I always enjoy announcement time at Rochester. Every year some marketing wallahs dressed in dark blue suits tell us what we should want. They usually get it half right, managing to please nobody and disappoint everybody. Well, I must say that this year is different. It looks like the bean counters over at IBM Corporate have finally got the message from you and I. It seems to have at last registered with someone that there are credible alternatives to the System i and at last the price of our beloved monolith has been subjected to a rationalisation that makes it part of a rational world.

Coupled with this pricing revolution, IBM has recognised that people really can vote with their feet when it comes to system power and at last it has given up on choking the performance of low-end machines (an act I always thought was a little bit mean of spirit, to be honest). The result is the bean counters have finally realised that reliability, performance and cost of ownership are part of a very complex model that means different things to different people. Telling a potential customer that the System i carries low cost of ownership doesn’t deliver sales when the initial purchase price is three times the cost of a Windows or Unix rig of comparable power.

Price matters and IBM is at last waking up to the fact that it can make the System i a credible investment opportunity if it changes its approach and as I have said before in this column, without a vibrant user base the System i will drop off the peg eventually. It is of crucial importance to all of us with interests in System i technology that IBM starts developing new sales and I honestly think it has now hit pay dirt with the new pricing strategies.

Encouraging too is the news that IBM is continuing to invest heavily in System i technology, to the point I really do think even IBM is convinced the machine has a credible future. This is all very nice for people like you and I who are wrapped up in the System i, but what about new users? As I said earlier, we desperately need new users to bolster the support industry that makes the System i a feasible technology choice so, personally, I will break open the champagne when I see herds of new System i customers coming over the hill and until that time we should all just be grateful for high availability.

For some years now and ever since I started writing this column I have been banging the reliability drum. The fact is most people who buy into the System i do so because it is reliable. There are cheaper alternatives and there are friendlier operating systems but I just don’t see a more reliable machine for the money and the fact is that if you want high availability the System i is a serious place to start.

Although announced in 1988, the System i (or AS/400) remains IBM’s most recent major architectural development, conceived and developed in-house. And in all honesty we are lucky to have it. You see, when Lou Gerstner joined IBM in 1993 he hammered out a radical change in the company’s perceptions of investment risk and made it clear to anyone who would listen (and he wasn’t known to tolerate people who didn’t) that massive developments (such as the System i) were just too damned risky, and by and large many blue-sky developments were canned soon after he landed in the CEO’s chair at Blue Heaven.

Making acquisitions

IBM now strategically maintains its pre-eminent position in the technology market space more by acquisition than by  new internal development per se (look at recent takeovers like Rational, Lotus, Informix, ISS or MRO). Thankfully, the System i has tended to buck the trend of development by bean counting and this has proved good for all of us over the past few years. We are fortunate that when Frank Soltis conceived the AS/400 before Lou became blue, he came up with a future-proof design. Sure, the initial development was a major risk but the outcome was not. IBM can change the hardware, software, whatever, and the outcome is always the same: an AS/400 with new badges (nisi Soltis frustra!).

I suppose one really good initiative kicked off by Blue Lou was the massive  product rationalisation programme that took place throughout the past few years. At one time, IBM had four main ranges of servers (x, i, p & z) and they shared about three common parts (all screws I think, but I seem to remember the nuts were all different). Through an aggressive programme of re-engineering, Lou brought the System i and pSeries together as a single entity and developed internal design rules to ensure that components were adopted in a sensible way ensuring minimum inventory, maximum reliability and minimum cost to IBM. Now, at last, IBM has started to treat us like customers instead of milk cows and is passing some of Lou’s hard won shekels back to us the customer.

Is it too late?

All this largesse begs the question: ‘Is it too late for the System i?’ Face it, the server’s user base is half what it should be, its share of the server market has dropped steadily over the last five years and, to be honest, I am not surprised. While IBM has been banging a drum about total cost of ownership, HP has been stacking Windows boxes a mile high and shipping them at incredible prices. Indeed, more than half of all server sales now involve Windows in some form or other. In March, 2005, Windows accounted for only a third of server sales, so what have Microsoft/HP/Dell done right and IBM done wrong? Nothing, actually.

Many servers sold today are simply over-specced desktops with rails bolted onto the sides. Put simply (and not at all glibly), they are just not the sort of thing you would want to stake your life on. The System i is a different matter. Since 1988, I have been somewhat overawed by the machine’s reliability. In nearly twenty years of working with the AS/400/iSeries/System i (whatever you want to call it) I can only remember a handful of system failures that were of sufficient magnitude to result in serious unplanned downtime. But when those failures occurred, the machine’s owners suffered horribly.

In the right niche, Windows servers are great. They could make terrific small business servers if they were simpler to manage and cheaper to own. They could make good core system servers if they were more reliable and if Windows could keep its pants on (security doesn’t exactly spring to mind when you consider any version of Windows, Vista included). If you want a Windows server to offer the levels of reliability we are used to from our System i you need to involve multiple machines, clustering and lots of sticky back plastic and I’m sorry but not even Valerie Singleton can make a Windows server reliable. No, call me old fashioned (and I am sure you will) but for reliability I need something rather more substantial than a plastic box loaded with Taiwanese spare parts and a ropey operating system.

Having enjoyed my daily Windows bashing, I must point out that I am not complacent about the System i either. Sure it is the most reliable machine outside of the mainframe world but is that enough? Well, here is where a bean counting approach does make some sense. System i reliability has to be enough if you can’t afford a hot standby HA infrastructure but if you can, then it would be churlish to deny the value of two System is running in a solid failover configuration. The new announcements by IBM have brought HA within reach of most System i operators. With prices like £5,000 for a credible back-up machine you need never lose sight of a data transaction again.

Unlucky for some

You see, system failures don’t just stop your core processes working properly. If you are really unlucky, you can lose the detail of transactions that have not been backed up and if your system fails 23 and three-quarter hours since the last backup you could lose orders, money, goodwill -- the list could be endless if you are really unlucky. For the majority of smaller System i operators robust HA has never really been an option, the costs were simply too hard to justify and instead people relied on the System i’s intrinsic reliability and unfortunately once in a while someone got badly burned. With a back up System i starting at five grand, a feasible HA infrastructure is now within reach of anyone with a spare eight or nine big ones going spare. Better than that, it can be financed for less than the cost of your company car (unless you drive an old Skoda).

IBM has really got with the programme in this area. In addition to making the price of a back-up server feasible to all but dear old Ebenezer, it has made it possible to transfer licences from the primary server to the secondary server should the unthinkable happen. This means that if your big box does die on you, you can failover to the secondary server, carry on working (possibly at a reduced throughput) and get your dead server resurrected. While all this is happening and if you have done your planning right, none of your customers will ever know there has been an issue. I suppose some tight-fisted bean counters out there will prefer to drive a Skoda but, hell, you can’t please everyone.

John Amans has been involved in supporting the AS/400, iSeries and System i family of midrange computers since their launch.
john.amans@x4solutions.co.uk

This article originally appeared in the May, 2007, edition of System i NEWS UK magazine.