Published: 01/05/2009 | Last Revision: 07/07/2010

It’s time to have that hard discussion about the new wave of cloud-based technologies that are tempting developers and system integrators with the promise of datacentre reliability and uptime, and price tags that won’t break the bank. To be honest, it’s a bit like sitting down with your Dad to talk about the birds and the bees. There’s lots of mumbling and embarrassment, and wishing that the whole topic would come to an end within the next three seconds. But the reality is that this is a topic that we cannot afford to just skip over – to assume that everyone has done their homework and that everything will be alright. I will state the stark truth upfront: some of the players in this new marketplace are demonstrating a level of naivety or sheer arrogance that makes my head explode. I can just about deal with the fact that it is entirely my fault if I choose to use Microsoft Excel for some number crunching. Perhaps I should have used Mathematica, but then both are wrapped up with service level agreements, liability clauses and chunkily impressive paragraphs of licence text written in ALL CAPITALS to ensure you understand that you use this product entirely at your own risk. Indeed I have put such terms on software that I’ve developed in the past: it is pretty standard stuff. Then we have online services such as the provision of ADSL. Back in the days of Kilostream and Megastream lines, a chap in a suit and briefcase would be dispatched at 3am on a Sunday morning to come and fix your line when a fault was found. Today, in the new cheap world of ADSL, someone will come ‘soon-ish’. If it is a big outage, then they will do their best but it could be some days before you are reconnected. Again, we accept this because we know that if we really need reliable data telecoms into our office then it’s down to us to install two feeds, preferably from different vendors and hopefully taking different routes to the local exchange if possible. Then we mitigate risk further by having our outward-facing servers held in rack space, either at our ISP or at one of the big hosting centres in London. Then, if we do have a line failure, our remote mail server and our e-shop are still up and running and we can access it from our cell phones if necessary, and certainly from a laptop with a 3G card. Preferably from the local pub! Again the risk element is well understood and the cost/benefit ratio and risk to business interruption have been well documented and demonstrated. Yes, people still get it horribly wrong, but a quick slap round the face normally resumes sensibility. You will notice that there is a common design theme going through this: it’s your data, held on your server at a location relatively close to you. It might not be easy to get access to the datacentre for a disaster recovery at 3am, but it is possible, or your ISP can pull it from the rack and hand it to you at the front door. The recovery and business continuity that we demand is there in our hands, and we can take responsibility for our ongoing connected state.

Into the unknown

However, with the move to cloud services, some very important and well trodden experience is being thrown away and replaced by assumptions. These might work. Or not. For example, if your Exchange Server/Navision/Sharepoint/Azure Cloud services virtual machine is running just fine, then we are all happy. If it breaks, then the SLA is frankly meaningless. Microsoft, for example, will refund you the month’s fees if the outage is serious [revised to 12 months as we went to press]. If it is terminally serious then, as far as we can tell, Microsoft doesn’t want to know: our understanding is that compensation does not scale with increasing failure time. In other words, your business might be being destroyed while the call centre prioritises other customers, and there is nothing you can do about this. Furthermore, you have no physical access: you can’t turn up at 3am and demand to fix your server. You have no control over or even idea as to where in the world your e-shop/hosted Exchange Server is actually running at any particular moment in time. Worse still, Microsoft views its cloud as being one big cloud that is under its control, which means it has the right to move your stuff anywhere it likes in the cloud. When I tackled Prashant Ketkar, Director of Product Marketing for Cloud Infrastructure Services at Microsoft, at the recent MIX09 conference, he wouldn’t budge on this. Microsoft has the right to move your stuff anywhere it wants, including out of the EU which would almost certainly mean that you, as a company director, are breaking the EU data protection laws unless you have taken special steps in advance to allow for this. Microsoft will not allow you to limit the location to Dublin and Amsterdam. Your data might even land in Singapore, which is even worse than the USA. What if your business infrastructure lands in a US datacentre and the FBI turns up and shuts it down? Tough – Microsoft takes no liability for this. Following a phone conference with myself and HardCopy contributor Kay Ewbank, I have a bunch of questions open with the UK product manager for hosted services. Some six weeks on and there has been no coherent reply. If you believe the hype, there is this huge myth that having a cloud service is intrinsically a good thing. One aspect of this claim is that they can scale up your Web service or e-shop on demand while you remain online and operational. No longer will you have to scramble for more servers and attempt to get the load-balancing optimised at midnight, simply because people in Los Angeles have gone nutty for your product. Of course, if your entire business is e-based, then this is just fine. But out in the real world of product manufacturing and shipping, having a Web farm that can scale up at the click of a button does not address the problem. The real problem is that your 500 stock items will soon be depleted, and no amount of server scale-up is going to replenish your warehouse, increase the size of your packing department, or create an accounts department able to cope with the flood. It’s all part of the e-myth. The more I look at the big players in the cloud space, the more I conclude that they have a very clear vision of the sort of customer they want. USA-based is a number one issue. E-based is number two (they don’t want any pesky real manufacturers because that dents the scale-up argument badly). If you are prepared to roll over and sing ‘The Stars And Stripes’ then that’s even better. Be an unhelpful European and ask awkward questions about SLA, time to recover, legal liability and so forth, and they just stare at your impassively and say “we have no announcements to make on this at this time.”

A different vision

At the recent VMworld conference, VMware came up with a rather different vision which was articulated by CEO Paul “I invented Microsoft’s lock-in strategy” Maritz with a wry smile: “It’s like Hotel California – you can check your data/business processes in to the big three cloud service providers, but you can’t check them out.” VMware is taking a different route that involves small and medium-sized hosting companies within your country of business, with the facilities to buy more space from other companies in other countries if you need it. Put your virtual machines into their hosting space and forget all about these grandiose plans for global cloud computing; the business framework doesn’t need it, the legal framework doesn’t allow for it and the liability framework is almost non-existent. Work with local applications and then move the VM into the cloud. Have local private clouds inside your organisation; have public clouds for public facing services – you choose. Put your stuff where you want it and when you want it there. Move the VM back to a local server if you want to check over its health and stability. Use a solution which requires you to make changes to your application so it can run on the Cloud? What a silly idea. Have a Cloud infrastructure that you can’t run locally? Who would be mad enough to come up with that idea? Mr. Maritz is smiling like the Cheshire Cat. So if your service is truly global facing, contains no personal data, has no customer SLA guaranteeing uptime, is an e-solution which isn’t a e-shop fronting a real warehouse of products, and it’s not a big issue if the whole thing goes down for an undetermined period of time, or goes to Singapore for a holiday, then cloud services are just the ticket. For the rest of us, the robot is shouting “Caution, Will Robinson” with good reason. UPDATE* As we went to press we received a statement from Microsoft suggesting that it would technically be possible to ensure that the data of a UK customer of Microsoft Online Services did not leave the EU and that, if a customer made such a request, Microsoft would work with them to satisfy their needs. Grey Matter is looking for further clarification on these issues. Visit www.hardcopymag.com for the latest information.*