The drive to combine the mainframe computer and the cloud – at first sight perhaps seen as an attempt to join chalk with cheese’ – continues to gather pace, and is attracting new players that are coming up with advantageous ways to exploit the mix.
One of those advantages comes in the form of the vast wealth of historical business data that the mainframe – as a collective entity – holds. One of the disadvantages is the currently non-trivial task of business users trying to access that data in ways that make it immediately available to important tools, like analytics, AI systems and machine learning tools.
That, in particular, is the primary target for Israeli-based Model9 and its founder and CEO Gil Peleg. It is perhaps worthwhile spending a second or two setting out his background, because once we start talking about having extensive experience of running and managing IBM mainframe systems, extensive experience of not just COBOL applications, but also much racier tasks such as real-time, mission-critical logistics management systems as well as being the co-author of multiple IBM Red Books on z/OS implementation, it is pardonable to assume that he must be at least as old as this writer. [Editor’s note - shome mishtake shurely?]
That, however, is not the case, as he told me:
I started with mainframes when I was 18. right out of high school. As you may know, in Israel, when you finish high school, you have to do a mandatory army service, and I was picked to serve in the army computer center.
The interesting thing here is that the Israeli military remains largely based on IBM mainframes. In fact, as Peleg points out, it is one of the largest mainframe sh0ps in the world. So, despite being well-versed in Linux, Windows and all things PC he, as he put, accepted his destiny:
I started learning mainframes, and I actually fell in love with that platform. I'm 40 this year, and I have more than 20 years experience with mainframes. My reserve duty is actually teaching mainframe computing to younger people and helping create the next generation of mainframe professionals.
So, Peleg’s experience goes much further than the typical mainframe user working in a bank, for example. The reliance of the Israeli military on mainframes meant that the management and deployment of just about any resource, in any type of military action iss a job for a mainframe system, and often a real time, mission-critical job.
Model9 was born out of that experience, set up with that goal of bringing the mainframe and cloud closer together. To do that Peleg brought together a team that has both deep mainframe expertise and the capability to move at the pace of the cloud. Not surprisingly most of them come from the same background and the same army unit. The original investment came from Intel Capital and the firm now has two locations - engineering is in Tel Aviv, while the standards and customer management organization in the USA.
Standing at the junction
Peleg positions Model9 at the junction of two very strong market forces. The first, of course, is mainframes, which are still at the core of the largest companies in the world and provide the backbone of the data centre platform; the other is the cloud, where the innovation, the analytics, the economics and everything that's ‘happening’ in IT can be found.
At present these two have not met well, if at all, and if anything are at risk of being increasingly at loggerheads, he suggests:
The hyper-scalers of the world are super interested in the large enterprise data centers, but the enterprise has this mainframe weight tied to the legs, keeping it on-premises. What the cloud players are offering the enterprises are long-term, expensive, risky migration projects, at the end of which they're going to be in the cloud. But that's something that's very hard to do and it takes many years. We speak to customers who are on the seventh year of a five-year project.
In practice, he sees this as enterprises betting on their most valuable mission critical systems by attempting to migrate them to entirely new applications on entirely different platforms, when what is required, he suggests, is a different approach - data-led cloud migration. The objective here is to make mainframe data available on the cloud and make it transparent to those applications that can exploit it effectively.
When Peleg uses the word cloud he is not talking about the public cloud but rather the private cloud services most large enterprises run. This is combined with a twist on the classic SaaS model, whereby a user company runs its data on the SaaS firm’s servers to be close and available to an instance of the application. Model9 is following the more modern trend of moving the compute to where the data resides, in this case installing an instance of its SaaS application onto an enterprise customer’s own private cloud.
The other difference is that it does not then work with a duplicate of the mainframe data, but the live data itself. The mainframe is cut back to being a compute system with enough storage to allow the application code to run properly, he explains:
Most of the mainframe data is actually stored on tape systems and we can replace the tape systems, including all the operations and the application operations on tape, directly with cloud storage. The cloud becomes your primary data storage for your mainframe, and that all happens transparently to your application.
This bypasses the common next step which is attempting to migrate mainframe application functionality to a cloud-native application, which is both risky and rarely accomplished with great success. It is a task not helped by the fact that mainframe applications developers are not now in plentiful supply, while developers out in the cloud environment simply don’t have the right skills to work with mainframes. This, in Peleg’s view, is why the trend towards grafting cloud-oriented dev/ops technologies and skills onto traditional mainframe development methods have emerged as a way of sidestepping the lack of suitable talent.
Real mainframe data...in the cloud
Model9’s pitch is that it builds an environment where the cloud is not a duplicate source of data but the source for the mainframe, a capability which allows applications to store more than available to them on the mainframe itself. As Peleg observes, it is now quite common for mainframe applications to write data direct to tape in order to operate effectively and avoid mainframe capacity limitations.
In addition, it also creates an environment where Model9 can also make the data directly available to a range of increasingly critical business applications, such as AI and Machine Learning initiatives and, of course, analytics tools, he adds:
So we're trying to tackle this from a different paradigm of making the data accessible to the tools that are already out there, instead of trying to port many of the tools to the mainframe ecosystem. We're saying it's going to be a hybrid world. The basis of application development these days is data, so how do we make data accessible to everyone, including through mainframe applications relying on sentence structures and reading in a certain way that’s very hard to change. That's how we do it - we make this data accessible without making ‘Big Bang’-type changes on the mainframe side by tackling the storage.
The solution seeks to provide a flexible system that shows users what types of data are on available domains and what types of files, volumes, databases, etc are in each domain, and what do they each contain? This information is delivered to the cloud developer or data analyst so they can pick and choose whatever they need. They then run their selection through the data transformation services, effectively an API layer, which makes mainframe data, in mainframe format, available to cloud applications via the same API's that they use to consume data, effectively making them like an XML file or a database.
As AI systems start to move center stage, the value of data starts to grow, especially when it comes to historical data from which much longer-cycle trends data can be inferred. So the more that tools can gain access to such data, the more value is derived and made available for exploitation. Peleg is a bit reluctant to name names when it comes to customers, but when asked for some measure of this capability in action, he cites a transportation company in the US, which is now able to use some 100x more data in its analytics:
It's something like they moved from 20 database tables to 2,000 database tables that participate in their analytics initiative now. When the data is on tape it remains untouched. It's simply not accessible to your analytics and AI Solutions.
This is a rather striking example of the importance of data and its pointlessness if it is impossible to access. Mainframe systems are now of huge importance to the enterprises that use them – and particularly have used them for years. The value is no longer in their ability to provide day-to-day back office services, though they are still very important. Instead, it is about the wealth of historical data they hold and what can be learned from it if that data can be accessed by the right tools.
Those tools exist, but are only widely found in the cloud, so putting mainframe data in the cloud alongside them is obviously sensible. The problem then is that the data exists in formats and languages that cannot be readily accessed by modern cloud applications. For example, the concept of a data ‘file’ is totally alien in the mainframe world. Bridge that gap and a whole new world of data opens up to AI, Machine Learning and analytical tools. That is the goal of Model9, and should it succeed, there is even the possibility that it could open up use of the mainframe - especially now IBM is producing new, small mainframe systems - to a whole new market for that venerable technology.