[Government] need[s] to be able to maintain order or else [citizens] can't learn. If the [citizens] have no respect for the [Government's] authority, they can't maintain order. Therefore, anything a [Citizen] does that causes other [Citizen's] to lose respect for their [Government] must be punished. The alternative is to have a situation where classes are completely out of control and the [citizen] who want to learn, can't.
Remember. Most school teachers are government employees.
> I am not aware of any internal FC implementations on standard server hardware, but as a rule, its an external JBOD application.
* SunBlade 1000 uses FCAL internally (workstation)
* SunFire 280R, v880, and 3500 use FVAL internally (servers)
More and more, FC is not used for JBOD, for RAIDed disks. And more and more prevalent is the use of FC Hubs/Switches to make managed of disks much easier (in a large server environment). FC is *almost* a necessity in (failover) cluster environments of more than two systems. BTW... FC uses the SCSI 3 protocal over copper/fiber.;-)
I think you're missing a few things. First of all, yes it has 32 CPUs... but each CPU has two cores on it, i.e... it's more comparable to a 64 CPU box. Secondly, the CPUs have considerably higher benchmarks than Suns USIIIi's. The 6800's can only support 4 domains, total, split between two hardware segments. This will handle up to 16 domains. We still need to compare it to an E10K or a F15K. While an E10k can handle a single CPU on a board, the F15K CPUs are soldered onto the CPU/Mem boards, and thus are only available in unit of 2 or 4. Also, as IBM invented partition (xx/360-390 and AS/400), I don't know that I'd compare this "first generation" partitioning with Sun's "fifth generation" partitioning. (And they didn't get help from Cray early on... Cray WROTE IT early on, sold it to SGI, who then sold the C6400 to Sun and was renamed the E10K.)
I bet you're right... I bet there are more apps for Solaris, however when it comes to the apps that actually run on boxen this large, you're typically talking about an Oracle or DB/2 database, which are available for both platforms.
Both the F15K and p690 appear to be fantastic boxes. Only time will tell how they fare against each other.
Actually, while giving a "case-study" of Veritas Cluster Server at the Vertex 2001 conference, an Ebay admin said they were running two fully loaded Sun E10ks running Oracle on the backed, and were looking at splitting the Oracle database into multiple chunks. No postgres, etc. =-)
- go to http://www.neverwinternights.com/about.html
- Click on VII The Techincal Stuff
- Click Edit, Find.
- Type "Linux".
- Do more research next time you open trap.
LVM only allows resizing of Logical Volumes, not filesystems. Using a filesystem that will allow dynamic resing (JFS, or ReiserFS, or ext2 with a specific kernel patch) will allow current Linux boxes to resize filesytems on the fly. I currently use Linux LVM with ReiserFS and have on many occasions resized filesystems live. Even without the ability to dynamically resize volumes, LVMs allow you to take better advantage of your disk space, however. It allows you to stripe volumes as well as span multiple disks. Solaris and Linux native disk management doesn't really allow for this (although MD will allow you to do both).
When you buy a Sun the damned thing just doesn't fall down unless you have a system mangler who keeps dicking around with it. And if a single Sun could not address the problem, then maybe it's time to buy some real iron, like a maxed out S/390. When you have a terabyte of data to process, you have to start paying a little more attention to things like I/O.
I can assure, as one with experience, that Suns most certainly *DO* fall over. The E10K is a nice box with quite a few redundant feature. The ability to remove a system board on the fly ranks pretty high up there. Assuming you lose a processor or memory and assuming (this is a big one) that your system doesn't fall over immediatly, you can most likely replace that processor hot. However... it's been my experience that failling processors or memory bring the box (or domain) down more often than not. And let me assure, sun processors *DO* fail periodically. In a large shop, you can exect to replace at least one or two processors per year.
I would have thought it to be wiser to setup Sun E10000's or something like that.. having 4 32 proc e 100000's in a cluster is a hell of alot easier to manage and cheaper.
Last I checked (this was about a year or so ago) a fully loaded (64/64) E10K ran around $12M and the base (2psr) system was running around $800,000. Even if that's off by a factor of 3 or 4, you're still talking $3-$4M a piece... at three of them, you're looking at between $12-$48M. On the other hand, the typical white box PC will run between $800-$1500. That amounts to $3.68M-$6.9M for 4600 nodes. This doesn't include the network infrastructure or administration costs, however, as someone who has administered large clusters (largest was an 80 node SP/2), it actually becomes easier to administer that many nodes in a cluster than it would that many servers. Keep in mind that there most certainly are groupings of nodes where they are kept identical except for IP.
