Re: [ccp4bb] Crystallographic computing platform recommendations?
Axel: Um... it works fine for us. I just tested it about 10 seconds ago (names have been changed to protect the guilty): [EMAIL PROTECTED] ssh [EMAIL PROTECTED] Password: macclient:~ myaccount$ pwd /Network/Servers/osxserverbox.some.domain/afp/myaccount macclient:~ myaccount$ echo $HOME /Network/Servers/osxserverbox.some.domain/afp/myaccount This is OS X Server 10.5.5 and a 10.5.5 client, with an extremely vanilla configuration. After fighting the setup for two days to try to get it to mount somewhere *other* than /Network/Servers/blahblahblah, I left it alone and put in the following link on the client (we only have one, but could have more without any problems, as far as I can tell): /people - /Network/Servers/osxserverbox.some.domain/afp And: macclient:~ myaccount$ df -h FilesystemSize Used Avail Capacity Mounted on /dev/disk0s2 465Gi 24Gi 441Gi 6%/ devfs107Ki 107Ki0Bi 100%/dev fdesc1.0Ki 1.0Ki0Bi 100%/dev map -hosts 0Bi0Bi0Bi 100%/net map auto_home 0Bi0Bi0Bi 100%/home map -fstab 0Bi0Bi0Bi 100% /Network/Servers trigger0Bi0Bi0Bi 100% /Network/Servers/osxserverbox.some.domain/afp afp_05oXza0007GeoMVU-1.2d3a 298Gi 82Gi 216Gi28% /Network/Servers/osxserverbox.some.domain/afp I think this homedir mounting feature is a function of using OSX Server, rather than vanilla OS X, to do the AFP exporting, which is why we are using Server. Best, -- Steve Lane System, Network and Security Administrator Doudna Lab Biomolecular Structure and Mechanism Group UC Berkeley On Tue, Nov 18, 2008 at 10:51:44PM -0800, Axel Brunger wrote: Brian, There is one disadvantage with using AFP rather than wide open (potentially insecure) NFS mounts. Remote login via ssh into a client computer won't by default mount the user's AFP home directory. While it is possible to manually mount the AFP home directory it may preclude other users from using the client computer from the console. This feature of AFP is due to user-specific mounting of the remote disk on the client computer. I assume the same feature would apply to Kerberized NFS mounts, but I haven't tried it. This limitation of AFP requires some thought when using idle client computers as compute servers. We're using the Mac Server Xgrid service, along with the freely available GridStuffer.app application to make submission of batch jobs to all our Macs relatively easy. Axel On Nov 18, 2008, at 9:10 PM, Brian Mark wrote: Francis, From your response and others to my question about OS X server 10.5, AFP seems to be the preferred networking protocol over NFS. Yes, in our case the RAID is connected to a G5 (via firewire 800 - which provides surprisingly good transfer rates BTW) that is running OS X server 10.5 . I'll try AFP for the user home directories. Thanks, Brian Axel T. Brunger Investigator, Howard Hughes Medical Institute Professor of Molecular and Cellular Physiology Stanford University Web:http://atbweb.stanford.edu Email: [EMAIL PROTECTED] Phone: +1 650-736-1031 Fax:+1 650-745-1463
Re: [ccp4bb] Crystallographic computing platform recommendations?
Hi Folks, Appropos to the comments below - if the folk who bundle CCP4 in this way would be kind enough to send the packages to me, I can test them the same way as we test our own binary builds, to make sure they work correctly. Saves me trying to search them out on the web. Many thanks, Graeme -Original Message- From: CCP4 bulletin board [mailto:[EMAIL PROTECTED] On Behalf Of Donnie Berkholz Sent: 17 November 2008 22:41 To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] Crystallographic computing platform recommendations? When choosing a Linux distribution, look for one with crystallographic software available in a package repository so you don't have to waste time dealing with that. No matter how convenient a software project like CCP4 tries to make it to install by hand, it will never integrate with a system as well as packaging designed to do so. Ubuntu, Fedora/RHEL, and Gentoo all have crystallographic packages available somewhere. From that point, it's just a question of what you're most comfortable with. -- Thanks, Donnie Donnie Berkholz Developer, Gentoo Linux Blog: http://dberkholz.wordpress.com
Re: [ccp4bb] Crystallographic computing platform recommendations?
Hi 1. Mac vs. Linux (which flavor?) vs. Windows Ubuntu - just works(TM). Gentoo - install only if know vi :-) 2. Graphics cards any NV quadro series. Buy at least one year old card so that linux will have drivers. Intel integrated cards also work well and have good drivers (Coot/ono shows good performance if you have 1.5GB system RAM). 3. Displays Budget dependent. 4. Processors - multiple processors, multiple cores? Speed? I would suggest a quad core Xeon or Opteron (with Debian - you need rock solid) that will be a workhorse (where I put images also) to do refinement jobs that users can ssh -X into. make sure the directories in the remote machine are available for exports via NFS/autofs. For local workstations (use ubuntu to get the bling) any AMD/Intel Core2 series would be OK (that way my office would be quiet and cool). Even I can hibernate to save power. Regards, Karthik About half of what I do involves ~1.0 A X-ray structures - data processing, rebuilding in Coot, refinement, and so forth - so my current desktop (Optiplex GX745, Radeon X1300) machine drags on graphics sometimes. I don't seem to need stereo these days, for what it's worth. Anybody have suggestions or specs they'd like to share? Thanks in anticipation of your advice. Regards, Anna Gardberg My workstations and data servers: 1. Linux (Fedora Core 8, currently, as this distro is well supported in xtallography circles--I'm tempted to try Ubuntu next upgrade cycle) 2. If stereo is not an issue, any 8000 or 9000 series GeForce NVidia card is fine. I avoid ATI because of flaky Linux drivers. The 9600GT Nvidia is a bargain, but an 8600 is OK, too. 3. I use a 24 Samsung LCD display. You need CRTs only if doing stereo, but then you would need a really expensive Quadro card with stereo outputs. 4. Quad cores are so cheap right now, there is no reason not to go this route. I built my workstations from really cheap Intel DG35EC boards with Intel Q9300 processors. 250Gbyte drive for the OS and programs, and a 400 Gbyte drive for data, all SATA-300. I loaded up with a memory card reader, SATA DVD-writer, and packaged it all up in a tiny cubical micro-ATX box. You can build the whole workstation for under $1500, including the LCD monitor. My data and backup servers are installed on junk machines salvaged from offices and labs. Anything with a Pentium 4 is fast enough to be a data or backup server. The nice thing about building your own computer is when it comes time for an upgrade, you can save many of the parts (case, power supply, DVD writer, etc.) and just upgrade the motherboard, CPU, and maybe the hard drives.
