Don't let cloud slurp all your data. Chew it on the edge, says HPE
Imagine playing office football with a compute rack on castors
Public cloud will become unaffordable for players who reckon the best thing to do with industrial data is shovel it en masse into the white 'n' fluffy stuff, reckons HPE.
Colin I'Anson, HPE's IoT evangelist, spoke to The Register at length earlier this week, mainly to extol the virtues of the firm's edge computing offering, which takes Internet of Things data crunching out of the traditional remote bit barn and instead puts it into your factory.
HPE's edge computing offering is catchily named the Micro Data Centre and was wheeled out in December last year. In I'Anson's words, it "is a box with the power supplies that it needs, it has sufficient cooling inside it, but it can go and stand in any environment you put it in and essentially there is a rack inside and you build into it whatever you want."
The MDC can be fitted with HPE Edgeline boxes, other HPE hyperconverged systems and is billed as "a converged IT and facilities solution for remote deployment", according to the marketing brochure [PDF]. It's basically a rack on wheels, and I'Anson said you can also put "partner solutions" into it.
"People are beginning to understand how our context-aware policies approach ... comes through our Clearpass Universal Profiler, actually allows universal visibility to be put onto the floor, workflow automation, and it gives a whole new way of doing network access control," added I'Anson.
The basic pitch is that HPE's gear can do compute on your shop floor without taking up large chunks of your floorspace, meaning you don't need to splash out on collecting and moving data back and forth, or spend megabucks on cloud services.
HPE are not alone in the edge computing space, however. Back in October Dell teased a similar concept with an identical name, which is much bigger and therefore doesn't come on four handy castors – though it can be left outside in the rain without becoming a literal data lake. It can also swallow up to 96 compute nodes.
Giving compute to the great unwashed "is causing real reaction from customers who are saying 'that's something I want to go and do right now; I want to know how to take the next step.' We've been running into customers who are already intelligent on the edge and beginning to say, 'right, I've done my first experiments, I've learned my first little bit, how do I grow up in this area?'" according to I'Anson.
He wouldn't, however, be drawn on how many of these HPE has sold. Last year the company said the Profiler would start from $5,000 and scale up $50,000, depending how many devices it was monitoring. It did not answer questions as to how much the new hardware and software bundles would cost.
"Strange little things have appeared inside it as well," he continued. "The most popular service coming up is 'how do I have augmented reality in my production environment?'"
Augmented reality is becoming an increasingly popular offering from big tech firms looking to get into the Internet of Things in a big way. As well as covering your factory machinery with sensors, the idea is that your shop floor techies can be equipped with, say, a mobile phone app that puts an overlay on the machine. That overlay then lets operators and maintainers "start to see what has been the problem on this piece of equipment," according to I'Anson. "On to the screen we can do green boxes and red boxes with numbers over the top, depending on where in the machine something has been detected. We can also show them the process by which they can dismantle some of the parts of the machine."
He added: "This part is not working yet but in the future the parts can already be in a van en route to [maintenance staff], so they can do the dismantling while the parts are on their way and they can immediately rebuild and get it up and running that much quicker."
Is this something that the industrial world wants? I'Anson was candid, admitting that some potential customers were "coming back and saying 'actually we've got great processes inside our manufacturing environment already'."
"I did run into a guy who had zero per cent failures in the last year inside his production facilities, so any attempt to tell him we would help him speed up his repairs on a broken-down line didn't work for him," said I'Anson. "But he was super interested in what he could learn about his machines, so that he could start to be much more predictive about when he goes and touches a machine – and particularly when not to touch a machine."
Going back to the comparison between the Dell and HPE micro data centres, HPE's is the physically smaller product. Of the customer who had 100 per cent uptime, I'Anson continued: "He did complain of getting too much data. He said 'I only need to see a very small amount' so he was telling a very good visibility story and he admitted – and he's not the only one who did this – everybody started to say 'I have visibility but not of the things I need'. Although there's the compute to do all the analytics... the customer value on day 1 is going to be about 'can I see it, can it understand what it is, can I take action?'"
With some IoT use cases – for example, Pratt and Whitney's aero engines – generating around 300TB of data per engine per flight, the sheer volume of data that needs to be stored for later analysis is a problem.
"Those large numbers mean that if you can't do the compute, you'll end up with data you can't use and which you're just throwing away," he said. HPE is now "starting to calculate how many racks of equipment everybody's going to need to buy to keep up with this and it's quite interesting to say that, your journey into digitisation, if you're going to introduce the services you want next year, this is the amount of compute you're going to need."
The standard response of "oh, we'll just lob it all into the cloud" is, so I'Anson reckoned, going to become very costly as data volumes grow more and more in the coming years.
"Many are saying 'oh well, I'm just going to put it into public cloud or whatever' and if you start to say 'do you know what your bill is going to be when you do that?'... You know, we've got to stretch our minds actively." ®