So that's AIX and Power PC heading for the rubbish bin then....
I'm guessing that leaves Windows (obviously), Linux, Solaris and IBM Z series?
Deutsche Bank is re-engineering its "excessively complex" IT – including slashing the number of its operating systems from 45 down to four – as part of a sweeping strategic overhaul. Under its major rationalisation plans the bank will also pull out of 10 countries and cut 9,000 jobs globally. Part of that will also include …
The BSD family (FreeBSD, OpenBSD, NetBSD) - which are sufficiently different to count as three different OSs (unlike Linux, which just defines a kernel)
(those three because it's Friday; hopefully they are not in use)
Beer icon, because TGIF! Edit: and because somebody beat me to it already
My assumption is that someone was asked to produce a large number so probably took the approach of counting OS versions so that Windows XP, 2k, Vista, 7 and 8 all count as different OSs as do Red Hat 5,6,7 and Centos 5,6,7 and however many other flavours of Linux are deployed.
To be fair when looking at consolidating what you have to support that isn't an unreasonably approach, IT generally only want to support once specific OS release version not any flavour of Linux that a random dev happens to favour.
I wonder how much of this is driven by legacy apps they can't replace and more modern apps that have their own system requirements. I knew of an old pick app that was mission critical and no one could manage to replace it. They tried. Oh they tried. Every young new CxO tried to get rid of that thing with something shiny and new. The new hotness could never do the job.
So forget operating systems. I wonder how many database platforms they support because of the aforementioned application system requirements...
I've worked on a few projects that were set up to replace legacy systems and they all failed because the businesses failed to appreciate how much these systems do and how much it costs to replace them.
What a lot of companies also fail to do is ask the question of whether they need to replace their legacy systems. Many of the owners of these platforms, like Rocket who own Universe, Unidata & D3 (Pick-like systems) have been developing them over the years to natively integrate into the modern world such that they can provide SaaS capabilities, web services, Object Orientated programming and easy interfaces into their databases, which are NoSQL.
To me it seems that many companies are wasting a lot of time and money trying to replace something that not only works, but is cost effective, cheap to run and can be engineered to provide the same facilities as any modern environment.
Some of the fault lies with the platform owners in that they fail to market these new developments to the people who use them. To me they should be knocking on the doors of their customers and promoting their latest enhancements and not just assume that the IT Managers understand what they are and leave it to them to do their marketing job for them.
Likewise, I have been involved with companies where I'm looking after a Pick-like system that they're trying to "sunset" (lovely euphemism for kill-off!).
In every case but one, they've failed. Even then they succeeded only because their business changed drastically when the recession hit and they were then able to adjust their business practices to suit a commercial package. However, it wasn't before they'd spent a lot, and I mean a lot, of money, all of which was for nought in the end.
One of the biggest problems is that most of the systems are decades old and have been highly tuned to match the requirements of the business as it grows. In fact, in many cases the businesses have become successful because they've had an application that they've had complete control over. In a sense, they've become victims of their own success when they then try to get rid of the application.
I understand the need for businesses to move on, particularly where they've got a system that is capable of being maintained and enhanced only by an ageing engineer population. They clearly find themselves between a rock and a hard place; dependent on a completely tailored system on the one hand and worried about the longevity of available resource on the other. Sooner or later they'll be forced to bite the bullet but it'll be a brave CIO (or whatever) that sees it through.
Re; between a rock and a hard place
I think part of the problem is that many people in (and entering) the industry only want to 'play' with the latest and greatest packages ie. brand names to put on their CV's in some belief that this will guarantee them a successful career - with Microsoft seemingly trying desperately to commit suicide, there will be many in the industry who may find life post 2020 not so comfortable...
I too have seen many attempts to remove PICK and MultiValue applications, usually very expensively and as complete failures.
The problem is, these systems are extremely reliable and require next to no maintenance: which means that pretty soon nobody knows anything about them. If they went wrong or needed daily feeding by an army of DBAs and support engineers crawling all over them, everyone would stay current with what they actually do - it's often the lack of business knowledge that both leads to the attempt to remove these systems and then explains why the attempt fails.
PICK/MultiValue handles complexity extremely well. It's an old model but, hey, so is SQL .. in fact, outside of mobile has there been anything REALLY new in the last 30 years? It often looks like these shiny 'new' systems just plastering more lipstick on the old piggies.
I'd forgotten Tru64, but I got up to 27 just counting Digital operating systems alone (well, counting every variant of RSX-11 which is, I think permissible as each one was inexplicably different from the others) and to at least 40 if I included third-party OS's (including Unics [sic]) for DEC kit.