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Tue, 2008-11-18 at 09:40 -0500, James M. Vergis wrote: 2) Graphics Cards: I like the Nvidia cards since they provide linux drivers. ATI also does now. I would say the only thing I don't like about them is when you do a kernel update you have to remake the driver. I would just add that for those using Fedora, you can avoid having to remake the driver by using the RPMFusion repositories to install the Nvidia drivers. This way updating the kernel automatically installs the corresponding Nvidia kernel module. The same applies to AMD/ATI drivers.
Re: [ccp4bb] Crystallographic computing platform recommendations?
Hi I would tend to go with a system type that you're familiar and happy with, rather than buying into someone else's idea of the best system. So if you are used to running any particular type (Windows, OSX, Linux), I'd spend a little time setting it up on what you've currently got and see if you can do what you want, and where you want it to perform better. Of course, you get a much wider range of available hardware if you go for Linux or Windows (unless you really want to build a Hackintosh). Most (or at least enough to be able to process data solve structures) of the software out there has been ported to the popular platforms, and it's likely that in the short term at least, that you'd be more productive in not having to learn a new environment, with all its idiosyncrasies... Dear list, I haven't seen the crystallographic computing platform thread come up for a while, and I've got a chance to upgrade my desktop to a workstation, so I thought I'd ask the CCP4BB for advice on: 1. Mac vs. Linux (which flavor?) vs. Windows 2. Graphics cards 3. Displays 4. Processors - multiple processors, multiple cores? Speed? About half of what I do involves ~1.0 A X-ray structures - data processing, rebuilding in Coot, refinement, and so forth - so my current desktop (Optiplex GX745, Radeon X1300) machine drags on graphics sometimes. I don't seem to need stereo these days, for what it's worth. Anybody have suggestions or specs they'd like to share? Thanks in anticipation of your advice. Regards, Anna Gardberg Harry -- Dr Harry Powell, MRC Laboratory of Molecular Biology, MRC Centre, Hills Road, Cambridge, CB2 2QH
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Nov 18, 2008, at 16:26, William G. Scott wrote: On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Same here. Lets also say that I have seen many people in the lab switch from Linux and Windows to OSX, but nobody from OSX back to Windows or Linux. That must be saying something. We still use btw Linux workstations for stereo and I am happy with them as well. As Harry says, choose what you are most comfortable with(*). A. (*) If thats Windows Vista though, take a break and meditate, think of the bigger picture of life, donate to the Bill Gates foundation which is what Bill Gates should be remembered for, and find the nearest MacStore. ;-) Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
They're all great! - Linux offers limitless possibility with its concomitant complexity and chaos. - Mac OS X offers the design integrity, consistency, and efficiency of centralized control. - Windows guarantees lowest common denominator functionality for a rock bottom price. So why not buy hardware software that can run all three? ...with native-like performance. ...simultaneously. -Original Message- From: CCP4 bulletin board on behalf of William G. Scott Sent: Tue 11/18/2008 7:37 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] Crystallographic computing platform recommendations? On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
After having dealt, over the years, with several dozens of 'crystallographic computing platforms' and having setup and maintained quite a few myself, I would recommend to not be cheap. I would recommend to go with well supported hardware and OS. For linux, I would recommend a commercial solution, and the hardware could come from a vendor such as Dell. In our lab, we use Macs practically exclusively, except for a few legacy Linux boxes. I don't think it is worth saving a few hundred dollars when you end up spending/wasting so much time down the road assembling and fixing the machine as well as trying to keep up with the latest OS patches and drivers. I'd rather spend my time doing something else than being a computer support person. I realize I am not using the latest, greatest, pimped-out number-crunching monster, but a quad-core Mac is plenty sufficient. I like the fact that a refinement takes a few minutes longer, because that gives me time to fetch a cup of coffee or chat with a colleague. Just a thought. Best - MM Dear list, I haven't seen the crystallographic computing platform thread come up for a while, and I've got a chance to upgrade my desktop to a workstation, so I thought I'd ask the CCP4BB for advice on: 1. Mac vs. Linux (which flavor?) vs. Windows 2. Graphics cards 3. Displays 4. Processors - multiple processors, multiple cores? Speed? About half of what I do involves ~1.0 A X-ray structures - data processing, rebuilding in Coot, refinement, and so forth - so my current desktop (Optiplex GX745, Radeon X1300) machine drags on graphics sometimes. I don't seem to need stereo these days, for what it's worth. Anybody have suggestions or specs they'd like to share? Thanks in anticipation of your advice. Regards, Anna Gardberg Mischa Machius, PhD Associate Professor Department of Biochemistry UT Southwestern Medical Center at Dallas 5323 Harry Hines Blvd.; ND10.214A Dallas, TX 75390-8816; U.S.A. Tel: +1 214 645 6381 Fax: +1 214 645 6353
Re: [ccp4bb] Crystallographic computing platform recommendations?
