We’re Integrated, We’re A Platform, Let’s Catch The Wave

May 19, 2014 Timothy Prickett Morgan

In my 25 years of watching the AS/400 and its progeny, I have had my share–some might say more than my fair share–of crazy ideas about how to modify, grow, stimulate, and vitalize the market for OS/400 and IBM i systems. As I sit here and contemplate the rising market for integrated systems, I am busting my brain to try to figure out how to leverage this wave on the behalf of all of us. While bare-bones rack and tower servers still dominate shipments worldwide, integrated machines, often with pre-tuned and pre-configured software stacks, are finding their places in the data center.

The integrated nature of the System/38 and AS/400 hardware systems, combined with the built-in relational database in the operating system and the application development tools designed for business logic, not low-level programming, seemed like an obvious thing to pay a slight premium for. The premium was crazy stupid large in the System/38 period from 1980 to 1988, several times more expensive than the IBM System/390 mainframes of the time, but with the advent of the AS/400 in 1988 the cost of relational database processing on the midrange platform fell well below that of the mainframe and has stayed there ever since.

IBM would have done well to have created the AS/400 as an integrated system, but to keep pricing for the software components separate from the hardware and from each other so companies could make apples-to-apples comparisons. Just because the box is integrated does not mean the invoice has to only have a few items on it. I would argue that the AS/400 had an ease of programming advantage as well as an ease of use advantage over other systems–the application programmer, system administrator, and database administrator were often the same person at small AS/400 shops. IBM never did quantify the cost savings from all of this in a concrete fashion. I have never understood why. IBMers talked about this ease of use all the time, but never with concrete numbers.

IBM has always been skittish about suggesting that computers can be used to replace people, whether in the lines of business or in the data center. But this is obviously true, and if you watch the hyperscale data center market as I do, this is precisely what has happened. Just to give you a sense of it: At the Facebook datacenter in North Carolina, there are three shifts of workers, with 80 people in total across those shifts. They have automated the living daylights out of system configuration with a tool called FBAR, which automates the work of hundreds of admins and allows for a single tech to babysit 25,000 servers. (The site has many more servers than this.) Computers create jobs for programmers and architects and all kinds of people, but they also destroy jobs. Period. Get over it. Or, don’t.

The wave of integrated systems coming out of all of the major server makers–IBM PureSystems, Hewlett-Packard AppSystem, Dell Active Systems, Oracle Exa family engineered systems, and others–are all about making machines that are tuned to very specific jobs, but also making it so you don’t need as many handlers for the computers. In some cases, the idea is to use the vendor to patch and maintain the machines and get users away from it. And this probably makes a lot of sense, particularly for very small businesses who are never going to have the skills to manage a complex system anyway.

My point in bringing this up is that the market for integrated platforms is growing, according to the latest data from IDC. For the past several quarters, IDC has been trying to figure out how to best qualify and quantify the market for what it calls integrated infrastructure–packages of servers, storage, and switching–and integrated platforms–which is said machinery with layers of systems and sometimes application software on it to do a very specific job. After messing around with the numbers for a bit, IDC thinks it has cased the market correctly for the integrated platforms part of the market.

According to IDC, Oracle has the lion’s share of this space with its Exadata, Exalogic, Exalytics, and Sparc SuperCluster machines, as you can see in this table:

IBM, as you can see, comes in second place in the integrated platforms market, with a 15 percent market share and $535.3 million in sales in 2013. IBM’s PureSystems business at the platform level–which includes the PureApplication setups as well as PureData (formerly Netezza data warehouse appliances and other preconfigured machines) machines. For some reason, IBM does not call a PureFlex node running IBM i a PureApplication machine, but technically speaking a cluster of one or more of the machines running IBM i is absolutely a PureApplication by any definition that makes sense. The good news is that IBM’s PureApplication business is growing nearly three times as fast as the market overall. There are obviously many players in this market, and many of them are only bringing in millions to tens of millions of dollars per quarter, if IDC’s casing of the market is correct. HP has a tiny slice of this space with its AppSystem, a variant of its BladeSystem blade servers with its Matrix automation environment with its Vertica NoSQL database put on it. The other vendors in this space have nearly half of the market, but again, they are all small and they are, when taken all together, not growing.

Basically, this is a two-horse race, and at the moment, Oracle is winning.

First of all, to get the data correct, IBM should be able to count any Power Systems machine running IBM i as an integrated platform, particularly because it is so obviously true. I am not sure how much that would change the IDC data. And second, IBM should be making more noise about the fact that it was integrated when integrated was not yet cool, that this particular IBM i integrated system can run legacy applications as well as new ones written in Java or PHP and with some help, in Ruby or Python or Node.js or whatever. (OK, Python and Node,js are iffy, but there are good reasons to support these programming languages.)

Maybe what IBM needs to do is just create a Linux shell that runs atop IBM i and makes it look like a Linux machine–get rid of all the Q commands and replace them with Linux commands that are converted to OS/400 commands on the backend. Make it look and feel like a modern, integrated Linux, and maybe put an Oracle and MySQL skinning layer atop the DB2 for i database. Use some of that excess Power8 processing power to make it look like a LAMP stack machine even though, under the covers, it is better. Call it a new emulation mode, but rather than emulating a System/36 or System/38, like the AS/400 did when it was first announced (and can still do), it would be emulating the functions of a Linux-Apache-MySQL-PHP box that can also run Java. Maybe we can even convince IBM to charge the same prices for this box as it does for the Linux-only versions of its Power Systems machines.

Hmmmm. Probably not.

If you add up the integrated infrastructure and platform businesses together, by the way, IDC says that this business grew at 45.6 percent in 2013 and hit $7.6 billion in sales. Just under three-quarters of the revenues in the integrated infrastructure market, which came to just under $4 billion in sales and which doubled last year, were based on compute nodes from Cisco Systems. The amazing thing is that Cisco has not tried to build its own application stack. It has enough money to buy Red Hat and a database maker and take a run at it, but it probably won’t do that. It is cheaper to partner with all of the software suppliers than try to take them on. At least for now. If the integrated platforms market starts riding a hockey stick up–and it is not doing that now, unlike the integrated infrastructure market is–then Cisco could change tactics. But right now, clearly Cisco called this one right and is benefitting from it.