To hear Intel Fellow Matt Adiletta tell it, Chipzilla not only invented the term microserver but saw the trend towards wimpy computing coming way ahead of the all this fawning over the ARM architecture and a half-dozen upstarts wanting to take big bites out of the Xeon server processor cash cow. When El Reg says "fawn", that's …
Cool writeup, Tim
No-one ever comments because one cannot have fun and silliness with stories about hardware and specs.
"Geez, it just hurts my head to think about all of the opportunities this could provide if we can realize it"
Yeah Andy. Then take a Valium but just stop playing the Überguru. Stuff that computes and is small. Woah, ey!
"it might take an Atom phone"
"it might take an Atom phone and netbook processor wrapped in Xeon drag to repel the onslaught of the ARMed rebels."
It might take a what? How are those Atom phones going?
Meanwhile, for those interested in where the MARKET is going, rather than in the latest Intel press release, go read about several years worth of products based on the "Computer on Module" concept.
Lots of ARM-based product based on lots of ARM chips from lots of ARM licencees, a few PowerPC from a few companies. And for the Windows-dependent (why, in this market?) a handful of Atom ones too.
Now, it's fair to say that CoM cards are not exactly datacentre servers out of the box, but they certainly resemble "a computer on a SIMM" as mentioned in the article. And did I mention that they're mostly not Intel based?
A bit late
"caught the microserver bug back in 2006"
So that means he wasn't paying attention for two years since in 2004 IBM put Blue Gene/L together from a bunch of embedded 440 PowerPC cores running 64 cores and 17 watts per node according to wikipedia. Heck it even piqued my interest back in '05 but hey, welcome to the party.
Re: A bit late
'ere, Otellini, is that an ARMy massing on the ridge? Oh shit, put out a press release.
MicroServers will change the game
Unless you have lived with MicroServers for a while, it is easy to miss the nuances and potential of their power. They are more than just small, powerful boxes. They are a new way at looking at computing. What would you do with a server that handles thousands of users and is the size of a chemistry text book and draws less than a car headlight on high beam? One answer - http://www.billionnodecloud.com
And there are many, many more applications just waiting to be discovered. In its own way, the entire MicroServer phenomenon is actually exciting because it opens up entirely new vistas of computing applications. Want a quiet, energy sipping, multi-node cloud platform on your desk that runs dozens of virtual machines ? http://www.crkit.info
Totally shameless promotion I must admit, but MicroServers will reduce energy consumption in huge ways if allowed to and I would think green techies would be all over that societally advantageous application of MicroServer technology. The electricity cost savings and reduction in coal and gas consumption could be substantial in the aggregate. How many old 1,000 watt servers are there out there sucking up electricity when they could be replaced with 50 Watt boxes that may have more compute power than what they replace? My guess is millions. That means billions of watts are being wasted right now because a lot of the MicroServer technology exists today.
Re: MicroServers will change the game
millions of those were already virtualized years ago - that's the way most folks will continue to operate. Micro servers are only useful if you have a specific workload that is targeted at them, they are not flexible. You can't say "oh I need more cpu cores" or memory or whatever.
Right now I have a bunch of servers each one has about 30VMs on it, and each server draws about 240W - or roughly 8W/VM (server). Even with 30VMs on it the physical host sits at around 15% cpu utilization (roughly 170GB of memory in use). This is conservative - in a 500W physical server I could easily quadruple the # of VMs. That gets me down to 4W/virtual server.
The world already has a way to save power and that way is virtualization.
Micro servers is just going backwards in most cases - unless again you are able to tailor your workload to run within the constraints of those limited systems. It's sort of like the Amazon cloud way of thinking - not flexible enough, leads to lower resource utilization.
But if your a facebook or a tweeter, or something like that and really know your workload then a microserver can be a good solution - for the rest of us though - x86-64 & virtualization will reign supreme for the foreseeable future.
Gotta be careful though - wouldn't want to deploy 10s of thousands of new systems only to find out a year later that they need a memory upgrade for your latest software feature and there's no way to upgrade them.
Re: MicroServers will change the game
Nate, doesn't that depend on the end point being prepared to pay for the virtualization "tax" and the ever increasing layers of management software that comes with it?
I'm not disagreeing with your assertion, it's just we've seen the same story play out before when mainframe virtualization ruled the roost. Ultimately is was the software cost/stack that precipitated the move, not the hardware cost, capacity or virtualization.
Re: MicroServers will change the game
1-2 watts is easily possible. I have 8-9 servers running on an i7T using 7-11 watts most of the time. I have cut my power consumption to 150 w for two serers running a farm that used to fill a rack and use about 8000w. Thats is what microservers are about and you are right the more people use them the better.
Take an empty Xeon BGA package
Stick in an Atom CPU
Add a 25W power resistor
Badge at as XEON-LTE
Gradually each year increase the resistor thus demonstrating a lower power, reduced TCO, and enhanced green credentials
shoot the proof-reader
"...holding his hands in his head..."
