Slashdot Mirror
Sun Releases Starcat
SilentChris writes: "Sun has released the Starcat server, a beast with up to 106 processors running Unix. Anyone have an extra couple [million] bucks lying around?" They're not cheap.
← Back to Stories (view on slashdot.org)
At last a platform to get descent java performance...
Je t'aime Stéphanie
until it shows up on e-bay from a disgruntled former dot bomb employee who five-fingered it from a linux shop which stole BSD code.
Lets remember, that this system is not intended to replace a beowolf cluster of cheap pc's. It is intended to do something that most beowolf clusters can never do: present a single OS image with half a terabyte of memory that any cpu can access at very high speed.
This is a system that is very good at things like fluid dynamics and massive database operations. It is not a good idea if all you want to do is get to the top of the list for the SETI@Home project
Sun themselves are in a need for boxes like that as their website seems severely slashdotted right now.
Your pizza just the way you ought to have it.
The system grows to 106 in the following way:
There are 18 "cpu/memory" boards that hold 4 cpu's each. This brings the system up to a total of 72 cpu's and 576GB of ram.
Now, if you want an server that just does number crunching and dont care about I/O, you can then add 'MaxCPU" modules. Each module holds two additional cpu's (no memory) and occupies the hPCI module slot (a hot swap PCI case that can hold what looks like two to four pci cards). You can use up to 17 of the hPCI module slots to hold MaxCPU modules. (there are 18 pci channels on the system, and at least one must be used for accessing the boot disk).
So there ya have it, 106 cpu's and half a terabyte of ram. I think that in most cases, folks will opt to not use the MaxCPU modules and just stick to the 72 cpu limit.
Dear Santa,
I've been a real good gEek this year. I wrote several white-hat worms to fix IIS holes. I defended IP rights in the Linux kernel. I also mirrored the LOTR trailer.
Could I please get just one little old Starcat Server from Sun? Please make sure it is the 106 processor version with 576 GB of RAM.
I will be real good and use my idle time for SETI.
Your pal,
digital_freedom
P.S. Chocolate chip cookies are your favorite right?
These can be either. It depends on how you configure it. And the fun this is, you can reconfigure it on the fly. You want a cluster in a box? You got it. You want 2 seperate instances of Solaris running, each using 1/3rd the resources of this box, while you pull out the hardware on the rest of the box for maintenance? You got it. This thing is _configurable_. You can hot swap everything except the backplane, pretty much. It's _sweet_.
I'm glad to see that some companies are at least trying to accomplish new things and come out with new products given the state of our economy and markets and such. Even if people think it is overpriced or under-powered and what not that still doesn't degrade the fact that it is a relatively new product in a squishy market. Personally, I own a few hundred shares of SGI stock, but I'm still happy to see any tech comapny suck in their gut, tighten their belt and release a new product in this market. Makes me want to believe the tech markets will turn around sooner than people believe. Kudos to Sun for still working on new products and trying to generally improve things. Now, if Cray would follow suit, I'd be a happy man...
Things you can say to your dog that you can't say to a girl: "How about a nice bone?"
"Wow, look at all the hits we're getting on the Starcat shopping cart! We're going to make a mint on these suckers!"
As someone who does nothing with these types of systems, nor follows them, I think it's great that you can have different processor speeds using "partitions."
I wonder if memory is treated the same way... i.e., separated by "partitions," or if you also have a choice to use it as one, large unified memory resource... or, I wonder if memory can be dynamically partitioned... hmm.
Actually, now that I'm thinking about it... are all of the processor partitions considered peers? I mean, are the partitions all treated as if they were a single processor... then treated equally?
to get on the National US ID Card database bandwagon with Oracle... It'll only need to store about 300 million records with DNA, fingerprint, picture for facial recognition software, key escrow, etc...
try { do() || do_not(); } catch (JediException err) { yoda(err); }
"Huh? I understand that the nation's air traffic controllers may need updated equipment in light of the existing crisis, but how hard can scheduling be? I could see a use for a massively parallel monster like this in, say, flow-through or structural analysis or something, but scheduling? "
What your missing is that this isn't a matter of airtraffic control. This is a matter of determining which planes and crews to fly to which locations at what times to maximize revenue. This is a classic, big, nasty travelling salesman problem. The bigger of a beast of a machine you get, the closer you get to an optimized solution. I.E. Most passengers willing to pay this most money with the least use of resources. It's a huge problem that needs massive computational power.
We may want to forgive Sun for being a bit slow in getting their DB server back up and pretty. A huge chuck of their support staff is helping bale out clients whose data centers got blown up out east. On the other hand, it's your brand new product, you gotta make sure it's available to be bought up. But then again, who in their right mind would just go online and buy one of these? I'd bet ALL of these are sold through meetings between sales reps and IT purchasers.
Um. I believe Starcat Cluster is a trademark owned by the Little Debbie Food Group Inc., Lubbock Texas. It consists of two Star Crunch patties glued together with Mallow Kreeme filling. Then dipped in a Chocolastic sealant.
Just watch out. I hear Debbie has quite a few lawyers.
Let's see...
Sparc 1 box, CPU pretty much part of the MB, unless you get one of those fancy Weitek things. CPU fails, tech replaces MB, no pins to bend.
