With the upcoming major OS upgrade (V6), the iSeries will not really exist as a system in its own right, but it will share hardware and software resources with many other environments that will run on what is now a unitised PowerPC platform. The announcements surrounding the new i5/OS version focused on issues such as virtualisation, security and web services support. Concurrent with this announcement was the news that a new version of the iSeries was being launched to run on the new PowerPC 6 – hailed by some as ‘the most powerful processor ever built’ (whatever that is).

If one reads through the announcements in detail, there really is little to interest the bulk of the iSeries world of users. Most of the new features are targeted towards ‘big shops’, where the iSeries annual investment is measured in hundreds of thousands of pounds, as opposed to the SMB world where systems are expected to run for more than forty minutes without being upgraded. So if there is so little detail in the announcements of real interest to the bulk of iSeries shops, we should ask the following: is IBM leaving us high and dry?

Back in the early Eighties, the AS/400 was most definitely an SMB offering; if you had a need to run financials, payroll, manufacturing and perhaps some reporting over a modest business, then the AS/400 was certainly for you – there was little else on the market outside of the overly complicated world of UNIX. If you had hundreds of thousands of customers and massive payrolls, the mainframe world was your preserve. These two worlds have effectively merged and a medium-sized modern iSeries is well capable of competing with the largest mainframes available at the time of  the AS/400’s conception.

When one places the iSeries in the context of big x86-based servers, it becomes clear that anyone reading these news announcements would be justified in questioning the viability of the iSeries for the SMB market space; let’s face it, the hardware still costs too much for what it is, the system requires support by specialists who know what they’re doing (as does any other system, but the fact is iSeries skills are probably becoming more difficult to find) and, on top of that, there is little application software being written for or ported to the iSeries that can compete on price with the Sugar CRMs and OpenTaps ERPs of this world (to save you from searching the net, I can tell you they’re both credible, feature-rich, open source enterprise applications). So (my monthly rhetorical question coming up) I hear you all shout: ‘Why stick with the iSeries?’

To address issues of investment viability, it’s important to reflect on the fact that the issues surrounding the viability of the iSeries are different from those surrounding other systems; UNIX can be had for free in various forms (Linux, OpenBSD and Solaris being just three) and will run on almost any processor base; Windows Server will run on almost any properly specced x86 box; i5/OS on the other hand is proprietary, it will only run on hardware specified by IBM and none other.

 
The iSeries, in contrast to Windows, is immensely scalable – whereas i5/OS needs only more resources (in the same box if you want) to scale, Windows has to start messing around with clustering and, to be honest, Microsoft have only recently started to get Windows clustering to the point where it’s both cost acceptable and reliable. The iSeries has supported a form of internal clustering (virtualisation) for many years, but it’s always been reliable and has been a major part of broadening the appeal of the machine in recent years. Even with its undeniable reliability, the question remains: ‘Would you build your core systems, such as financials and manufacturing, on the iSeries today if you were starting from scratch?’

This question cannot be easily answered; for every answer there will be many other arguments. For me, a major issue lies in the area of application support. When I started in IT, all systems were bespoke, written to meet the express needs of each individual user company and, to be honest, they never seemed to reach a point of conclusion. Within the UNIX and Windows world, there are literally thousands of competing applications – so many applications that the need to tailor software has diminished enormously, but the iSeries world has not moved in this direction.

Let me explain by means of a very simple example; there’s a good friend of mine who works in Glasgow as an independent IT consultant. He’s recently been involved in helping one of his clients to select an enterprise CRM solution; the client (a medium-sized manufacturing company) has a significant investment in both iSeries and Windows technology and briefed the consultant to look at all viable technologies. Unsurprisingly, my friend found the Windows and UNIX worlds are awash with good, feature-rich CRM systems, but the iSeries world is a Third-world country in comparison.

IBM only endorses a single CRM product for the iSeries (WinTouch CRM); my friend looked at WinTouch’s feature set and found it ‘stunted’ (these are his words, not mine) when compared with the CRM market leaders (most of whom focus on Windows or Unix/Linux). In fact, after comparing WinTouch  with CentricCRM and SugarCRM (both of which are open source offerings), our intrepid consultant struggled to look further. This clear lack of ‘iSeries native choice’ should resonate loudly (and with great anxiety) with anybody looking at the iSeries as a potential platform for deploying simple packaged software.

Products paucity

There must be good reasons why there is a comparative dearth of good CRM products for the iSeries; surely the rewards of deployment to the iSeries are substantial? Well, I’d say the market is telling us something – it does look as though the commercial risks intrinsic to application development on the iSeries are unacceptable and, if this is true, why so?