Another significant expense is that hardware support costs associated with such systems. If you have 4600 nodes, it's trivial to simply keep (MANY) spare systems floating around. Also, you can disable a node with negligible impact. Even if you're subdomaining an E10K, there are (a small few) single points of failure on the platform (regardless of what Suns documentation says). If you're not subdomaining it, you're simply talking a 32way SMP box (might as well just use a 6500 for that configuration). If you were to lose the backplane for whatever reason, you've lost a singificant portion of your compute resources.
How is this any different than a Sun E10K or some Mainframes and Crays (amongst others) that support the ability to dynamically add/remove not only CPUs, but memory, and cards. Even most new Intel servers (and NT) support at least some ability to dynamically add/remove PCI cards.
Sun Microsystems claims to be a leader in system reliability and more reliable than Windows.
A bold statement. Yet they immediatly go into discussion about the unreliability of Sun's hardware and how stables Windows is. (They're probably right though. I doubt that a Windows keyboard or mouse has causes a mission critical system to crash.) They also discuss how the Gartner Group (pah) doesn't recommend Sun's for environments that require high availability. HA is a buzzword that could mean several things. They're probably refering to fail-over clustering, in which case, Sun's behind the game on it... but other companies that make HA products for Sun are quite examplary (Qualix, Veritas FirstWatch, etc.). I'd like to see that context of the Gartner Groups' (pah) claims.
Actually, as this site says, EBay runs on a pair of E10K (upto 64 processors and 64 GB of ram. woohoo.;-) for the backend and some NT boxes on the front-end. I remember talking to our local Sun SE after the EBay incidents, and he had stated that EBay was not installing critical patch updates and wasn't listening to their Sun techs. If that's truly the case, then it's interesting that MS put that as an example. I doubt they want bad implementations of NT pointed out to them in a public forum. *grin*
There is already a project underway to get an LVM for linux. I've been successfully using it for approximately a year now. Take a look at http://linux.msede.com/lvm/ The command structure is very similar to HP-UX, and the theory behind it consitent with AIX's and HP-UX's LVMs.
I didn't notice this in any of the other comments, so I thought I'd comment on it. =-)
First, they state that there is no GUI for Samba with RedHat. This is simply wrong. There is one that is built into LinuxConf. Not only that, though, redhat includes SWAT -- all that is needed to turn it on is to edit/etc/inetd.conf and restart inetd. As they mention that RedHat (and Unix's) ability to turn on/off individual ports, they have at least SOME understanding of this file and its purpose.
Second, they mention that RedHat lacks any graphical monitoring tools (and they only mention vmstat and iostat). RedHat includes several different graphical monitoring tools including xosview, xsysinfo, and gtop. That being said, I don't believe that the NT monitoring tools give quite the depth of information that iostat, vmstat, and sar give.
Third, they mention W2K's active directory and the ability to manage it inside it's MMC. RedHat linux ships with OpenLDAP, (ADS is basically a superset of LDAP) as well as the ability to configure it, add/delete users, etc. in LinuxConf. Not only that, but RedHat ships with a pam module so that you can authenticate against an LDAP server (as well as a RADIUS server, SMB server, etc. with unshipped pam modules).
Forth, documentation... RedHat not only ships their three modules, but it ships a copy the LDP's list of HOWTO's, the Apache Manual, the PHP Manual, Man pages, as well as/usr/doc (which will ~100m on my install). These include not only basic installation and maintenance, but basically everything most admins ever wanted to know in a reasonably readable format. Add to that that almost all of it is availble in an HTML browser window, something that they marked against W2K.
Lastly... Since when did RedHat only cover the "Bare-bones network services"? Generally speaking, the problems I've run across with RedHat installs is that the number of network services installed FAR exceeds the needs of the system in question. It includes SMB file/printer sharing, Netware file/printer sharing, Appletalk, etc. It includes a Web and FTP server, allows for remote admin, etc. etc. etc. (I'm sorry for pointing out things ya'll already know... but this really strikes me as a poorly researched article that judges the contestants in a very subjective way.)
