back to article When rip and replace is not an option

Today’s brochures about data centre design show pictures of row upon row of rack-mounted devices, and come with the clear message that in networking, servers and storage, conformity is the answer, efficiency is the goal and value is the inevitable result. This message is reinforced by many parties. Networking companies such as …


This topic is closed for new posts.
  1. Pete Silver badge

    So which do you think is cheaper?

    > Unification, consolidation, simplification – it sounds attractive. But not when you take the real world into account

    On the one hand, we have buying in bulk from a couple of preferred vendors (obviously, not going with a single supplier - that would be utterly insane, for so many well-known reasons) where you can play one off against the other. With the discounts that come from buying significant amounts of "stuff" and with the leverage to extract cheap or free support with the tacit threat that you'll drop them instantly if they fail to deliver.

    On the other we have every designer or project choosing their own hardware, O/S and applications based on what glossy they got through the post that morning, or which vendor's sales lady (or gent) is prettiest, or even what course they did at secondary school. Ending up with a mix of Dell, HP, Sun (ok, that was a joke), various flavours of Linux, W/Server 200x, Unix, Oracle, SAP, and a dozen others. All needing specialised support staff on site, all needing their own upgrade schedules, test environments (ok, that was the other joke), backups, DR, patches, bug-fixes and support contracts. None of which amount to a sizable enough stake for any particular vendor to hand out special terms and with third party support which, likewise, involves hanging on the phone for a slow-reading voice to ask if you've tried rebooting.

    While this first option may well tie you in to a particular regime, the costs involved in being able to deploy cheap, generic hardware - that can be swapped around, reallocated and load-balanced in next to no time, and cheap generic support staff who will line up outside your head office at the merest hint of a vacancy,, is hugely attractive. Especially when you consider that the average engagement time of an IT director is measured in a few years, so they'll reap the rewards of the cost savings without making the mistake of sticking around to clean the **it off the fan - and you can see why data centres look more like a stockroom for blade servers, than the diverse environment of a computer fairground.

  2. Anonymous Coward

    *Journalist in rehashing known facts shocker!*

    I don't think that companies like VMware are advocating wholesale change or offering a way of "rip and replace" at all. They certainly don't suggest that you need "conformity" in your infrastructure. The point (or one of them) that virtualization companies make is about the increase in efficiency you can get from consolidation: whether this is on new OR existing hardware (apart from hyper-v which needs 64bit hardware).

    VMware in particular advocates "peaceful co-existence" - a kind of detante for physical and virtual machines and all types of OS. Sure, you can replace physical machines with virtual but you still need physical machines to run them: the fact VMware supports a shed load of OSes (unlike other rivals) would support the suggestion that in fact, you're way off beam with your "conformity" angle.

    Your point about the cloud, whilst valid, is already well known. The IDC and Gartner have both published papers which already state exactly the point you make: that no-one is going to buy EVERYTHING from the cloud and that what will most likely evolve is a hybrid model with IT departments using virtualization, the cloud and traditional physical machines to supply their IT and business needs.

    I'm not sure I get the point of this article at all. It doesn't really advocate anything other than the final message: don't panic. We all know that businesses will take a good look at the benefits: only a fool would go headlong into a project to virtualize or unify or whatever ALL of their datacenter without weighing up the pros and cons.

    As you were then, eh?

  3. Jon Collins


    Hello Anonymous Coward,

    Thanks for your feedback. For the record, I agree wholeheartedly. I brought the topic up with reason, given that my last conversations with VMware and a number of other vendors most certainly weren't advocating peaceful co-existence - indeed, the debate I participated in at VMworld (and instigated) was that VMware was suggesting everything could be done inside the virtual sphere, without considering the virtual-physical hybrid you mention. Also participating were representatives from Gartner and IDC, who were confirming the same thing, i.e.: "VMware, you need to recognise the world isn't going to look like your vision, and act accordingly."

    Not everybody has access to Gartner and IDC papers, hence one of the reasons for this column, written by analysts, not journalists. I do concur with what you are saying, but the point remains that there exist organisations (including other analyst firms) presenting a view of the world that suggests otherwise. It is therefore worth putting the more grounded view to balance things out, just as I have, and so have you.

    As we were indeed ;-)

  4. Jon Collins


    Indeed, sir, and all for cheap, generic hardware. It's about making the right choices for the right decisions, for sure. I think "Anonymous Coward" hit the nail on the head with his "Don't Panic" remark - I would extend that to "Don't believe any one technology is going to solve all problems". And while this may be an obvious thing to say, anyone who lived through the mistakes made during the outsourcing wave, say, or the dot-com boom, is quite right to be nervous of the hyperbolic rhetoric around virtualisation, cloud, unified data centres and so on.

  5. Anonymous Coward


    disagree in general. I think the article could be profitably read by more than a few specific tech/OS fanbois, none of them Apple.

    In big corporate sites I have seen a small number of unix servers ripped out and replaced by literally thousands of X86 boxes, all because of a CIO convinced of a beautific vision by brochure ware. Cost immense, performance worse, energy efficiency duh.

    I agree with the author, because of the multiple sites I have worked on over the decades, all have had different origins, work loads, uptime needs. This meant they all had very different histories. All changes made were, wait for it..., designed, planned, and sometimes, even tested. With varying degrees of ineffectiveness , naturally.

    I can see why this might be confused with accumulating random changes, given the average CIO is a sucker for salesdroids.

This topic is closed for new posts.

Other stories you might like