Serverless architectures - a great idea, but some caution advised
Content Copyright © 2017 Bloor. All Rights Reserved.
Also posted on: The Norfolk Punt
I am quite enthusiastic about the Serverless Architecture option – partly because it sounds just like Cloud computing always promised it’d be like; and partly because if mainstream IT doesn’t take the concept seriously, it is a powerful tool for Shadow IT to build something akin to the late 20th century “spreadsheet hell“. Mainstream IT needs to embrace and manage Serverless Architectures (with a fairly light touch) while it still can.
I was reminded of this when Bloor’s Research Director for IT Infrastructure, Paul Bevan, tweeted about “zombie servers”, which prompted me to find an Adobe blog which apparently claims that your business won’t use a server in 5 years’ time. In broad terms, this is probably about right – a lot of today’s servers are expensively idle much of the time (“zombie servers“). But, in 5 years’ time, I bet big banks are still running on big-iron mainframes even after “everyone” has gone Serverless – Serverless is a powerful technology option, when it is appropriate, but it is not always appropriate and it does need management.
So, I bet your business does use servers in 5 years time, but it’ll just be using them more effectively – and it’ll probably use fewer of them…
I remember “spreadsheet hell” back in the 1990’s – and the look of horror which went over our spreadsheet guru’s face (this was the guy who kept telling me that spreadsheets were going to replace all of our IT systems), when I pointed out that his favourite product, being used on mission critical banking systems had a serious reported bug, around doing maths correctly, and that he’d better find everywhere it was being used and get it fixed and make sure no-one was using the old version…
IT is good at reinventing wheels; and sometimes it reinvents square wheels. “Re-inventing the wheel” is almost IT “good practice” as long as you change the name to obfuscate the reinvention – but let’s not recreate “serverless hell” in homage to “spreadsheet hell… It’s not just a development issue, but also a governance issue.
I’ve had a useful comment from the folks at IBM:
“We think David is making a good point about governance, but we don’t believe the ‘spreadsheet’ hell analogy really works. The ‘spreadsheet hell’ issue is actually no different to risk of just copying and modifying any code – it is really a ‘copy by value’ issue where you lose connection to the original versus ‘copy by reference’. Serverless does not inherently suffer from this anymore than other environments. Without adequate governance you can always lose control. Indeed with a cloud based environment such OpenWhisk on Bluemix where the actions are stored in a central repository you are in a better place. You should still have versioning, code control, a proper automated tools chain – perhaps using something like the serverless framework to support it but actually if you do lose control everything is in one place and therefore you never really have the equivalent of 1,000s of spreadsheet each with ‘own’ copy of a macro…“.
I think they are probably right, in the context of a well-managed mature organisation (and I was trying to be provocative). However, never underestimate the possibility of chaos resulting from clever people operating in an immature organisation, and there are still some of those out there! A colleague has also complained about my implication that “spreadsheet hell” is over. He thinks it is still here – and I guess he is probably right too. One of the big problems in IT is that superceded technology (OK, I admit that I dislike spreadsheets for anything important, sorry) never goes away, it just moves into “maintenance”…
Comments are closed.