My colleague suggests that perhaps the risk of developing software applications in the Windows and Linux environment is less than would be the case for the iSeries, principally (he argues) because there are many more non-iSeries developers in the world. Right? Absolute nonsense. In fact, I can pretty well guarantee that almost any application developed for the Linux and Windows platforms can be easily ported to the iSeries platform – there only needs to be a commercially sound reason for doing it.

Most software now is (or can be) written independently of the hardware it runs on; Java apps can be ported to any machine equipped with a standard JVM; there’s a reverse-engineered .Net platform for the Linux world that is currently being ported to the iSeries platform and, of course, this will allow any application built on the .Net platform to run on the Linux and (eventually) iSeries platform and, whilst you might not see much movement in the near future, I think it is no leap of imagination to predict that due to technologies such as virtualisation (and in the final analysis all Java and .Net apps run on a virtual machine of some kind) will render these distinctions innocuous within perhaps five years or so.

We already know that due to hardware abstraction strategies we can continually change the hardware elements of the iSeries without informing or otherwise involving users; I believe this will evolve to the point where hardware simply doesn’t matter; so long as there is some kind of platform, and the correct virtualisation layers exist, any software can and will be made to operate on that hardware within its native platform environment. If this comes to pass (and I think it will), then it must (again) beg the question: ‘Why stick with the iSeries?’

For me, the answer starts to become obvious when one matches investment against performance and functionality. I can get an iSeries machine specifically tuned to exactly the right size and capability to reliably run my financials, planning, warehouse, email and groupware and any other core software – and given that in the future (pointed hat on) I’ll be able to run other ‘non-iSeries’ applications from my server, I really believe the question must be reversed and asked in a different way: ‘Why wouldn’t you stick with the iSeries?’

Now, in continuing I must declare an interest: I am a member of the *noMax UK team. It would be wrong of me to endorse the *noMax product (particularly in this issue), other than to say I am proud to be associated with both the product and the company. With my honest declarations over, I want us to look at system reliability. We live in a fickle world that tends to be run by bean counters of one form or another; all they want from their ‘IT team’ is a perfect system that never fails – and, oh yes, one that you don’t need to spend any money on either! I know of few bean counters who are really aware of the value of their IT infrastructure; they tend to become aware when it doesn’t work, but at all other times it’s taken completely for granted.

Transaction preservation

Few non-iSeries people seem to understand the value of transaction preservation – they talk in terms of data protection. Most good IT departments have solid back-up plans; they take regular backups and make sure the back-up tapes are taken offsite. The really good IT departments will have a disaster recovery plan that lays down exactly what must be done to restore systems from bare metal and get the business up and running after a disaster. The best IT departments realise a disaster must not be allowed to compromise either data, transaction or system integrity. But the fact is building systems from bare metal takes days, weeks even, and if core manufacturing or distribution processes are dependent on system availability, then serious wedge and credibility will be lost while you find out your back-up tapes simply aren’t up to scratch.

HA is now an imperative for most businesses, whereas twenty years ago batch systems were king and you could always run your batch jobs the next day (maybe on someone else’s machine). The world has changed and transactional systems are the order of the day – lose a transaction and you may well lose an order; lose hundreds of transactions and the implications are obvious.

I have worked with a few Windows-based HA systems and for the most part they’ve not been inspiring; they generally have to introduce a layer of technology that monitors the system I/O ports and then uses this as a trigger to migrate data changes across the network. The iSeries, of course, achieves this by means of remote journaling and many respected studies have shown that HA via Remote Journals incurs a surprisingly low impact on system overhead.

The arguments regarding the eminence of remote journals as an effective means of managing availability has (I believe) been laid to rest. For example, a couple of years ago EMC introduced their ‘Recover Point’ network appliance for Microsoft Exchange and SQL Servers. Recover Point effectively introduces a remote journaling system for Windows technologies by means of a hardware intervention solution. It’s expensive and commercially viable only for large Exchange installations. EMC could have used many strategies to achieve its goal, but chose to go with what is effectively a local/remote journaling model – we get all this wonderful functionality on the iSeries natively; all we need to make it work is an appropriate application to manage the journaling effort for us.

With our worlds converging, we will see the emergence of true platform-independence, but if we want cheap, effective and reliable HA, look no further than the iSeries – even if you decide to run a .Net, LAMP or J2EE application stack on it.

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