I've gone off computers completely after finding that if I stare hard enough at my diffraction pattern I can sketch the density by hand :-)! Actually it's OSX for me because I view myself as a scientist who uses computers rather than a computer expert who uses science. Linux is just that bit too complicated and windows still lacks function. Simon On 18 Nov 2008, at 16:09, Warren DeLano wrote: They're all great! - Linux offers limitless possibility with its concomitant complexity and chaos. - Mac OS X offers the design integrity, consistency, and efficiency of centralized control. - Windows guarantees lowest common denominator functionality for a rock bottom price. So why not buy hardware software that can run all three? ...with native-like performance. ...simultaneously. -Original Message- From: CCP4 bulletin board on behalf of William G. Scott Sent: Tue 11/18/2008 7:37 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] Crystallographic computing platform recommendations? On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Tue, 18 Nov 2008, Michel Fodje wrote: On Tue, 2008-11-18 at 09:40 -0500, James M. Vergis wrote: 2) Graphics Cards: I like the Nvidia cards since they provide linux drivers. ATI also does now. I would say the only thing I don't like about them is when you do a kernel update you have to remake the driver. I would just add that for those using Fedora, you can avoid having to remake the driver by using the RPMFusion repositories to install the Nvidia drivers. This way updating the kernel automatically installs the corresponding Nvidia kernel module. The same applies to AMD/ATI drivers. The same is true for SUSE distros: you don't have to recompile the NVIDIA drivers (or the ATI ones, AFAIK) on a kernel upgrade as long as you have enabled the correct package repository for the driver. Regards, Peter. -- Peter Keller Tel.: +44 (0)1223 353033 Global Phasing Ltd., Fax.: +44 (0)1223 366889 Sheraton House, Castle Park, Cambridge CB3 0AX United Kingdom
Re: [ccp4bb] Crystallographic computing platform recommendations?
We also use OSX almost exclusively for our virus crystallography projects, for the same reasons given by Bill and Anastassis. Every now and then we have to spend time re-compiling some programs with larger parameters or up to 64-bit, so don't want the added hassle of tinkering with the hardware. More expensive, yes, but cost savings in other areas, namely time and maintenance. Also provides wide software compatibility between productivity/office and command line applications on a single machine. The Macs have good 64-bit hardware and OSX 10.5 runs 64-bit on the command line - the 32-bit and 64-bit programs run together transparently. Currently having fun with an 8- core Mac Pro using 64-bit multi-threaded CNS v1.21. We can now run a capsid refinement in just under 50 minutes (uses all 8 cores and 3Gb of RAM), a remarkable speed increase. Many thanks to the CNS authors and contributors, and all the other software authors for OSX compatibility and support . Linux would work fine as well, but for the reasons given OSX on the Macs has been a highly productive solution for our needs. Cheers, Jeff On Nov 18, 2008, at 7:42 AM, Anastassis Perrakis wrote: On Nov 18, 2008, at 16:26, William G. Scott wrote: On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Same here. Lets also say that I have seen many people in the lab switch from Linux and Windows to OSX, but nobody from OSX back to Windows or Linux. That must be saying something. We still use btw Linux workstations for stereo and I am happy with them as well. As Harry says, choose what you are most comfortable with(*). A. (*) If thats Windows Vista though, take a break and meditate, think of the bigger picture of life, donate to the Bill Gates foundation which is what Bill Gates should be remembered for, and find the nearest MacStore. ;-) Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
Hi all, With all the talk about Mac OS X, I've not heard much mention about OS X Server and networking Macs together. Is anyone using the 10.5 server and LDAP to centrally house user directories on a RAID connected to a Mac server for example? We had this setup running with 10.4, but we now seem to have issues getting the server 10.5 to export the RAID to the clients via NFS. I know NSF is old, but our network is behind a firewall and it has worked well for us. It seem if you don't set things up the way Apple recommends (using AFP for example), things can get difficult... It's quite possible that we are missing something during the setup, but frankly I'm surprised at how difficult it has been to work with OS X server. I would be happy to hear about other people's experience with OS X server 10.5. Perhaps a website describing OS X server for crystallographic computing/networking would be nice if it doesn't exist already. Cheers, Brian = Brian L. Mark, MSc, PhD Assistant Professor Department of Microbiology Room 418, Buller Building University of Manitoba Winnipeg, Manitoba CANADA R3T 2N2 Phone (204) 480-1430 Fax (204) 474-7603 Web: http://www.umanitoba.ca/science/microbiology/staff/mark/ On 18-Nov-08, at 9:42 AM, Anastassis Perrakis wrote: On Nov 18, 2008, at 16:26, William G. Scott wrote: On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Same here. Lets also say that I have seen many people in the lab switch from Linux and Windows to OSX, but nobody from OSX back to Windows or Linux. That must be saying something. We still use btw Linux workstations for stereo and I am happy with them as well. As Harry says, choose what you are most comfortable with(*). A. (*) If thats Windows Vista though, take a break and meditate, think of the bigger picture of life, donate to the Bill Gates foundation which is what Bill Gates should be remembered for, and find the nearest MacStore. ;-) Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Nov 18, 2008, at 8:47 AM, Jeff Speir wrote: We also use OSX almost exclusively for our virus crystallography projects, for the same reasons given by Bill and Anastassis. Every now and then we have to spend time re-compiling some programs with larger parameters or up to 64-bit, so don't want the added hassle of tinkering with the hardware. More expensive, yes, but cost savings in other areas, namely time and maintenance. Also provides wide software compatibility between productivity/office and command line applications on a single machine. The Macs have good 64-bit hardware and OSX 10.5 runs 64-bit on the command line - the 32-bit and 64-bit programs run together transparently. MacOS X.5 runs 64-bit GUI apps as well.