I've worked for quite a few IT managers in some rather large shops (1000+ servers). Not once have I had one not willing for us to install any number of patches, just so long as they have been tested in a test environment. I have to wonder where ZD is getting the idea that enterprise businesses CARE how many patches are being installed (or what OS they're running, for that matter). Most companies simply ant a stable platform to run their applications on.
I'd like to comment on the Solaris x86 support for hardware. I work in a rather large IT shop that uses x86 on quite a number of boxen. There seems to be some misunderstandings about x86 hardware support. It's been my experience that Linux support a considerably larger variety of hardware that Solaris x86 does. Even down to simple things as CD-ROM drives. On multiple ocassionas I've had x86 not "see" an ATAPI CD-ROM drive that Linux will. I've only had one ATAPI CDROM that Linux didn't recognize. This is doubly so for video cards. Also bear in mind that the package managers for Linux are light years ahead of Solaris' package system. (RPM, etc.) As well as any Linux distro is far more "complete" that Solaris is, at the moment. That's not to say that there aren't benefits to Solaris x86. If you're already a Solaris shop, it's easier for admins to deal with one Unix. Also CDE/Motif are included in the package as well as better support for multiple heads (and a front panel on both!).
Indeedy, you are correct. Either a PPC or a Power series chip. And Blue Pacific uses 4 way 604e silvers. I thought I'd read that they moved to 2 way power3's, but that was probably another of the larger SP/2s. In the site you mentioned, I find it odd that they only display 176 nodes, as opposed to all 1464, though. Do you know if they partitioned their SP?
Actually, Clustering has been around for quite some time. The largest 'supercomputers' (Blue Mountain and Blue Pacific, specifically) are built on massive clusters of various nodes. Blue Pacific (an IBM RS/6000 SP/2) is built upon 1500 nodes (at a sum cost of ~$93 of taxpayer dollars *grin*) using a combination of PPC 604e and Power3 processors. Blue Mountain is a comparable SGI Origin system using MIPS processors. For more information on large systems, check out http://www.top500.org/ .
BTW... RS/6000 is a product line, not a CPU. The RS/6000 line uses the PowerPC series (601,603,603e,604,604e) for interger math and Power, Power2, Power2sc, and Power3 processors for floating point math.
Slashdot Mirror
[Government] need[s] to be able to maintain order or else [citizens] can't learn. If the [citizens] have no respect for the [Government's] authority, they can't maintain order. Therefore, anything a [Citizen] does that causes other [Citizen's] to lose respect for their [Government] must be punished. The alternative is to have a situation where classes are completely out of control and the [citizen] who want to learn, can't.
Remember. Most school teachers are government employees.
I think we could characterize them in the words "Fibre Channel". ;-)
> I am not aware of any internal FC implementations on standard server hardware, but as a rule, its an external JBOD application.
;-)
* SunBlade 1000 uses FCAL internally (workstation)
* SunFire 280R, v880, and 3500 use FVAL internally (servers)
More and more, FC is not used for JBOD, for RAIDed disks. And more and more prevalent is the use of FC Hubs/Switches to make managed of disks much easier (in a large server environment). FC is *almost* a necessity in (failover) cluster environments of more than two systems. BTW... FC uses the SCSI 3 protocal over copper/fiber.
I think you're missing a few things. First of all, yes it has 32 CPUs... but each CPU has two cores on it, i.e... it's more comparable to a 64 CPU box. Secondly, the CPUs have considerably higher benchmarks than Suns USIIIi's. The 6800's can only support 4 domains, total, split between two hardware segments. This will handle up to 16 domains. We still need to compare it to an E10K or a F15K. While an E10k can handle a single CPU on a board, the F15K CPUs are soldered onto the CPU/Mem boards, and thus are only available in unit of 2 or 4. Also, as IBM invented partition (xx/360-390 and AS/400), I don't know that I'd compare this "first generation" partitioning with Sun's "fifth generation" partitioning. (And they didn't get help from Cray early on... Cray WROTE IT early on, sold it to SGI, who then sold the C6400 to Sun and was renamed the E10K.)
I bet you're right... I bet there are more apps for Solaris, however when it comes to the apps that actually run on boxen this large, you're typically talking about an Oracle or DB/2 database, which are available for both platforms.
Both the F15K and p690 appear to be fantastic boxes. Only time will tell how they fare against each other.
