Re: [sage-devel] Re: dream machine ideas
On Fri, Feb 26, 2010 at 7:20 AM, mhampton hampto...@gmail.com wrote: One thing that would be nice is to have a faster machine than t2 running solaris. We have disk.math.washington.edu, which is an 8-core 2.3Ghz opteron with 32GB of RAM, which runs *OpenSolaris*. Also, one could setup a Solaris 10 x86 virtual machine on boxen, which would be another way to get x86 solaris that is way faster than t2. Regarding sparc solaris, there are fast machines on skynet. -- William -Marshall On Jan 22, 1:52 pm, William Stein wst...@gmail.com wrote: Hi, We are considering purchasing a new computer for the sage.math cluster, which will act partly as a Sage notebook server. The budget is about $20-30K (!). If you're a hardware lover, and have been looking at what one could get for that much money (tons of RAM? cores?) these days, let me know. We could just got 1 or 2 computers like sage.math.washington.edu (i.e., 24 2.6Ghz cores,128GBRAM, etc.)... but maybe there is something new or about to come out that we should be aware of? -- William -- William Stein Associate Professor of Mathematics University of Washingtonhttp://wstein.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -- William Stein Associate Professor of Mathematics University of Washington http://wstein.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
Re: [sage-devel] Re: dream machine ideas
On 26 February 2010 15:31, William Stein wst...@gmail.com wrote: Regarding sparc solaris, there are fast machines on skynet. -- William Is there anything quicker than the Blade 2500, which is quite old? The fastest processor that machine could have is 1.6 GHz, which is pretty damm slow by today's standards. Something like an M3000 http://www.oracle.com/us/products/servers-storage/servers/sparc-enterprise/031588.htm with a much faster and modern CPU would be a lot quicker. Perhaps it is not possible, given 't2' was donated, but Sun might accept it in part-ex. In fact, a new M3000 is cheaper than a new T5240! Dave -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
Re: [sage-devel] Re: dream machine ideas
One should definitely look into possibility of buying a BladCenter. The support for InfiniBand and Server RAID, many storage options and highly configurable interfaces make this system an outstanding hardware to work on. There is always an opportunity for hardware diversity also. That makes possible even experiments with hardware. #Serge William Stein wrote: On Sun, Jan 24, 2010 at 7:42 AM, Joshua Herman zitterbeweg...@gmail.com wrote: Man i'm drooling over this thread already. What about some type of blade system like from IBM? http://www-03.ibm.com/systems/bladecenter/ I consider that approach last year (in late 2008). Then, without surprisingly expensive add-ons, blade systems appear like a bunch of separate machines with a (very) fast network. This is more work to manage for a sysadmin, and much more difficult to use for end users. It may be worth looking into this again. william On Sun, Jan 24, 2010 at 4:10 AM, Robert Bradshaw rober...@math.washington.edu wrote: On Jan 22, 2010, at 2:49 PM, Tom Boothby wrote: On Fri, Jan 22, 2010 at 1:48 PM, Simon King simon.k...@nuigalway.ie wrote: Would it be possible to have a faster disk system in *general* (i.e., in the /home part)? I don't know, I am no hardware expert, perhaps NFS==slow. But that would be a nice thing to have. Of course our disk server had other major issues recently, so I don't know if that's a good measure of how well NFS works. While we're dreaming, it would be nice if all the machines had a /scratch, and if we're looking at new ones maybe something better than a usb drive hanging out of the back (perhaps even local raid). Actually, for some applications, sage.math isn't the fastest. Excellent points! I'm definitely in favor of adding a few larger hard drives to the machines, because I really hate our current setup with USB drives cluttering up the rack. With regard to speed... there are faster processors out there, but this comes at the cost of cores. We could run 16 cores at 3.4GHz, which I could get behind. We could push this further, and cut down to 8 cores at 3.7GHz, but I don't think that's a worthwhile tradeoff. I like the idea of having a box with fewer but faster cores for the not-as-parallelizeable tasks. 16/3.4GHz seems like a nice compromise. (What is the speed of a new 24-core setup?) Of course well-used notebook servers like sagenb.org are very parallelizeable. By the way, I am -1 to having just two more identical machines. Why not foster some diversity? Because the hardware is a *dream* to work with. The design is modular with captive screws, the rails snap into the rack and slide out smoothly, the ILOM makes it possible to hard boot / diagnose hardware from a remote location (we can flash the BIOS from anywhere in the world!) I really like being able to switch between machines and use the same binaries, so that's a reason to reduce hardware diversity. - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