* The CLI/GUI split that you mention was a limitation of MacOS X.4. Ian *Marketing link: http://www.apple.com/macosx/technology/64bit.html Currently having fun with an 8-core Mac Pro using 64-bit multi- threaded CNS v1.21. We can now run a capsid refinement in just under 50 minutes (uses all 8 cores and 3Gb of RAM), a remarkable speed increase. Many thanks to the CNS authors and contributors, and all the other software authors for OSX compatibility and support . Linux would work fine as well, but for the reasons given OSX on the Macs has been a highly productive solution for our needs. Cheers, Jeff On Nov 18, 2008, at 7:42 AM, Anastassis Perrakis wrote: On Nov 18, 2008, at 16:26, William G. Scott wrote: On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Same here. Lets also say that I have seen many people in the lab switch from Linux and Windows to OSX, but nobody from OSX back to Windows or Linux. That must be saying something. We still use btw Linux workstations for stereo and I am happy with them as well. As Harry says, choose what you are most comfortable with(*). A. (*) If thats Windows Vista though, take a break and meditate, think of the bigger picture of life, donate to the Bill Gates foundation which is what Bill Gates should be remembered for, and find the nearest MacStore. ;-) Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
I have zero experience with OS X server, but there were many changes (mostly for the better) in NFS with 10.5. I have put some information here, and as this is a wiki, I would be grateful for more contributions... http://sage.ucsc.edu/xtal/wiki/index.php/NFS On Nov 18, 2008, at 8:57 AM, Brian Mark wrote: Hi all, With all the talk about Mac OS X, I've not heard much mention about OS X Server and networking Macs together. Is anyone using the 10.5 server and LDAP to centrally house user directories on a RAID connected to a Mac server for example? We had this setup running with 10.4, but we now seem to have issues getting the server 10.5 to export the RAID to the clients via NFS. I know NSF is old, but our network is behind a firewall and it has worked well for us. It seem if you don't set things up the way Apple recommends (using AFP for example), things can get difficult... It's quite possible that we are missing something during the setup, but frankly I'm surprised at how difficult it has been to work with OS X server. I would be happy to hear about other people's experience with OS X server 10.5. Perhaps a website describing OS X server for crystallographic computing/networking would be nice if it doesn't exist already. Cheers, Brian = Brian L. Mark, MSc, PhD Assistant Professor Department of Microbiology Room 418, Buller Building University of Manitoba Winnipeg, Manitoba CANADA R3T 2N2 Phone (204) 480-1430 Fax (204) 474-7603 Web: http://www.umanitoba.ca/science/microbiology/staff/mark/ On 18-Nov-08, at 9:42 AM, Anastassis Perrakis wrote: On Nov 18, 2008, at 16:26, William G. Scott wrote: On Nov 18, 2008, at 7:01 AM, Mischa Machius wrote: For linux, I would recommend a commercial solution, For linux, I actually ditched the commercial solution for Ubuntu, because it was vastly easier as a non-expert to maintain. Having said that, like you, I have found running Mac OS X to be the most cost-effective in terms of time and utility. Same here. Lets also say that I have seen many people in the lab switch from Linux and Windows to OSX, but nobody from OSX back to Windows or Linux. That must be saying something. We still use btw Linux workstations for stereo and I am happy with them as well. As Harry says, choose what you are most comfortable with(*). A. (*) If thats Windows Vista though, take a break and meditate, think of the bigger picture of life, donate to the Bill Gates foundation which is what Bill Gates should be remembered for, and find the nearest MacStore. ;-) Bill
Re: [ccp4bb] Crystallographic computing platform recommendations?
I followed Kay's advice (after deciding that I knew better, of course, but we won't elaborate on that :-) and I am very pleased AND have had no trouble (knock on wood) to get everything working just fine. We make sure we have both 64 bit and 32 bit libraries and so far everything has worked out of the box, no hassle. And that includes coot. Mark -Original Message- From: Kevin Cowtan [EMAIL PROTECTED] To: CCP4BB@JISCMAIL.AC.UK Sent: Tue, 18 Nov 2008 3:21 am Subject: Re: [ccp4bb] Crystallographic computing platform recommendations? And I would give exactly the opposite advice, unless you are or have a guru who can devote time to fixing all the little things which still don't work under 64 bit OSs.? ? (Does anyone else have any clues on why 64-bit compiled coot can't calculate a map? I need to look into it, but have a huge backlog of work at the moment.)? ? Kay Diederichs wrote:? Dear Anna,? you didn't ask about that, but I would definitely recommend a 64bit operating system.? My specific recommendations are mostly in the articles Computer_hardware and CentOS, to be found under the more general topic Xtal_computing of the CCP4 wiki (http://strucbio.biologie.uni-konstanz.de/ccp4wiki/index.php/Xtal_computing) HTH,? Kay? Anna S Gardberg schrieb:? Dear list,? I haven't seen the crystallographic computing platform thread come up for a while, and I've got a chance to upgrade my desktop to a workstation, so I thought I'd ask the CCP4BB for advice on:? ? 1. Mac vs. Linux (which flavor?) vs. Windows? 2. Graphics cards? 3. Displays? 4. Processors - multiple processors, multiple cores? Speed?? ? About half of what I do involves ~1.0 A X-ray structures - data processing, rebuilding in Coot, refinement, and so forth - so my current desktop (Optiplex GX745, Radeon X1300) machine drags on graphics sometimes. I don't seem to need stereo these days, for what it's worth.? ? Anybody have suggestions or specs they'd like to share? Thanks in anticipation of your advice.? ? Regards,? Anna Gardberg?
Re: [ccp4bb] Crystallographic computing platform recommendations?