Actually, while giving a "case-study" of Veritas Cluster Server at the Vertex 2001 conference, an Ebay admin said they were running two fully loaded Sun E10ks running Oracle on the backed, and were looking at splitting the Oracle database into multiple chunks. No postgres, etc. =-)
- go to http://www.neverwinternights.com/about.html
- Click on VII The Techincal Stuff
- Click Edit, Find.
- Type "Linux".
- Do more research next time you open trap.
I *DO* work for a certain nameless long distance provider... and the employees are most certainly celebrating. =-)
LVM only allows resizing of Logical Volumes, not filesystems. Using a filesystem that will allow dynamic resing (JFS, or ReiserFS, or ext2 with a specific kernel patch) will allow current Linux boxes to resize filesytems on the fly. I currently use Linux LVM with ReiserFS and have on many occasions resized filesystems live. Even without the ability to dynamically resize volumes, LVMs allow you to take better advantage of your disk space, however. It allows you to stripe volumes as well as span multiple disks. Solaris and Linux native disk management doesn't really allow for this (although MD will allow you to do both).
When you buy a Sun the damned thing just doesn't fall down unless you have a system mangler who keeps dicking around with it. And if a single Sun could not address the problem, then maybe it's time to buy some real iron, like a maxed out S/390. When you have a terabyte of data to process, you have to start paying a little more attention to things like I/O.
I can assure, as one with experience, that Suns most certainly *DO* fall over. The E10K is a nice box with quite a few redundant feature. The ability to remove a system board on the fly ranks pretty high up there. Assuming you lose a processor or memory and assuming (this is a big one) that your system doesn't fall over immediatly, you can most likely replace that processor hot. However... it's been my experience that failling processors or memory bring the box (or domain) down more often than not. And let me assure, sun processors *DO* fail periodically. In a large shop, you can exect to replace at least one or two processors per year.
I would have thought it to be wiser to setup Sun E10000's or something like that.. having 4 32 proc e 100000's in a cluster is a hell of alot easier to manage and cheaper.
Last I checked (this was about a year or so ago) a fully loaded (64/64) E10K ran around $12M and the base (2psr) system was running around $800,000. Even if that's off by a factor of 3 or 4, you're still talking $3-$4M a piece... at three of them, you're looking at between $12-$48M. On the other hand, the typical white box PC will run between $800-$1500. That amounts to $3.68M-$6.9M for 4600 nodes. This doesn't include the network infrastructure or administration costs, however, as someone who has administered large clusters (largest was an 80 node SP/2), it actually becomes easier to administer that many nodes in a cluster than it would that many servers. Keep in mind that there most certainly are groupings of nodes where they are kept identical except for IP.
Another significant expense is that hardware support costs associated with such systems. If you have 4600 nodes, it's trivial to simply keep (MANY) spare systems floating around. Also, you can disable a node with negligible impact. Even if you're subdomaining an E10K, there are (a small few) single points of failure on the platform (regardless of what Suns documentation says). If you're not subdomaining it, you're simply talking a 32way SMP box (might as well just use a 6500 for that configuration). If you were to lose the backplane for whatever reason, you've lost a singificant portion of your compute resources.
How is this any different than a Sun E10K or some Mainframes and Crays (amongst others) that support the ability to dynamically add/remove not only CPUs, but memory, and cards. Even most new Intel servers (and NT) support at least some ability to dynamically add/remove PCI cards.
Sun Microsystems claims to be a leader in system reliability and more reliable than Windows.
A bold statement. Yet they immediatly go into discussion about the unreliability of Sun's hardware and how stables Windows is. (They're probably right though. I doubt that a Windows keyboard or mouse has causes a mission critical system to crash.) They also discuss how the Gartner Group (pah) doesn't recommend Sun's for environments that require high availability. HA is a buzzword that could mean several things. They're probably refering to fail-over clustering, in which case, Sun's behind the game on it... but other companies that make HA products for Sun are quite examplary (Qualix, Veritas FirstWatch, etc.). I'd like to see that context of the Gartner Groups' (pah) claims.
Actually, as this site says, EBay runs on a pair of E10K (upto 64 processors and 64 GB of ram. woohoo. ;-) for the backend and some NT boxes on the front-end. I remember talking to our local Sun SE after the EBay incidents, and he had stated that EBay was not installing critical patch updates and wasn't listening to their Sun techs. If that's truly the case, then it's interesting that MS put that as an example. I doubt they want bad implementations of NT pointed out to them in a public forum. *grin*
There is already a project underway to get an LVM for linux. I've been successfully using it for approximately a year now. Take a look at http://linux.msede.com/lvm/ The command structure is very similar to HP-UX, and the theory behind it consitent with AIX's and HP-UX's LVMs.