Re: [sage-devel] Re: dream machine ideas
On Jan 22, 2010, at 2:49 PM, Tom Boothby wrote: On Fri, Jan 22, 2010 at 1:48 PM, Simon King simon.k...@nuigalway.ie wrote: Would it be possible to have a faster disk system in *general* (i.e., in the /home part)? I don't know, I am no hardware expert, perhaps NFS==slow. But that would be a nice thing to have. Of course our disk server had other major issues recently, so I don't know if that's a good measure of how well NFS works. While we're dreaming, it would be nice if all the machines had a /scratch, and if we're looking at new ones maybe something better than a usb drive hanging out of the back (perhaps even local raid). Actually, for some applications, sage.math isn't the fastest. Excellent points! I'm definitely in favor of adding a few larger hard drives to the machines, because I really hate our current setup with USB drives cluttering up the rack. With regard to speed... there are faster processors out there, but this comes at the cost of cores. We could run 16 cores at 3.4GHz, which I could get behind. We could push this further, and cut down to 8 cores at 3.7GHz, but I don't think that's a worthwhile tradeoff. I like the idea of having a box with fewer but faster cores for the not-as-parallelizeable tasks. 16/3.4GHz seems like a nice compromise. (What is the speed of a new 24-core setup?) Of course well-used notebook servers like sagenb.org are very parallelizeable. By the way, I am -1 to having just two more identical machines. Why not foster some diversity? Because the hardware is a *dream* to work with. The design is modular with captive screws, the rails snap into the rack and slide out smoothly, the ILOM makes it possible to hard boot / diagnose hardware from a remote location (we can flash the BIOS from anywhere in the world!) I really like being able to switch between machines and use the same binaries, so that's a reason to reduce hardware diversity. - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
Re: [sage-devel] Re: dream machine ideas
Man i'm drooling over this thread already. What about some type of blade system like from IBM? http://www-03.ibm.com/systems/bladecenter/ On Sun, Jan 24, 2010 at 4:10 AM, Robert Bradshaw rober...@math.washington.edu wrote: On Jan 22, 2010, at 2:49 PM, Tom Boothby wrote: On Fri, Jan 22, 2010 at 1:48 PM, Simon King simon.k...@nuigalway.ie wrote: Would it be possible to have a faster disk system in *general* (i.e., in the /home part)? I don't know, I am no hardware expert, perhaps NFS==slow. But that would be a nice thing to have. Of course our disk server had other major issues recently, so I don't know if that's a good measure of how well NFS works. While we're dreaming, it would be nice if all the machines had a /scratch, and if we're looking at new ones maybe something better than a usb drive hanging out of the back (perhaps even local raid). Actually, for some applications, sage.math isn't the fastest. Excellent points! I'm definitely in favor of adding a few larger hard drives to the machines, because I really hate our current setup with USB drives cluttering up the rack. With regard to speed... there are faster processors out there, but this comes at the cost of cores. We could run 16 cores at 3.4GHz, which I could get behind. We could push this further, and cut down to 8 cores at 3.7GHz, but I don't think that's a worthwhile tradeoff. I like the idea of having a box with fewer but faster cores for the not-as-parallelizeable tasks. 16/3.4GHz seems like a nice compromise. (What is the speed of a new 24-core setup?) Of course well-used notebook servers like sagenb.org are very parallelizeable. By the way, I am -1 to having just two more identical machines. Why not foster some diversity? Because the hardware is a *dream* to work with. The design is modular with captive screws, the rails snap into the rack and slide out smoothly, the ILOM makes it possible to hard boot / diagnose hardware from a remote location (we can flash the BIOS from anywhere in the world!) I really like being able to switch between machines and use the same binaries, so that's a reason to reduce hardware diversity. - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
