HPE CEO pledges $4bn Edge R&D splurge Hewlett Packard Enterprise will make a US$4bn bet on edge computing, CEO Antonio Neri confirmed at the Discover CIO conference in Las Vegas today. Neri, who took over the business in February, pointed to fresh data* from analyst firm Gartner that 75 per cent of data will be created and processed outside of the traditional data …
Yeah, good luck with that. What do you want for edge computing? Great performance with low power in a small form factor, so you don't have to send as much data back to the cloud.
What is modern HP completely incapable of making? Anything that isn't a bloated enterprisey hot mess. Sorry, your edge router's going to need a Xeon, 32GB of RAM, and a 512GB SSD just to run HP's drivers and management suite.
This sounds an awful lot like what happened in the 80s & 90s. Instead of having dumb terminals talking to the central mainframe, you'd have smart PCs that could add more value to the data in the mainframe.
Having gone back to the old dumb client/central server setup (Cloud & web browser*) we're now heading back to the fat client.
* If the web browser were truly dumb, we'd probably avoid a lot of the security problems with them.
Imagine, data processed and consumed at the point of necessity...jeez
Fast food cloud data tends to get expensive and unhealthy.
Not everything you need for your business requires a cloudy excursion to be effective.
That said, can HPE facilitate a workforce to do such things? Doing R&D hasn't been an HP thing for decades.
They (not just HP, but the IT industry as a whole) are always going to sell the "next big thing" and it will always be something that's a totally different direction than what they were selling you a few years ago.
"Don't buy standalone servers anyone, you want blade servers"
"Don't set up servers dedicated to a task, you want a private cloud"
"Don't ship all your raw data to the cloud, you need edge computing to do the heavy lifting before it reaches your cloud"
I wonder if next they'll try to tell people that you should have servers dedicated to a task so you can be sure critical tasks get sufficient performance, and we'll be back at standalone servers again...
I'm certain that they can't possibly pull it off. Not enough money especially as they'd need to buy at least two firms. They've also serious lack of in-house talent in two of the required areas of knowledge and I've doubts on two more. This is seriously cross-cutting domains and it's going to make them bleed.
As I'd rather not provide them with a business plan, I'll leave it at that.
Yes and no, having seen what happens in large cooperates you have a pool of talent that thinks a certain way and can no longer innovate. In that instance your best bet maybe to layoff the dinosaurs and start again, preferable by moving location and keeping them at arms length and staff them with new employees who don't get encumbered with existing culture. Best thing would be to not even brand it as HPE.
The problem with most company purchases that are made for innovation reasons is that the bigger organisation tries to impose their outdated policies instead of leaving the smaller more nimble company as they are.
"Edge computing, for the uninitiated, puts compute and storage capacity closer to where data is created, so it can be processed locally. Doing so avoids the cost of shipping data to a cloud and improves response times too. If an edge system spots something worthy of more attention, it will ship that data to a core cloud so that more processing power can be brought to bear."
Doesn't this just mean a local data centre? Like what we always used to have?
Joking about HP aside for a moment, there's a bit of difference in theory and intent.
The intent, basically, is that the edge computing devices only contain transitory data, while all permanent data is still in the cloud.
You want edge computing when there's a firehose of locally generated data, like you've got hundreds of sensors hooked up all over your buildings generating video, temperature, proximity, etc etc. What you're trying to do is avoid having to send every byte of data back to the cloud and only send it the *interesting* data.
Or let's say you're doing facial recognition on that video - sending all the streams of video back to the cloud and having the cloud tag faces is silly. Maybe you'd LIKE to have all the video from all the cameras, but if you don't have the NSA's budget you need to make some tradeoffs. So you'd have local machines which are configured from cloud data then just tell the cloud which people they see and video only for Persons of Interest. Then you keep the streams local for a month before deleting.
This isn't all that different from things we've seen before, but it is different from a local data center in that the local data center is intended to be The Canonical Repository, but now that's the cloud. And if you have a bunch of local data centers that coordinate to be the canonical repository, well that's a cloud.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