For future proofing your computing platform , I would go with a 64 bit system. You will not regret it. All crystallography applications live very happily on a 64bit platform . There were a few teething troubles , but nothing an email to ccp4bb , phenixbb , coot or pymol newsgroups cant fix. We recently upgraded all our computers to 64bit Here we use two Mac Pros running Leopard OSX -64bit , after a few teething problems we are at equilibrium. Our third machine is a 64bit dual - quad- core , 16GB workstation from Appro Linux with a nvidia graphics card that they installed for us during the build and it does stereo ( Quadro 3450 bought from ebay for $250 OR Quadro 4600 newer version for a whole lot more $2300-major overkill) . It runs Ubuntu Hardy-Heron 64 bit , installed from downloaded iso image and almost minimal tweaking. Since most projects release builds for ubuntu , I am partial to Ubuntu . But Gentoo is also a great distribution. Just my two cents Hari On Tue, Nov 18, 2008 at 12:15 PM, [EMAIL PROTECTED] wrote: I followed Kay's advice (after deciding that I knew better, of course, but we won't elaborate on that :-) and I am very pleased AND have had no trouble (knock on wood) to get everything working just fine. We make sure we have both 64 bit and 32 bit libraries and so far everything has worked out of the box, no hassle. And that includes coot. Mark -Original Message- From: Kevin Cowtan [EMAIL PROTECTED] To: CCP4BB@JISCMAIL.AC.UK Sent: Tue, 18 Nov 2008 3:21 am Subject: Re: [ccp4bb] Crystallographic computing platform recommendations? And I would give exactly the opposite advice, unless you are or have a guru who can devote time to fixing all the little things which still don't work under 64 bit OSs. (Does anyone else have any clues on why 64-bit compiled coot can't calculate a map? I need to look into it, but have a huge backlog of work at the moment.) Kay Diederichs wrote: Dear Anna, you didn't ask about that, but I would definitely recommend a 64bit operating system. My specific recommendations are mostly in the articles Computer_hardware and CentOS, to be found under the more general topic Xtal_computing of the CCP4 wiki ( http://strucbio.biologie.uni-konstanz.de/ccp4wiki/index.php/Xtal_computing) HTH, Kay Anna S Gardberg schrieb: Dear list, I haven't seen the crystallographic computing platform thread come up for a while, and I've got a chance to upgrade my desktop to a workstation, so I thought I'd ask the CCP4BB for advice on: 1. Mac vs. Linux (which flavor?) vs. Windows 2. Graphics cards 3. Displays 4. Processors - multiple processors, multiple cores? Speed? About half of what I do involves ~1.0 A X-ray structures - data processing, rebuilding in Coot, refinement, and so forth - so my current desktop (Optiplex GX745, Radeon X1300) machine drags on graphics sometimes. I don't seem to need stereo these days, for what it's worth. Anybody have suggestions or specs they'd like to share? Thanks in anticipation of your advice. Regards, Anna Gardberg -- Instant access to the latest most popular FREE games while you browse with the Games Toolbar - Download Now!http://pr.atwola.com/promoclk/10075x1212904500x1200818240/aol?redir=http://toolbar.aol.com/games/download.html?ncid=emlweusdown0004
Re: [ccp4bb] Crystallographic computing platform recommendations?
I completely agree with Marius. Our (my) constraints are not $500 in price difference, but the fact that I maintain a system for scientific computing AND an X-ray system AND I am expected to be a scientist who publishes and writes grants. Thus, our approach has been to automate and minimize my time and effort spent. We buy out-of-the-box Dell systems WITH support. This has already paid off once when (for hard to explain reasons) a motherboard went out. I diagnosed it, called India, and Dell sent someone out to replace the motherboard and with minimal effort we were back in business. Our Dell boxes came with standard NVIDIA quadro graphics cards. Hardware-wise our only real problem is the non-existance of CRTs that can do stereo and the exhorbitant prices for the flat screens. We run CentOS because (I think) RH is impossible to deal with, but I do like the stability. We have shared, network attached file storage? (which backs up automagically and is RAIDed) and we authenticate off a Windows server via LDAP, so I do not need to maintain user records, someone else already does that and all I need to do is e-mail a note that person X (or Y , gender neutral) can have access to the Linux systems and voila work is done. The system creates everything needed. It took a LOT of effort to get where we are. We thought long and hard about what exactly we needed and why. The most difficult operation is currently to get data from synchrotrons to our system. All the firewalls (existing for good reasons, of course) make it so hard these days to do anything useful in data transfer. We use Rsync and then I manually archive raw image data away. We sometimes have dreams about distributed computing (not so much for crystallography but for small angle scattering). Not enough experience to tell you how it should be done well. Mark -Original Message- From: Mischa Machius [EMAIL PROTECTED] To: CCP4BB@JISCMAIL.AC.UK Sent: Tue, 18 Nov 2008 8:01 am Subject: Re: [ccp4bb] Crystallographic computing platform recommendations? After having dealt, over the years, with several dozens of 'crystallographic computing platforms' and having setup and maintained quite a few myself, I would recommend to not be cheap. I would recommend to go with well supported hardware and OS. For linux, I would recommend a commercial solution, and the hardware could come from a vendor such as Dell. In our lab, we use Macs practically exclusively, except for a few legacy Linux boxes.? ? I don't think it is worth saving a few hundred dollars when you end up spending/wasting so much time down the road assembling and fixing the machine as well as trying to keep up with the latest OS patches and drivers. I'd rather spend my time doing something else than being a computer support person. I realize I am not using the latest, greatest, pimped-out number-crunching monster, but a quad-core Mac is plenty sufficient. I like the fact that a refinement takes a few minutes longer, because that gives me time to fetch a cup of coffee or chat with a colleague.? ? Just a thought.? ? Best - MM? ? ? Dear list,? I haven't seen the crystallographic computing platform thread come? up for a while, and I've got a chance to upgrade my desktop to a? workstation, so I thought I'd ask the CCP4BB for advice on:? ? 1. Mac vs. Linux (which flavor?) vs. Windows? 2. Graphics cards? 3. Displays? 4. Processors - multiple processors, multiple cores? Speed?? ? About half of what I do involves ~1.0 A X-ray structures - data? processing, rebuilding in Coot, refinement, and so forth - so my? current desktop (Optiplex GX745, Radeon X1300) machine drags on? graphics sometimes. I don't seem to need stereo these days, for what? it's worth.? ? Anybody have suggestions or specs they'd like to share? Thanks in? anticipation of your advice.? ? Regards,? Anna Gardberg? ? ? ? Mischa Machius, PhD? Associate Professor? Department of Biochemistry? UT Southwestern Medical Center at Dallas? 5323 Harry Hines Blvd.; ND10.214A? Dallas, TX 75390-8816; U.S.A.? Tel: +1 214 645 6381? Fax: +1 214 645 6353?
