Content Copyright © 2018 Bloor. All Rights Reserved.
Also posted on: IT Infrastructure
A couple of weeks ago I had one of those ‘ah-ha’ moments. I had been researching the market for infrastructure performance management (IPM) solutions across hybrid cloud and on-premises environments. Digging beneath the bland similarity of the marketing messages it appeared that there was a difference, broadly, between those vendors that sample data from the infrastructure and those that capture streams of data at the per-second level. So what? Isn’t sampling data good enough to help me optimise the performance of my infrastructure? Do I really need all that stream data? My instincts were saying no to the first question and, yes to the second. But I couldn’t quite articulate the reason why.
Then, separately, I was discussing data centre energy optimisation with my good friend Stefan Norberg. Apart from being very knowledgeable on the topic, Stefan is a Swede living and working in France and, like many Northern continental Europeans is completely at home in multiple languages. I suspect that English is his third language so, one or two strange words do slip in from time to time and I’ll stop him to ensure I have understood what he meant.
He was briefing me on a small French company that provides energy optimisation solutions for data centres and other industrial buildings. This company, Kapsdata, was using IoT Sensors and machine learning to capture huge amounts of real-time data, linking it to server, storage and network load information to understand exactly how the cooling systems were working. As with IPM there seemed to be a difference between those who merely sample data and provide benchmarks and those who capture much more detailed real-time data.
As he was trying to describe why this was important, he used a word I wasn’t sure I had heard correctly. That word was ‘decode’. In essence what he was saying was that you need to have a forensic understanding of what is happening all the time…and then do something with that information. I’ll come back to the bit about doing something, but I was really struck by the word decode. Too often we make assumptions we know what is going on, or we accept that the information we have is ‘good enough’. You might know that I/O performance on a particular disk array has degraded, but how quickly can you find out what specific micro-service relationships are causing the problem and, will you be able to do that before end-users start to complain about application response times. We need to decode the flood of data we receive in such a way that we can not only react more quickly, but also start to predict issues before they occur.
This is where advances in technology are making life easier. Data Lakes and a wide variety of new cost effective and highly performant storage solutions, provide the ability to be able to ingest and store vast amounts of real-time data from IoT sensors, software and hardware probes. Machine Learning and advanced data analytics solutions combine to give new insights and also begin to predict issues and problems ahead of time.
When the scientists at Bletchley Park cracked the Enigma code, they got the ability not only to react to enemy actions more quickly, but also to start to understand and predict their future plans. Apply this decoding approach to IT Infrastructure environments. Less downtime or performance degradation, more effective capacity management, step changes in energy efficiency. In a digital world where IT is the business and improving IT productivity has a significant impact on the top and bottom line, summoning up your inner Bletchley Park can bring double digit improvement gains on areas you thought were already working optimally. Per second data feeds from multiple sources and machine learning augmented analytics should be on your shopping list for any IPM or data centre energy optimisation projects.
So, thank you Stefan. You can’t trademark the word decode, but I can acknowledge you as my inspiration here.