When DCIM first arrived, expectations and enthusiasm were high. Yet despite early industry exuberance, first generation solutions failed to meet customer expectations and enthusiasm waned. At one point, data centre infrastructure juggernaut Schneider Electric, themselves an early DCIM investor, considered pulling the plug on DCIM completely. But in May last year, the company returned with Ecostruxure IT Advisor, touted as a “next-gen” DCIM that addresses previous customer pain points and accommodates today’s realities of distributed and hybrid IT. Is it time for DCIM to shine? At Data Center Dynamics London in November, Techerati’s deputy editor James Orme spoke to Kevin Brown, SVP Innovation and CTO, Secure Power Division at Schneider Electric, to discuss the changing face of DCIM
James Orme: How has the DCIM conversation changed in recent years?
Kevin Brown: If you reflect on traditional DCIM, you pretty much have to say it failed to meet expectations. You go back a few years ago and there are people saying it’s going to be anywhere from a $3 to a $7 billion market. It’s probably more like a $700 million market. If you take a good number of the installations, there are studies that show they didn’t meet customer expectations.
Candidly, two or three years ago, we were looking at it as one of these vendors and saying, “Hey, is this an area that we should even continue to invest in? Is it even relevant in this new hybrid environment” Our conclusion was, if you took a look at the hybrid environment, the challenges that DCIM was trying to address still existed, and there’s probably more. For instance, how to get visibility into the thousands of these little edge and micro data centres is a significant problem.
Ultimately, to service and manage the ecosystem, DCIM needed to be much more open than maybe it was in the past. How do you evolve from something that was built for an enterprise data centre then get that to apply in this hybrid environment with the cloud and the edge that exists today?
JO: Are there any other reasons why you think it failed to meet expectations?
KB: Well, part of it is that the market moved. Right about the time DCIM was starting to hit maturity, it was going after enterprise data centres. Then the market started moving towards colo and cloud. Also, there were three key pain points: DCIMs were too hard to get started with, they were very complex and very difficult to maintain. We started looking and saying, “Well, we’ve got to solve those three problems.”
JO: That leads us to Schneider’s “next-gen” DCIM. What are the main pillars of your next-gen DCIM?
KB: What realised we needed a new architecture if we wanted to make this thing easy to use and address the customer problems. We needed a next-generation. We start calling it the next generation and we found that everybody else was using the term “next generation.” It’s a fairly common term. We just released a whitepaper that defines the five attributes we think next generation DCIM needs to have.
First of all, you’ve got to leverage cloud technologies in order to get the scalability and usability. Second, you need to be connected with a data lake so that you can start taking advantage of analytics. To do artificial intelligence and machine learning, you need a big amount of data. Getting that data into a data lake is really the only practical way that you can do it today. The third thing was, it had to be designed for mobile and ease of use.
Today, I swear to God, I had a customer come in and they pulled up their phone, they had the EcoStruxure IT app running. He was telling me about how he was away in Mexico City and was able to deal with the problem while in Mexico City. Traditional DCIM was not architected to do that. You need to think about mobile first.
The second thing is thinking about simplicity as the core of the design. We were having conversations internally when DCIM was rolling out that were like, “Well, look, you have to understand, this is complex software. It’s enterprise. It requires training, it requires consultants to come in and install it.” That was basically the business model that I think many vendors built. For the most part, people don’t tolerate that anymore for this type of product. It’s got to be as easy to use as your phone.
Then, the last thing that we identified — this really was learning that came in the last few months when we started looking at the data collected into our data lake — was that our customers were having a big problem with compliance and cyber security. There’s basically a couple of things we can show customers: Is your software updated? Is your firmware up to date? What protocols do you have enabled on a device that shouldn’t be enabled? And what devices can’t meet your cyber security policies that you have in place? This is something that we’re just rolling out now.
For the next generation DCIM to be credible in the market, you probably need those five attributes for it really to be considered next generation.
JO: When did you realise Ecostruxure IT Advisor’s potential to assist with cyber security and compliance?
KB: We picked up on it when we went into the cloud and started collecting all this data on the devices. That was not an area of focus for DCIM. It was straightforward for us because we had this cloud architecture and we had some data scientists looking at the data and they said, “Holy cow, look at this problem”. That was not part of the original plan, but all of a sudden, we pivoted towards it.
So we prototyped some stuff then we got a guy that helps us who’s a security expert. When he went and deployed it on his own network, he found 50 devices that had protocols enabled that he wasn’t aware of. This is a guy who does this for a living. He’s a professional. Even he had things missing.
JO: How deep is AI and machine learning integrated at present?
KB: For now, we’re rolling out some battery failure prediction algorithms that we’ve put in place. They are looking at not only when do we think the battery is going to fail but what might be causing it. Is it because of the temperature? Is it because of cycling? Is it just because it’s old?
You might be able to see in the future, “Hey, I’ve got this one that looks like it’s failing faster than the others.” That’s great to know, but actually, the algorithms and the data has allowed us to say “why”. Was it being driven because of temperatures? Was it being driven because of cycles? You can do different actions depending on what you’re learning from that. We’re going to keep rolling out new algorithms.
Ultimately, we’re doing a lot of research around the normal operating circumstances where problems are experienced, and what they may be correlated with. We’re doing not only just predicting when the failure is going to happen but also analysing when you might experience a problem that you shouldn’t expect. This is really where the promise of machine learning and AI comes.