A few obvious omissions:
OS/390 - I bet as a bank they still have stuff that won't move to z/OS
Bull - common in parts of Europe:
GCOS7 and 8
Bull supplied edition of AIX (I've forgotten its name).
They've probably also got some Unisys/Sperry/Burroughs kit tucked away doing essential stuff, so that's another couple of OS's.
Also shouldn't forget Siemens produced a range of kit at one point, as did NCR.
And as some others have pointed out there are quite a few legacy OS's under the HP name, so could be some: MPE, Tandem non-stop...
However, I doubt there would be any ICL/Fujitsu kit...
Some may also be including the OS's in the SAN controllers(*), ePOS, major network appliances.
(*) I kid you not, I remember one CIO getting very upset when he discovered the SAN controllers didn't use one of their strategic OS's...
Certainly it would be interesting to see the list.
They probably are counting Linux several times in that 44, as they may have a half dozen different distributions, and will standardize on a single one. They might even have some on different architectures (i.e. Linux on POWER, Linux on mainframe) and for other OSes as well (Windows on IA64, Windows on Alpha) There are various different OSes you can run on a mainframe like VM/CMS, they will get rid of all the really ancient ones and go z/OS only.
Then you have boutique high availability stuff like Tandem, you have ancient servers sitting in a closet somewhere (Novell, OS/2, SunOS)
It would have difficult to get to 44, but I think it is possible even without having anything dating from earlier than the 90s) and without having to count "Windows 2012" different from "Windows 2008" or RHEL5 different from RHEL6. If they count like that they'll never have only four, because you will always have multiple versions of Windows and Linux at any point in a large enterprise. It isn't feasible or even possible to ever reach a point where they're all the same rev.
Remember: work outsourced is work not done. So just save the money and don't do it.
Too late. Deutsche Bank signed an infrastructure deal with HP in February.
HP provide my company's infrastructure, desktop & support. Standard of service is bloody diabolical, and our poor IT is a major competitive advantage for those competitors who have been sold snake oil by outsourcers. Unsurprisingly to everybody (except the C-level halfwits) is the fact that the cost of our baseline services increased, and to keep total IT costs within the allowed cost profile, we've had to slash the IT projects budget. So our HP-delivered cost saving is in net terms delivering less, and to a lower standard than our previous in source arrangements.
...or get fired in a round of cost cutting. I was working at such an outsourcer when they decide to rearrange the deck chairs. They ditched telecommuting in favor of a physical location that no good IT would want to relocate to. Then they fired all of the legacy staff that had all of the tribal knowledge. Of course it was a total disaster. What was left of the telecommuter admin pool had to make up for the idiots sitting in the middle of Montana. Eventually they just ditched the idiots in Montana and were left with the telecommuters minus anyone who had a clue.
Sometimes I wonder why corporate America doesn't just implode.
HP provide my company's infrastructure, desktop & support. Standard of service is bloody diabolical,
One night, after a physical move of a handful of racks with Alphas and SANs in it, one of the SANs decided it didn't want to be serving the cluster any more. Which was a rather undesirable situation, from a redundancy point of view. So after our own troubleshooting skills were exhausted with no result, I got on the blower with HP, and after a short interlude ended up with Sanjay* in Bangalore. Who, after having had the problem and the symptoms explained to him, responded with "What's a SAN?". The words "You're bloody fucking useless" must have been heard in Bangalore even without the benefit of telecomms, and a dent appeared in the far wall caused by a very mobile telephone
* Some Indian name anyway, at least not "George" or "James".
Would this be part of returning to the traditional role of moving money around. Where once we had thousands of tellers we need only someone to turn the lights off in the data centre.
Freed of all the ridiculously paid people inventing and gambling with their instruments or selling PPI they might actually concentrate on delivering a simple transaction service efficiently, reliably and at negligible cost. Like RBS/Ulster Bank found impossible for how long? Or even finding a better way to connect lenders and borrowers?
DB won't see out another 24 months as a solvent, functioning bank, so trying to trim a few costs here and there is really a bit pointless.
As an aside, simplifying IT should be done for the right reasons. In DBs case, that would be to open up new opportunities instead of just cost cutting.
Really? - care to explain what great insight you have into DB that lead you to that conclusion? The only way DB wont be here in 2 years is if the entire financial system implodes and if that happens we'll all have more important things to think about.
Its underlying business is generally profitable - revenues are around 30-40Bn pa - the vast bulk of its losses have come from Regulatory fines and Accounting write downs.
