So one of the largest public clouds has suffered a major outage; it was going to happen as were the inevitable recriminations, crowing competitors and a general Nelson like ‘Ha ha’ from many. As a dawning of reality on many about the nature of single sourcing, lack of meaningful SLAs, credulous users and lack of architectural vision from many; do we find ourselves at a crossroads with regards to Cloud?
Amazon should be and are taking some heat on this; their lack of transparency has been frightening many for some time. The AWS dashboard is about as useful as your average ISPs and needs to be much more detailed or at least allow the expert user to drill down further. In this kind of event, accurate information is essential.
Public Cloud Providers such as Amazon need to take a lead in guiding their customers into how to design ‘infrastructure’ on top of Public Clouds for availability. This needs a full breakdown of risks and mitigation that a user can take to remove SPOFs.
For many the promise of Cloud was that you can get rid of your costly and obstructive infrastructure teams; these teams can be obstructive and costly because they carry the wounds of the past but don’t for one moment believe that getting rid of these teams will stop you suffering the wounds. The principles of availability, reliability and supportability are as valid now as they were in the decades gone by.
Anyone who has worked in Enterprise IT will tell you; ‘proof of concepts’ rapidly become products and you are suddenly running an infrastructure which takes no account of your non-functional requirements. The agility that Amazon almost formalizes and legitimises this practise. Your business model maybe predicated on Public Cloud but don’t imagine that this does not mean you can yet absolve yourself of the understanding of infrastructure and it’s impact on your applications and business.
We aren’t yet at the stage where computing is provided as a pure utility like electricity and even then, companies still have back-up generators, M&E teams, Facilities teams etc.
Yet, despite this the advantages of something like AWS to getting product to market can be amazing; we’ve seen a startling renaissance of new businesses getting up and running very quickly; many of these businesses may actually fail but lets hope that they fail because they were bad ideas and not because they built their infrastructure on sand.
I think that we are at cross-roads or at least at a significant point in the journey; I think we shall see some shoring-up and some rethinking but ultimately, the pain will be worthwhile. The utility providers such as Amazon will have to think a bit more about how they build their infrastructure, the same way the traditional utility companies had to; users will have to understand the nature of the utility and what it means if that utility is not available.
However, it’s far too early to say whether the model is flawed or not; expect more failures but expect their impact to diminish as people understand better how to design for failure. Expect new development techniques but also traditional architectural disciplines to have a place.
And expect more navel gazing from the likes of me and other wannabe pundits…if things like Amazon didn’t fail once in a whilst; we’d have to write about successes and that’s just boring!