I think we’re still in the early days. This isn’t going to be like one day I wake up and I’ve got it, and yesterday I didn’t. It’s going to evolve. We’re going to keep getting better and keep rolling out more and more sophisticated algorithms. What I can tell you is that it takes a fairly significant investment to get to where we are now. Now, we’re getting the fruits of the labour, so to speak.
We’ve identified three major categories when you talk about analytics and algorithms and machine learning. One is around just the predictive stuff. When are things going to fail, when aren’t they going to fail, and so forth? The second one is more in the category of operations management. You have to think about, “What can I do? How can I identify when a human error may have occurred?” The third one is around efficiency. Google, a few years ago, put out that they developed a machine learning algorithm and it reduced the waste on their HVAC system by 40 percent. That’s another category. What we’re doing is looking at all three of those and saying where are the areas that we have the biggest impact.
JO: Which area are you most excited about?
KB: An area that we’re very interested in is how do we make the Edge more efficient? How do we make sure it’s being managed properly? How about the whole life cycle of how do we make sure that when something fails, it’s lithium-ion batteries or lead-acid batteries? How do we get them back? Are things being operated in a way that extend their life as long as possible? I can tell you, just as part of Schneider, that’s an area that we’re highly focused on and we think the industry needs to be focused on as well.
JO: Do you think this has been something that’s not been touched on enough in discussions about the Edge?
KB: I would argue it’s barely been touched on at all. I think this is the next big challenge for the industry. I know of one retailer, they’ve got a data centre I believe is 8 MW. If you add up all its retail stores, it’s 35 MW. If you actually look at the IT energy consumption, it far exceeds what’s happening inside the data centre. My guess is that math holds for everything.
There is no PUE for Edge environments that’s meaningful. There is no measurement for energy consumption or whether you’re consuming more than your neighbours do. Are you typical or atypical? Every month, I get a report from my utility that says, “Here’s how much energy you consume compared to other households like yours.” Those are some of the benchmarking that I think you’ll see us drive on as time moves forward, particularly around the Edge. Data centres, for the most part, the hyperscalers in particular, they’re highly focused on this issue of efficiency. They have been for a long time. You look at the Edge, it’s a vacuum at this point.
JO: Moving on to data safety and security, we talked about the fact that EcoStruxure leverages cloud and data lakes. What are some of the concerns that your customers have raised in regards to security and how is Schneider working to address them?
KB: Well, there’s no shortage of security concerns. Some of it is just around, is the data being done securely? What’s the encryption that you’re using? How are you doing multifactor authentication? What are the systems? Then, the second category is, what are you doing with the data? Is it my data, your data, GDPR? What’s machine data versus personal data? How do I make sure I can get rid of the personal data if requested? What’s the audit trail to verify that you actually did that?
These are all new things that come when you go into a cloud environment. What we’ve been experiencing is for some of the customers who are using the tool within their own company, they need help dealing with their cybersecurity guys. It’s not unusual to have a 200-page document that gets handed to you asking cybersecurity questions. We’ve done a lot of work. We have very good answers to all of these questions, but really, we want to enable our customers to go and answer their questions they’re getting internally, and sometimes, these are 200-page documents. That’s a new part of the business model. We’ve done a good job adjusting to it, but it’s an adjustment for our customers as well.
JO: Are you seeing any particular sectors in which EcoStruxure IT Advisor is getting more traction than others?
KB: We when we first developed [EcoStruxure IT Advisor] we were targeting people who had Edge environments that they were not managing. We’re getting a good uptick in those types of customers, whether with universities, schools in general, and retailers.
It’s not like any one specific segment, but it’s people who have distributed environments that they aren’t managing today. This is a great tool for them to be able to get this under control very easily. Then, as we’re developing the algorithms and everything else, now, we’re seeing the big guys get more interested. The traditional customers are seeing things like the compliance tool, they’re starting to see the prediction models that we’re developing, and they want to be able to leverage that as well.
JO: Obviously you’ve not solved all the challenges yet. What challenges haven’t you solved? And what’s on Schneider’s roadmap for next year?
KB: One thing that we’re working on pretty aggressively at this point is trying to make liquid cooling more practical for people to be able to deploy. If they could look at energy efficiency, not just energy efficiency but energy consumption, liquid cooling on its own, we have some models that indicate that it could be a 10 percent to 15 percent energy reduction for the data centre.
There’s not that many technologies out there that could give us that kind of an increase, but so far, our view has been most of the liquid cooling technologies out there, they work, but they’re not designed in a way that they could be deployed at scale very easily. Here’s where we’ve announced our partnership with Iceotope and Avnet where we’re working towards, how can I take and get a much more practical, deployable solution that addresses a lot of the challenges that we’ve seen with liquid cooling in the past because we see this as a really great technology to help drive better performance, lower energy consumption.
We’re excited about the progress we’re making. Again, we made the announcement a few weeks ago at our Innovation Summit with Avnet and Iceotope. You’re getting the key pieces together, the ecosystem. We need a few other players coming in, but I’m pretty excited about the direction it’s moving. Avnet brings an important part to it because they bring in the system integration capabilities that we didn’t have. We bring in the ability to make this thing real in terms of supply chain and supplier. Iceotope’s got great technology. I think you’ll see some more interesting stuff from us before the end of the year.