Re: [sage-devel] Re: dream machine ideas
On Sun, Jan 24, 2010 at 7:42 AM, Joshua Herman zitterbeweg...@gmail.com wrote: Man i'm drooling over this thread already. What about some type of blade system like from IBM? http://www-03.ibm.com/systems/bladecenter/ I consider that approach last year (in late 2008). Then, without surprisingly expensive add-ons, blade systems appear like a bunch of separate machines with a (very) fast network. This is more work to manage for a sysadmin, and much more difficult to use for end users. It may be worth looking into this again. william On Sun, Jan 24, 2010 at 4:10 AM, Robert Bradshaw rober...@math.washington.edu wrote: On Jan 22, 2010, at 2:49 PM, Tom Boothby wrote: On Fri, Jan 22, 2010 at 1:48 PM, Simon King simon.k...@nuigalway.ie wrote: Would it be possible to have a faster disk system in *general* (i.e., in the /home part)? I don't know, I am no hardware expert, perhaps NFS==slow. But that would be a nice thing to have. Of course our disk server had other major issues recently, so I don't know if that's a good measure of how well NFS works. While we're dreaming, it would be nice if all the machines had a /scratch, and if we're looking at new ones maybe something better than a usb drive hanging out of the back (perhaps even local raid). Actually, for some applications, sage.math isn't the fastest. Excellent points! I'm definitely in favor of adding a few larger hard drives to the machines, because I really hate our current setup with USB drives cluttering up the rack. With regard to speed... there are faster processors out there, but this comes at the cost of cores. We could run 16 cores at 3.4GHz, which I could get behind. We could push this further, and cut down to 8 cores at 3.7GHz, but I don't think that's a worthwhile tradeoff. I like the idea of having a box with fewer but faster cores for the not-as-parallelizeable tasks. 16/3.4GHz seems like a nice compromise. (What is the speed of a new 24-core setup?) Of course well-used notebook servers like sagenb.org are very parallelizeable. By the way, I am -1 to having just two more identical machines. Why not foster some diversity? Because the hardware is a *dream* to work with. The design is modular with captive screws, the rails snap into the rack and slide out smoothly, the ILOM makes it possible to hard boot / diagnose hardware from a remote location (we can flash the BIOS from anywhere in the world!) I really like being able to switch between machines and use the same binaries, so that's a reason to reduce hardware diversity. - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -- William Stein Associate Professor of Mathematics University of Washington http://wstein.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
Re: [sage-devel] Re: dream machine ideas
On Fri, Jan 22, 2010 at 1:48 PM, Simon King simon.k...@nuigalway.ie wrote: Would it be possible to have a faster disk system in *general* (i.e., in the /home part)? I don't know, I am no hardware expert, perhaps NFS==slow. But that would be a nice thing to have. Actually, for some applications, sage.math isn't the fastest. Excellent points! I'm definitely in favor of adding a few larger hard drives to the machines, because I really hate our current setup with USB drives cluttering up the rack. With regard to speed... there are faster processors out there, but this comes at the cost of cores. We could run 16 cores at 3.4GHz, which I could get behind. We could push this further, and cut down to 8 cores at 3.7GHz, but I don't think that's a worthwhile tradeoff. By the way, I am -1 to having just two more identical machines. Why not foster some diversity? Because the hardware is a *dream* to work with. The design is modular with captive screws, the rails snap into the rack and slide out smoothly, the ILOM makes it possible to hard boot / diagnose hardware from a remote location (we can flash the BIOS from anywhere in the world!) --tom -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org