Slashdot Mirror
Sun to Give Niagara Servers to Reviewers
abdulzis writes "Sun Micro's Jonathan Schwartz says that Sun is giving away free servers to bloggers who do a good job reviewing their servers. From the blog article: 'if you write a blog that fairly assesses the machine's performance (positively or negatively), send us a pointer, we're likely to let you keep the machine'" Mr. Schwartz, if you're reading this, feel free to send us one with "Attn: CowboyNeal" on the label.
Sun to Give Nigeria Servers to Reviewers
hello my friend! i am a humble nigerian prince with millions of dollars and have selected you to....
Pinochet used to have this deal for journalists too - if you wrote an article that fairly reviewed the Chilean government, he wouldn't kill you.
paintball
In this day and age of super fast personal computers, what is to differentiate a server from a PC?
Is it the CPU architecture? That can't be the case because many servers run on plain old x86 motherboards.
Is it the OS? While you can say that we can delineate Windows servers into Windows Server and non-Windows Server versions, many places stick Linux on as the OS which blurs the line completely.
Is it the speed? A decade ago, we were looking at servers which weren't half as fast as our low end PCs today. If it is speed, do we have some magical cutoff which just keeps moving forward?
So I get a server from Sun. Does that just mean I get a fast computer with a shitty audio and video card? Limited expansion slots?
I'd rather get a PC.
The difference between a server and a PC is:
- A server is designed to serve data, and has nothing I don't need for it. That means that that damn video card that's not even hooked to a monitor can't break and take my website down with it's million dollars a day revenue.
- A server is designed to serve data reliably, and has enterprise class components. That means no cheap-ass western digital hard drives. If you don't think there's a difference, you've never used Enterprise hardware.
- A server is designed to serve data cheaply. This means low TCO, not low purchase price. Which means an OS that pushes the most bits per cpu, while requiring the least system administrator time. Is Solaris that OS? Debatable, since time has ensured that Apache is highly optimized for Linux. But if you can't run Linux on these yet, you will be able to soon. However, the CPU architecture on these is pretty highly parallel, and Solaris may work better than Linux. Sun is presenting some impressive numbers for these. And they're cheap (as servers go).
In other words, this may be a good time to buy SUNW, at least if you can grow a beard."He who would learn astronomy, and other recondite arts, let him go elsewhere. " -- John Calvin, commenting on Genesis 1
But you wouldn't be able to run World of Warcraft on it...
Please help metamoderate.
I've been doing AIX admin work for years. IBM has long had a program to let people try out their stuff first that they thought was very compelling. Most people wound up buying rs/6000 gear because it simply toasted other unix boxes. IBM actually let a dot com I worked for try out a fully loaded M80 ($250,000), 2 B80s and an F80 and we bought the M80 and 2 B80s because their Java implementation and 64bit copper chips toasted Sun at the time and IBM was willing to put their money where their mouth was... Sun has to be very confident that this will generate much needed postive press and reviews for them.
Contrary to popular belief, Unix is user friendly. It just happens to be particular about who it makes friends with.
What we have here is an eight core CPU, running four threads on a single core. Total power consumption at peak? 80 watts. (CPU only, of course; the system itself will need more -- 220 to 400 watts, depending on the specs.) Clock speed is "only" 1 GHz or so. One floating point unit on the entire chip.
So for scientific work, or other stuff that's seriously hammering the FPU, it's going to be a dog. Sun has never denied this. You're not going to take weather simulations and throw them on this thing; it'd be a waste of money. But for other applications -- database; web server; maybe financial simulations -- there's a hell of a lot of grunt, for very little power consumption.
Sun has effectively opened up a new niche. Anything you have written for Sparc before will still run on this thing, but if you can manage to get a good degree of parallelism in your workload, it will positively fly.
In my opinion (not having seen one of these in action), it's going to be either a massive flop, or a massive win for Sun. My money's on a massive win. They've thought long and hard about common workloads, and have come up with a CPU optimised for those workloads, without too much overhead from making a "general purpose" CPU that can handle anything you throw at it reasonably well. I can't help but wonder how long it will be before we see similar designs out of IBM and Intel.
The other question I have is: what's the IO on these systems like? Poor IO would cripple it, but again, it depends on your workload. The T1000 has a single expansion slot (PCI-E), but four gigabit ethernet ports; the T2000 has three PCI-E and two PCI-X with four gigabit ethernet. On paper, it looks good; time will tell, though, if the systems live up to the expectations.
Before leaving the factory, the platters on every single enterprise class drive receive extensive testing. That is why SCSIs still have a 5 year warranty from Seagate, because every single drive has been tested and meets certain criteria.
In case you weren't aware, Seagate's SATA drives also come with 5 year replacement warranties.
Want to improve your Karma? Instead of "Post Anonymously", try the "Post Humously" option.
There are differences in the hardware of mainframes, unix/as400 servers, pc servers and pc's. Most differences surround reliability (redundancy etc.) and parallel processing (multiple CPU's, multiple specialized processors controlling IO, etc. etc.) Here are a couple examples:
IO controller cache with error correcting checksum in memory and redundant power supply to ensure zero loss of data short of taking a sledgehammer to the thing.
Mainframe CPU - parity checking with automatic transaction rollback on error detection at the hardware level (on the CPU). This is why banks use mainframes, so they know the transaction completed, or didn't, no in between.
You get a LOANER server. At the end of 30 days, you have the option of buying it, or mailing it back, insured, at your expense, or taking the chance they like your bribed-for review. For 99% of the people that read Slashdot, that means you're out $60 bucks. That's a *long* way from getting a free server.
I'm not aware of any manufacturer outside the milspec arena that guarantees to test every component individually.
Modern manufacturing is statistical. You test n components out of each lot of 1000. If more than m fail, the lot is "rejected". In the case of high-cost manufacturing, the "rejected" lot will be individually tested so any good pieces can be salvaged.
If you want tested components, the "grey" refurb/retest units are the ones that have actually been tested. Those which "passed" the lot sampling were not individually tested.
Warranties are also purely statistical. They don't guarantee the drive will actually last that long, they just provide MTBF numbers, figure 24x7 server operation, and that provides the number of years the drive is expected to survive. You still get occasional failures, hence RAID-5/6 storage servers.
I do not fail; I succeed at finding out what does not work.