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Tue, 18 Nov 2008 18:04:11 + Kevin Cowtan [EMAIL PROTECTED] wrote: From a system management point of view there is one very significant benefit to Ubuntu: The LTS releases come out every 2 years and are supported for 3 years. Compare this with Fedora: releases are only supported for 18 months. The comparable distribution to LTS is centos or the official rhel releases, which are on a 7 year support cycle. The standard ubuntu releases follow almost the same 18 month cycle as fedora. -Tim -- - Tim Fenn [EMAIL PROTECTED] Stanford University, School of Medicine James H. Clark Center 318 Campus Drive, Room E300 Stanford, CA 94305-5432 Phone: (650) 736-1714 FAX: (650) 736-1961 -
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Nov 18, 2008, at 10:12 AM, Tim Fenn wrote: On Tue, 18 Nov 2008 10:21:29 + Kevin Cowtan [EMAIL PROTECTED] wrote: And I would give exactly the opposite advice, unless you are or have a guru who can devote time to fixing all the little things which still don't work under 64 bit OSs. (Does anyone else have any clues on why 64-bit compiled coot can't calculate a map? I need to look into it, but have a huge backlog of work at the moment.) Try -fno-strict-aliasing, or dropping the optimization to -O0 or -O1. Fixed the exact same problem for me. I also noticed several of the m4 macros coot uses (mmdb/ssm/guile-gtk) are broken such that 32 and 64 bit libraries can get mixed up, so you may not be using a 64 bit binary. Should you decide to actually go fix the code, one of the flags that I use religiously for 64-bit code is -Wshorten-64-to-32. It flags implicit conversions from 64- to 32-bit types. This catches two types of problems: Inadvertent truncation of 64-bit types (pointers, size_t, etc.) cause various correctness/crashing issues random intermixing of single and double precision operations: e.g. float x, y = x + 1.0; // 1.0 is double precision, and will force x to be up converted to double, a double precision add is then done, then downconverted to float these are a performance problem if you use XMM to do single and double precision arithmetic (as MacOS X does). Also, at least on my system, -fstrict-aliasing will cause a warning to be emitted when you have an aliasing violation. Hopefully that should help you find these things and clean up the code. Is the author of coot still working on the project? Ian
Re: [ccp4bb] Crystallographic computing platform recommendations?
Dear list, Thank you all for the helpful and thoughtful advice about crystallographic computing systems. Before I get into the (first day's) summary, I have a followup question: Has anyone worked with refinement or coot rebuilding of 1.0 Ang. structures on an iMac? If so, what was your experience of it? Here is a summary of the advice so far: OPERATING SYSTEM 1. There was nearly unanimous agreement that Windows isn't really an option. Since I feel antipathy for Vista anyway, that part of the decision will be easy! 2. While many replies suggested 64-bit operating systems, especially if RAM exceeds 4 GB, many also warned of problems I might face compiling coot and CCP4 under a 64-bit OS, and their solutions seem to require more Linux-fu than I currently possess. However, it seems that compiling programs as 32 bit will work fine under a 64-bit OS, and can be recompiled as 64-bit (with lots of cursing and asking for advice, presumably) when an application demands it. 3. For ease of maintenance, Ubuntu Linux was generally recommended over others, but there were several Red Hat/Fedora proponents, as well as support for other Linux systems (Linuces?). 4. Many recommended MacOSX for least-hassle use by a non-guru. I must say, this sounds blissful, for I have lost many days of my life to the hassles of stressful software installation. HARDWARE 1. Many of the suggestions involved constructing systems from components, a feat beyond my current skill and inclination. Thus, I will purchase a pre-built system. 2. If going with OSX, there was disagreement about the necessity for a Mac Pro; some responders thought a high-end iMac would be fine. Has anyone worked with 1.0 A structures on an iMac? 3. Eddie Snell made the point that most crystallographic software cannot take advantage of multiple processors yet. Still, as some people pointed, it is often convenient to try several refinement options at the same time, so there would be greater convenience with more processors. GRAPHICS/ DISPLAY 1. NVidia graphics cards were recommended over ATI. There was concern that Linux drivers would not be available for the very newest ones. 2. Some recommended dual-monitors, to give you a place to put all of your open windows.
Re: [ccp4bb] Crystallographic computing platform recommendations?
Hi Anna: I think the main thing to emphasize in summary is that any of the options various people have suggested would be far more than adequate to do refinement or model building with coot. In my limited personal experience, unless you have something as big as the ribosome, you won't be forced to resort to 64-bit OS. I have the slowest first- generation iMac at home, and that is more than adequate for anything we have done. Computers simply aren't rate-limiting anymore, so finding an operating system and hardware that you are comfortable with should probably be your main priority. Bill William G. Scott Contact info: http://chemistry.ucsc.edu/~wgscott/ On Nov 18, 2008, at 10:41 AM, Anna S Gardberg wrote: Dear list, Thank you all for the helpful and thoughtful advice about crystallographic computing systems. Before I get into the (first day's) summary, I have a followup question: Has anyone worked with refinement or coot rebuilding of 1.0 Ang. structures on an iMac? If so, what was your experience of it? Here is a summary of the advice so far: OPERATING SYSTEM 1. There was nearly unanimous agreement that Windows isn't really an option. Since I feel antipathy for Vista anyway, that part of the decision will be easy! 2. While many replies suggested 64-bit operating systems, especially if RAM exceeds 4 GB, many also warned of problems I might face compiling coot and CCP4 under a 64-bit OS, and their solutions seem to require more Linux-fu than I currently possess. However, it seems that compiling programs as 32 bit will work fine under a 64-bit OS, and can be recompiled as 64- bit (with lots of cursing and asking for advice, presumably) when an application demands it. 3. For ease of maintenance, Ubuntu Linux was generally recommended over others, but there were several Red Hat/Fedora proponents, as well as support for other Linux systems (Linuces?). 4. Many recommended MacOSX for least-hassle use by a non-guru. I must say, this sounds blissful, for I have lost many days of my life to the hassles of stressful software installation. HARDWARE 1. Many of the suggestions involved constructing systems from components, a feat beyond my current skill and inclination. Thus, I will purchase a pre-built system. 2. If going with OSX, there was disagreement about the necessity for a Mac Pro; some responders thought a high-end iMac would be fine. Has anyone worked with 1.0 A structures on an iMac? 3. Eddie Snell made the point that most crystallographic software cannot take advantage of multiple processors yet. Still, as some people pointed, it is often convenient to try several refinement options at the same time, so there would be greater convenience with more processors. GRAPHICS/ DISPLAY 1. NVidia graphics cards were recommended over ATI. There was concern that Linux drivers would not be available for the very newest ones. 2. Some recommended dual-monitors, to give you a place to put all of your open windows.
