software tools and runtime assists.

For instance, a new application programming interface called Zos Connect is designed to let application development teams easily access services running on the mainframe.

"The speed of mobile is dictating this," Desens said. "No longer can you put an application development project in place and a year later you're done. These things are changing on a monthly basis. Customers want to deploy new apps fast, see how customers react, and then make modifications."

On the analytics side, IBM is peddling its DB2 analytics accelerator, which is a couple of years old, to speed up the processing of complex queries. It's recommending its SPSS software for modeling and scoring and Cognos for reporting.

"We have the lifecycle of data so that now you can do in-transaction analytics," Desens said. For banks, fraud detection is the most logical first application of this, followed by customer analytics (e.g. sending a relevant product offer to the customer during a transaction).

Insurance providers could use this, too. Some insurance companies can only do fraud detection for 10% of claims because of the demands of their service level agreements, Desens said. The other 90% of claims flow through and fraud detection is done later.

"Then they have to chase down and try to get reimbursed for fraudulent claims, which you can imagine is not very effective," Desens. "Now with SPSS scoring and modeling capabilities on the mainframe, they can score and detect fraud on 100% of those transactions in real time."

The z13 itself also supports something called single instruction multiple data (SIMD), which lets one microinstruction operate at the same time on multiple data items. "When you start doing advanced modeling and analytics, you can get an 80% performance improvement in z13 because of SIMD," Desens said.

On the security side, Desens said IBM doubled encryption performance for the z13. "We have unique logic down to the microprocessor core that helps the advanced encryption algorithms work fast," Desens said.

Fichera sees the z13 as an "iterative improvement" of IBM's current mainframe technology, providing faster processing than earlier models.

"In essence a completely redesigned system, it has improvements in CPU performance, memory capacity and bandwidth at multiple levels," he said. "Current mainframe users can look forward to improvements in all processing metrics.

"This re-positioning of the mainframe seems to me to be a pretty sensible and low-risk strategy for IBM," Fichera said.

Incumbent customers are likely to migrate to the new platform and the company is poised to capture some of the rapidly growing mobile transaction volume.

"For developers of mobile applications that are heavily dependent on mainframe data resources — banking apps come immediately to mind — the z13 and associated software may provide an interesting platform alternative," he said. "Cost models are tricky for these environments, but IBM will be pulling out all the stops to capture this business, and potential customers for this new orientation of the mainframe will find themselves able to drive better bargains than they would for simple capacity upgrades of currently installed systems."