ARM server chip upstart Calxeda just bagged $55m in funding last week, and now we know what the company is going to do with the dough: plot a steady course to boost the performance of its ARM processors and the scalability of its on-die integrate Layer 2 distribute switch fabric until there is no reason to buy an x86 server chip …
When the discussion of ARM servers first came up, I didn't think too much of it. Maybe it would work for a niche situation like hosting for hundreds/thousands of tiny, low traffic sites... but not really game for real/significant workloads.
Looking at the roadmap and architecture direction... this could have more potential. Amdahl's law, of course, is still an issue so I can't really see ARM HPC as viable, but MapReduce... maybe?
We have a number of branch sites, some with no more than half a dozen workstations. All of them have some hulking great tower server sitting under a desk making a racket and helping to keep the place uncomfortably warm. Most of them just act as a DC, serve files and printing and the CPU is idling 99% of the time. I've always thought that this would perfect application for a low-power, fanless ARM server.
Microsoft have missed a trick by wasting their ARM licence on Windows RT. They should be porting Windows Server to ARM. Though Intel might have something to say about that..
Re: Small servers
From what little I've heard on it, ARM is out for Server 2012 but hasn't definitively been ruled out for a 2012 R2.
What the heck that means exactly - ex: do we even know if Microsoft is going to release a 2012 R2? - is anyone's guess. I wouldn't hold my breath waiting on it at least. Maybe someday...
In the meantime, ARM NAS devices are about as good as it gets I guess.
I would not rule out ARM HPC. ARM should be at least as good as the PowerPC cores in BlueGene.
You would need to add a much better FPU than Neon and might need to tweak the memory interface but it would still be an ARM. Lots of people are thinking about architectures with lots of GPU like vector units driven by a very small lightweight cpu core. Unless you are Intel IBM or AMD then the only sensible choice for the lightweigh core is ARM.
What they need to work on...
... is common images. In a nutshell they have to create a way to make ARM systems of multiple manufacturers compatible and then promote that to other ARM SoC manufacturers. Nobody in the server business wants to be locked down to one single vendor dependent version of his operating system. What people want, and what is crucial for the market is that everything is compatible.
Some early steps are already done inside the Linux Kernel, but I believe it would be best to have some "BIOS" or "Open Firmware" to bootstrap the operating system so it can probe the hardware and load things from mass storage.
Re: What they need to work on...
True, that is of course why x86 became so popular in the first place - it was nothing to do with absolute performance, and everything to do with being able to run common binaries.
Unless and until a common "ARM Server" platform exists allowing binary compatibility, it'll stay rather niche.
I say ARM Server because the majority of other manufacturers of ARM kit won't benefit much from common platform - STBs, mobiles, and tablets are (mostly) deliberately incompatible with each other, and the majority of users don't have any reason to care.
Re: What they need to work on...
Well a common platform might also be great for STBs, mobiles and tablets. After all just think of the Android update debacle where security fixes only come out if the manufacturer of the device can be assed to provide them.
What we need is the separation of hard- and software. I buy the hardware from one company and the software from another one, and it should just click together. If the mobile phone world would be like this, we'd have progress again.
Linus has just had a rant about this. It's being worked on, and has been quietly worked on for ages anyway. Search for a recent El Reg article featuring the name Dave Rusling.
Meanwhile, what's wrong with redboot or UEFI, depending on target environment?
And in the tablet/desktop and some parts of the consumer equipment market, I guess "whatever Microsoft specified, minus OS lockin" might be a good starting point?
It'll come, and when it does, it'll be good for everybody - vendors, developers, punters. Well, everybody except Intel and Microsoft, but their time at the top is nearly up anyway.