If DB stopped trading tomorrow the inertia of that 30-40Bn of revenue flow would keep segments of it trading for years to come.
With pathological need to reinvent the wheel in his favourite language.
Simply refuses to use anything Microsoft. It must be a framework on a blog, completely unsupported and with no fixed code base, but has lots of programming by side effect. Doesn't do half the things C# does, but it's not Microsoft because Bill is the devil.
Treats the whole organisation as a place which is there to fund his "creativity"
Repeat until bankrupt.
I haven't seen many, if any, Colins making an impact at Banks, but I have seen plenty of business folks stamping their feet until they get their latest favourite tech fluff installed. Case in point, over the last 20 years 100% of the banks I've worked at have had one or more Crays powered up but idle in their datacentres for several years. In 100% of the cases the purchase was driven by the business, no Colins necessary. :)
Seeing as "virtualised" == the same hardware just with a hypervisor burning 50% of the resources, and private cloud == the same hardware just with a different name on it, plus - as many have already mentioned - it all admined by a small subset of the Gurus supplemented by cheap idiots....GLWT!
"You obviously have no clue about how hypervisors work if you don't realize the massive savings in resources that result."
I know how they work. In the raw, they have zero savings -- they have (claimed) ~1% CPU overhead and (measured) more like 10%+ CPU overhead. Single-digit-percentage disk and network I/O overhead with the right paravirtualization drivers and support, or pretty high overhead without it.
The highest savings are probably with Windows deployments -- where there's a high likelihood of app A requiring it's own server, app B requiring it's own server, and app C requiring it's own server, not due to resource usage but due to these apps (maybe) stepping all over the filesystem. It's definitely better to have one 75% loaded server due to 3 VMs, than 3 20% loaded servers.
Linux software tends to not play the "you must use your own server" game, but nevertheless virtualization would help if they preferred different distros or what have you.
Don't get me wrong, I'm a fan of virtualization. But the virtualization has significant overhead, it's the higher server utilization that saves. (Plus, of course, being able to move things around more easily in a virtualized environment is nice too.)
It is rare for servers to be highly utilized in both memory and CPU, and even if they are during peak times you have off-peak times where a lot of resources go unneeded. Virtualization smooths a lot of that out, so you need far less resources. If you'd ever been involved in a large scale virtualization project and seen the before/after even when servers that were fairly well matched to their needs were used, you'd know that.
It depends on how they're charged and whether you own your own kit or not; I've been in situations where the server charge was on number of operating system instances rather than physical boxes.
It makes a kind of sense if you assume the admin effort per OS is constant and the boxes cheap.
"Linux software tends to not play the "you must use your own server" game"
You obviously haven't done this in the real world. Linux is way worse for requiring distribution X, kernel version Y, with dependency Z that isn't compatible with anything else. Windows is a breeze in comparison.
1. Should I be worried that so far no one has listed/mentioned any Apple-OS? Surely there are some 'creative' guys with macs at DB, not to mention all the fondleslabs that are bound to be there. No, wait - an Apple IIe a spreadsheet and a database containing sensitive internal information datig back to Josef Abs.
2. There are probably some very old Siemens / Nixdorf boxes around and humming. Possible even one so old that it's technically a Zuse.
3. I would really like to read a new BOFH story in which BOFH & PFY go to Frankfurt to sort out a couple of legacy systems. Ah, the possibilites...
I'm sure there are Macs somewhere in the company being used as servers. Maybe even some ancient Mac running A/UX, or a NeXT in some dank corner, who knows?
If they know they've got 44 operating systems they must have done a pretty decent audit and found a lot of those hidey holes to get the count that high.
Bull GCOS 7
Bull GCOS 8
Fujitsu BS2000 (almost its entire customer base is in Germany)
Unisys OS 2200
Reliant UNIX (SINIX)
Fujitsu VME (unlikely in Germany tho)
Nixdorf Niros family
And that's where I start to run out of systems that are likely to be common in a bank.
You forgot a few:
I'm sure others can think of more
Plus they probably count different architectures as different:
Windows on IA64 & Alpha
Linux on POWER & mainframe
Solaris on x86
VMS on Vax vs Alpha
HP-UX on PA-RISC vs IA64
Might have a few supers out there, like Cray, those always run some special OS.
Maybe, just maybe, they're planning on returning to the 1950''s or 60's. Have a bunch of novice accountants and those old hand cranked adding machines.(Can't be electric as electricity costs money.. and no you can't have another lump of coal on the fire.) The few computer bits left would just be for calculating the C-suite's bonuses.