Re: [ccp4bb] Crystallographic computing platform recommendations?
The main differences aren't cosmetic, but rather free accessibility of updates and distributions, as well as attitude aka philosophy. Advantages of Ubuntu include frequent OS updates, debian-based package management system (which admittedly is a matter of taste, but it is one of the original package management systems), an active user-based support community, extremely good internationalization support, and a major financial commitment to keeping everything free and open: http://www.ubuntu.com/community/ubuntustory/foundation from the Canonical corporation, headed by Mark Shuttleworth, a South African anti-racist entrepreneur who manages to defy the cynics: http://rixstep.com/2/1/20060712,00.shtml Any software that runs on RH/Fedora will run on Ubuntu and vice versa, so that should not be an issue. I'm running Xubuntu by the way (xfce4), but I also have the options of logging into KDE or gnome (Kubuntu and vanila Ubuntu) as well as other window managers. For me, installing vanilla ubuntu first and then xfce4 worked fine. I made the transition by installing on a new partition (new disk) and booting into it, back when Breezy was introduced. I never looked back. I've put some more propaganda here: http://sage.ucsc.edu/xtal/wiki/index.php/Scientific_Computing_on_Ubuntu Bill On Nov 18, 2008, at 9:23 AM, Andreas Förster wrote: Re: Switching from Fedora to Ubuntu. Don't. Stick with Fedora if it works for you. If you're tempted by Ubuntu because of the coolness, the bling, the color, the limited customizability or the lack of root, a Mac will get you all of this much more abundantly. If you're used to Redhat/Fedora, stick with it. Everything crystallography related works on it. I tried to set an ancient Vaio up with xUbuntu and went back to Redhat 9 after too much frustration. Things are not where they're supposed to be and don't work as they should. This is from a Redhat perspective, obviously, which I've been using since 5.2. Andreas Edward Snell wrote: That would have saved me a lot of time and frustration! Now if only I can remember this for the next time JI have a 64bit Fedora 9 system running on an Lenovo Thinkstation D10 – you can buy this without the operating system. I used the 64bit option as I’m processing very high resolution X-ray data and needed to use more memory. I was able to get a 16 GB system for a reasonable cost. If you are a tinkerer and want a few days of ‘fun’ this is a way to go. If not, commercially available Linux systems are available but I found few with a large memory capacity. Coot and CCP4 were a pain to compile in 64 bit mode (not the developers problem, just mine for trying to get them to work in 64 bit). I had to play with some header definitions, lop h’s off etc, add using namespace std and, install a lot of libraries but they now work well. Coot reads all the svg files and displays all the pretty icons. I like the PC Linux route mainly because the hardware is less expensive and easier to swap out components. I am used to Fedora and manage my own system so for the moment I plan to stick with that but it sounds like Ubuntu looks like an interesting way to go. Has anyone had any experiences in transitioning from Fedora to Ubuntu? Is it something to think about, a Thanksgiving project mainly? Cheers, Eddie Edward Snell Ph.D. Assistant Prof. Department of Structural Biology, SUNY Buffalo, Hauptman-Woodward Medical Research Institute 700 Ellicott Street, Buffalo, NY 14203-1102 Phone: (716) 898 8631 Fax: (716) 898 8660 Email: [EMAIL PROTECTED] Telepathy: 42.2 GHz Heisenberg was probably here! *From:* CCP4 bulletin board [mailto:[EMAIL PROTECTED] *On Behalf Of *Michel Fodje *Sent:* Tuesday, November 18, 2008 10:28 AM *To:* CCP4BB@JISCMAIL.AC.UK *Subject:* Re: [ccp4bb] Crystallographic computing platform recommendations? On Tue, 2008-11-18 at 09:40 -0500, James M. Vergis wrote: 2) Graphics Cards: I like the Nvidia cards since they provide linux drivers. ATI also does now. I would say the only thing I don't like about them is when you do a kernel update you have to remake the driver. I would just add that for those using Fedora, you can avoid having to remake the driver by using the RPMFusion repositories to install the Nvidia drivers. This way updating the kernel automatically installs the corresponding Nvidia kernel module. The same applies to AMD/ATI drivers. -- Andreas Förster, Research Associate Paul Freemont Xiaodong Zhang Labs Department of Biochemistry, Imperial College London
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Nov 18, 2008, at 11:41 AM, Anna S Gardberg wrote: 3. Eddie Snell made the point that most crystallographic software cannot take advantage of multiple processors yet. Still, as some people pointed, it is often convenient to try several refinement options at the same time, so there would be greater convenience with more processors. I've used phenix to calculate iterative build composite omit maps. I've also used phenix's autobuild after autosol option to do some automated building at around 2.0A. Tom Terwilliger said that scoring multiple heavy atom sites with AutoSol will be supported sometime in the future. All of these methods benefit with multiple cores. I personally would invest in a multicore system, or better yet a small cluster. Nobody here has talked about RAID. When it comes to integrating images, the speed up to using a mirrored setup or a raid 5 array is quite noticeable than just a single drive. Cheers FR - Francis Reyes M.Sc. 215 UCB University of Colorado at Boulder gpg --keyserver pgp.mit.edu --recv-keys 67BA8D5D 8AE2 F2F4 90F7 9640 28BC 686F 78FD 6669 67BA 8D5D
Re: [ccp4bb] Crystallographic computing platform recommendations?