Re: shoot the proof-reader
Shooting seems a bit extreme. As the writer, I can tell you I have a bit of word poisoning after a year of this, so you might as well shoot me, too. I think perhaps some coffee would work better than bullets.
I mean, we don't shoot programmers when their code has a bug. Not yet, anyway.
Re: shoot the proof-reader
"I mean, we don't shoot programmers when their code has a bug. Not yet, anyway."
If we did the Redmond campus would be a slaughterhouse.
Where AMD can win
They have abandoned ideological ties to x86. They're still committed to what makes money, but they have no embarrassment over punting ARM.
More to the point, they are lower-price players which means they have no cash cow they will try to protect by hobbling the low-end.
Its interesting to see the focus on microservers (rather than phones/tablets) where power consumption is just a little less critical and more oomph might be desirable.
My main concern is that microservers are still not cheap enough. For consumers we need RAID and enough disk-to-network power to saturate a couple of gigabit nics. We need a database for mythtv / amarok.
I'd like to see stuff come out on PCIe cards with SATA ports, external power and NICs - let's keep those costs down!
Re: Where AMD can win
"They have abandoned ideological ties to x86. They're still committed to what makes money, but they have no embarrassment over punting ARM."
It's ironic to see this giving AMD an edge now, considering how Intel tried to leave x86 behind while plugging Itanium - only to have that same AMD come along and eat their lunch by offering x86-64! Back then, of course, the small server market really needed "x86, but 64 bit", so they could run and port their legacy x86 environments and applications over; these days, for mobile and server applications, x86 isn't even advantageous let alone essential.
A shame Microsoft tried to give x86 an artificial edge in Windows 8, by restricting ARM to the feeble subset 'RT' rather than a full-blown port. (Yes, regular Windows-ARM software would still be hard to find - but that's no reason to ban it outright!)
If we are looking at low power server class computing how about a Feb 2004 "mini-cluster"
Very impressive with the consumer tech available at the time!
" stuff .... on PCIe cards with SATA ports, external power and NICs"
"I'd like to see stuff come out on PCIe cards with SATA ports, external power and NICs"
Computer on Module was mentioned earlier but the comment didn't say in detail what interfaces the COMexpress standards (from the PICMG standards people) supports. You've mentioned a list of interfaces you want, but not mentioned a form factor (other than "on a PCIe card"). Do you just need PCIe bus capability, or PCIe size and shape too?
You might want to have a look at a few CoMexpress products.That's express as in PCI express. They're not the usual size of a PCIe card though. Some of them support all your requested interfaces electrically, but depending on your needs may require a bit of adaptation to get the connectors exactly per your requirements. You could see whether the CoM formfactor could be adapted to meet your requirements. You can even get Atom-based CoM cards, if your requirements still need legacy Wintel compatibility.
If COMexpress is way too expensive (and it might be at the moment), there's plenty similar or lighter stuff without the COMexpress seal of approval.
"let's keep those costs down!"
Get the volume up and (ideally) get off legacy Wintel. Today, COMexpress modules are relatively low volume products premium priced for the industrial/commercial market. Like it or not, Raspberry Pi shows what you can get for not much money when the volumes are right and you don't need legacy Wintel.
[minor update acknowledging "on a PCI card"]
micro server is not new
There were the low power MIPS severs (64bit, 6 core, 10w and the features ARM promises to have in their coming chips) back to 2006. So, I don't think the micro server is new concept but just a new name.
The question is if the MIPS server which had everything micro server promise to have (power, program support) was unable to catch the market, why the renamed ARM/ATOM server would be a Success?
Re: micro server is not new
Because in 2006 picking ARM meant using a processor for cell phones with that weird Linux stuff to run our business? That sounds like a career limiting risky move.
Today using an Arm and Linux means getting off that legacy COBOL-era Wintel platform into the bright new dawn of Android, iThingy, Cloudy goodness makes me look like a cutting edge dynamic thrusting rockstar.
Big, massive, blinking FAIL
The Atom are a complete failure, not even good for netbooks and this Adiletta guy is able to say with a straight face that it's going to be great for microservers? Yeah sure. Call him the Romney of the IT industry.
Mushroom cloud because of the atom bomb, obviously.
The problem is not the hardware
It's the software. One would need something like the "BIOS" on PCs. Something allowing you to boot your operating system without having to build a special kernel. The kernel needs to be able to discover the hardware.
That is the current problem with ARM.
Re: The problem is not the hardware
Does "device tree" mean anything to you? I'm guessing not.
- NASA boffin: RIDDLE of odd BULGE FOUND on MOON is SOLVED
- SOULLESS machine-intelligence ROBOT cars to hit Blighty in 2015
- BuzzGasm! Thirteen Astonishing True Facts You Never Knew About SCREWS
- Worstall on Wednesday YES, iPhones ARE getting slower with each new release of iOS
- Microsoft's Euro cloud darkens: Redmond must let feds into foreign servers