Sparc 2, Sparc IPX, etc, same story.
Sparc 20, suddenly we have CPUs with pins on them. Coulda happened. Of course, that hasn't been current tech for several years.
UltraSparc line comes out...pressure fittings for the Enterprise servers, no pins to bend. Deskside UltraSparc (like E250, E450), no pins to bend, the CPU is on a card just like Intel does these days.
Ultrasparc III line comes out, big servers don't even use pressure fittings--if you lose a CPU, you get a new system board. Deskside US III (SunBlade 1000) uses a card similar to older deskside units, and has rails to line it up and a torque tool to seat it. Don't see too many bent pins there.
So, apparently you got burned once a very long time ago with a Sparc 20. Don't you think it's time to get over it?
7 November 2006: The day Americans realized corruption and incompetence weren't addressing 11 September 2001
I thought we weren't supposed to use strcat anymore, because it's subject to buffer overflows?
The SGI origin has a ccNUMA architecture, which makes it great for some tasks, ok for others, and awful for yet others. (the trick is to make sure that your particular app falls under the 'great' category)
The sun system is an smp based system, everything connects to a common backplane and each board has equal access to all of the other boards. With the sgi, the speed of accessing memory on the local board or boards in the same cabinet is much faster than hits to memory in remote cabinets.
From what I can tell, Sun is planing on producing a special system board that goes into one of those 18 slots. Thus, with 19 StarCats you can create one big system with 1836 cpu's and 9.7TB of ram. (think of a system in the middle that acts as the center of a star) it will most likely be based on a COMA architecture rather than a ccNUMA. Like the SGI, memory access will depend on the distance between the requesting cpu and the storage location. The difference is that under COMA, if a cpu requests a particular bit of memory a lot, that page is either migrated or copied to a memory bank on that cpu's memory board (so if 5 cpu's all need read only access to the same bit of memory, then they can each have their own copy in a local memory bank. write updates are what make the system a pain in the ass to manage ).
We all know strncat() is better.
Liberty in your lifetime
Just wait for six months. This is the first beast in a series of pseudo-clustered Sunfires. This is roughly a stack of 6800's, and there's going to be a MUCH larger machine released very soon.
"People who do stupid things with hazardous materials often die." -- Jim Davidson on alt.folklore.urban
What are *you* smoking? The Origin 3800 is certainly *not* using the "same back plane a the sun was 2 gens back". In fact, if you were anything other than a hopeless troll, you'd realize that the Origin 3800 doesn't *use* a backplane at all. You get, on a 512p system, 128 I/O channels, each of them supporting up to 12 PCI slots or 4 XIO slots. I can't remember off the top of my head what the bandwidth per channel is but it's on the order of a gigabyte/second (I wanna say it's 1.6 GB/sec, but I might be wrong).
Go Badgers! -- #include "std/disclaimer.h"
Given the way other Sun boxen like the E3500 work, I expect that's the 15K has 18 boards, each of which takes 3 modules, either 2xCPU of 8GB RAM.
That means that 72 CPU / 288 GB memory is 18 boards, each with 2 2xCPU modules and one 18 GB memory module, and the box is full.
Since you always need some memory, the most CPUs you can get is 17 boards w/ 6 each and one with 4. Of course, that leaves you with 8GB of memory for your 106 CPUs.
The other end is (17 x 3 + 1 x 2) 8 GB memory modules for 424 GB on a pair of CPUs
But that's just a guess...
Yes, there's a buy online button. But that's used to get info so one of their sales droids can contact you. It's not like you can slap it on your Visa card. :)
(Disclaimer, I work a lot with E10Ks, so this post is written mostly from my experience with those.)
The 15K is basically just an improvement on the E10K architecture, from what I've seen and heard from Sun's SSEs. The E10K started out life as the Cray SuperServer, and was sold to Sun for a song. It's not architecturally perfect. The E10K is set up to allow individual system boards to be part of domains (aka partitions), which can make for some great scalability in the domains. I've seen tiny little one-system-board domains, and domains with 13 fully populated system boards in them.
One of the major advantages to this platform is the fact that you can hot-swap everything except the centerplane. (Of course, I've never seen a centerplane fail.) The E10K also has Dynamic Reconfiguration, where you can remove system boards from a running domain, but unless your platform is set up in a certain, specific way, this doesn't work as well as advertised. I've personally never used it. The best thing about the E10K is the use of the System Service Processor, which handles all the administrative tasks for the entire cabinet. I've heard that the SSP is now integrated into the 15K, thus eliminating the need for a separate system to perform these tasks and monitoring.
The only thing I've ever seen this class of system used for is data warehousing. No modeling, no graphics rendering, just Oracle databases. Just because it has a large number of processors, doesn't mean they're going to be suitable for every task imaginable. (I used to have a 180MHz Indy R5000, that got 68kkeys/sec in d.net. My 166MMX got something like 350kkeys/sec.) These are workhorse processors, not sports-car style processors.
Though I wonder if Sun's gotten around to fixing that nasty ecache parity error problem with their processors... Having a domain randomly crash because the parity bit on a processor got flipped is no fun when you're dealing with a large production database. I have a feeling that problem will continue to plague them in the 15K.
A good traveller has no fixed plans and is not intent on arriving.