On Nov 18, 2008, at 9:57 AM, Brian Mark wrote: Hi all, With all the talk about Mac OS X, I've not heard much mention about OS X Server and networking Macs together. Is anyone using the 10.5 server and LDAP to centrally house user directories on a RAID connected to a Mac server for example? We run 10.5 server with OpenDirectory (LDAP for Apple) over AFP (no raid) with 10.5 clients. We had this setup running with 10.4, but we now seem to have issues getting the server 10.5 to export the RAID to the clients via NFS. I know NSF is old, but our network is behind a firewall and it has worked well for us. It seem if you don't set things up the way Apple recommends (using AFP for example), things can get difficult... It's quite possible that we are missing something during the setup, but frankly I'm surprised at how difficult it has been to work with OS X server. Is the RAID directly connected to the server? i.e. does OS X server host the nfs server or is it a unix machine? the requirements for specifying an nfs home dir are tricky. Check the Admin reference volume 1 book. I would be happy to hear about other people's experience with OS X server 10.5. Perhaps a website describing OS X server for crystallographic computing/networking would be nice if it doesn't exist already. As I said above we have leopard clients bound to an OSX server using OD and AFP home directories. All the clients are imaged to the same version of OSX Leopard and crystallography software (managed by fink mostly). If there's a prudent need for upgrading, I have a bootable firewire drive that I load onto one of the clients and do the updating, then image the software to the server and use Netboot/ Netinstall to distribute the image to the clients. Any more questions let me know FR - Francis Reyes M.Sc. 215 UCB University of Colorado at Boulder gpg --keyserver pgp.mit.edu --recv-keys 67BA8D5D 8AE2 F2F4 90F7 9640 28BC 686F 78FD 6669 67BA 8D5D
Re: [ccp4bb] Crystallographic computing platform recommendations?
Brian, There is one disadvantage with using AFP rather than wide open (potentially insecure) NFS mounts. Remote login via ssh into a client computer won't by default mount the user's AFP home directory. While it is possible to manually mount the AFP home directory it may preclude other users from using the client computer from the console. This feature of AFP is due to user-specific mounting of the remote disk on the client computer. I assume the same feature would apply to Kerberized NFS mounts, but I haven't tried it. This limitation of AFP requires some thought when using idle client computers as compute servers. We're using the Mac Server Xgrid service, along with the freely available GridStuffer.app application to make submission of batch jobs to all our Macs relatively easy. Axel On Nov 18, 2008, at 9:10 PM, Brian Mark wrote: Francis, From your response and others to my question about OS X server 10.5, AFP seems to be the preferred networking protocol over NFS. Yes, in our case the RAID is connected to a G5 (via firewire 800 - which provides surprisingly good transfer rates BTW) that is running OS X server 10.5 . I'll try AFP for the user home directories. Thanks, Brian Axel T. Brunger Investigator, Howard Hughes Medical Institute Professor of Molecular and Cellular Physiology Stanford University Web:http://atbweb.stanford.edu Email: [EMAIL PROTECTED] Phone: +1 650-736-1031 Fax:+1 650-745-1463
Re: [ccp4bb] Crystallographic computing platform recommendations?
Anna S Gardberg wrote: Dear list, I haven't seen the "crystallographic computing platform" thread come up for a while, and I've got a chance to upgrade my desktop to a workstation, so I thought I'd ask the CCP4BB for advice on: 1. Mac vs. Linux (which flavor?) vs. Windows 2. Graphics cards 3. Displays 4. Processors - multiple processors, multiple cores? Speed? About half of what I do involves ~1.0 A X-ray structures - data processing, rebuilding in Coot, refinement, and so forth - so my current desktop (Optiplex GX745, Radeon X1300) machine drags on graphics sometimes. I don't seem to need stereo these days, for what it's worth. Anybody have suggestions or specs they'd like to share? Thanks in anticipation of your advice. Regards, Anna Gardberg My workstations and data servers: 1. Linux (Fedora Core 8, currently, as this distro is well supported in xtallography circles--I'm tempted to try Ubuntu next upgrade cycle) 2. If stereo is not an issue, any 8000 or 9000 series GeForce NVidia card is fine. I avoid ATI because of flaky Linux drivers. The 9600GT Nvidia is a bargain, but an 8600 is OK, too. 3. I use a 24" Samsung LCD display. You need CRTs only if doing stereo, but then you would need a really expensive Quadro card with stereo outputs. 4. Quad cores are so cheap right now, there is no reason not to go this route. I built my workstations from really cheap Intel DG35EC boards with Intel Q9300 processors. 250Gbyte drive for the OS and programs, and a 400 Gbyte drive for data, all SATA-300. I loaded up with a memory card reader, SATA DVD-writer, and packaged it all up in a tiny cubical micro-ATX box. You can build the whole workstation for under $1500, including the LCD monitor. My data and backup servers are installed on junk machines salvaged from offices and labs. Anything with a Pentium 4 is fast enough to be a data or backup server. The nice thing about building your own computer is when it comes time for an upgrade, you can save many of the parts (case, power supply, DVD writer, etc.) and just upgrade the motherboard, CPU, and maybe the hard drives. Cheers, -- Roger S. Rowlett Professor Colgate University Presidential Scholar Department of Chemistry Colgate University 13 Oak Drive Hamilton, NY 13346 tel: (315)-228-7245 ofc: (315)-228-7395 fax: (315)-228-7935 email: [EMAIL PROTECTED]