--Daniel
I didn't notice this in any of the other comments, so I thought I'd comment on it. =-)
/etc/inetd.conf and restart inetd. As they mention that RedHat (and Unix's) ability to turn on/off individual ports, they have at least SOME understanding of this file and its purpose.
/usr/doc (which will ~100m on my install). These include not only basic installation and maintenance, but basically everything most admins ever wanted to know in a reasonably readable format. Add to that that almost all of it is availble in an HTML browser window, something that they marked against W2K.
First, they state that there is no GUI for Samba with RedHat. This is simply wrong. There is one that is built into LinuxConf. Not only that, though, redhat includes SWAT -- all that is needed to turn it on is to edit
Second, they mention that RedHat lacks any graphical monitoring tools (and they only mention vmstat and iostat). RedHat includes several different graphical monitoring tools including xosview, xsysinfo, and gtop. That being said, I don't believe that the NT monitoring tools give quite the depth of information that iostat, vmstat, and sar give.
Third, they mention W2K's active directory and the ability to manage it inside it's MMC. RedHat linux ships with OpenLDAP, (ADS is basically a superset of LDAP) as well as the ability to configure it, add/delete users, etc. in LinuxConf. Not only that, but RedHat ships with a pam module so that you can authenticate against an LDAP server (as well as a RADIUS server, SMB server, etc. with unshipped pam modules).
Forth, documentation... RedHat not only ships their three modules, but it ships a copy the LDP's list of HOWTO's, the Apache Manual, the PHP Manual, Man pages, as well as
Lastly... Since when did RedHat only cover the "Bare-bones network services"? Generally speaking, the problems I've run across with RedHat installs is that the number of network services installed FAR exceeds the needs of the system in question. It includes SMB file/printer sharing, Netware file/printer sharing, Appletalk, etc. It includes a Web and FTP server, allows for remote admin, etc. etc. etc. (I'm sorry for pointing out things ya'll already know... but this really strikes me as a poorly researched article that judges the contestants in a very subjective way.)
I've worked for quite a few IT managers in some rather large shops (1000+ servers). Not once have I had one not willing for us to install any number of patches, just so long as they have been tested in a test environment. I have to wonder where ZD is getting the idea that enterprise businesses CARE how many patches are being installed (or what OS they're running, for that matter). Most companies simply ant a stable platform to run their applications on.
I'd like to comment on the Solaris x86 support for hardware. I work in a rather large IT shop that uses x86 on quite a number of boxen. There seems to be some misunderstandings about x86 hardware support. It's been my experience that Linux support a considerably larger variety of hardware that Solaris x86 does. Even down to simple things as CD-ROM drives. On multiple ocassionas I've had x86 not "see" an ATAPI CD-ROM drive that Linux will. I've only had one ATAPI CDROM that Linux didn't recognize. This is doubly so for video cards. Also bear in mind that the package managers for Linux are light years ahead of Solaris' package system. (RPM, etc.) As well as any Linux distro is far more "complete" that Solaris is, at the moment. That's not to say that there aren't benefits to Solaris x86. If you're already a Solaris shop, it's easier for admins to deal with one Unix. Also CDE/Motif are included in the package as well as better support for multiple heads (and a front panel on both!).
;-)
Of course I'm an AIX bigot... but hey.
Indeedy, you are correct. Either a PPC or a Power series chip. And Blue Pacific uses 4 way 604e silvers. I thought I'd read that they moved to 2 way power3's, but that was probably another of the larger SP/2s. In the site you mentioned, I find it odd that they only display 176 nodes, as opposed to all 1464, though. Do you know if they partitioned their SP?
Actually, Clustering has been around for quite some time. The largest 'supercomputers' (Blue Mountain and Blue Pacific, specifically) are built on massive clusters of various nodes. Blue Pacific (an IBM RS/6000 SP/2) is built upon 1500 nodes (at a sum cost of ~$93 of taxpayer dollars *grin*) using a combination of PPC 604e and Power3 processors. Blue Mountain is a comparable SGI Origin system using MIPS processors. For more information on large systems, check out http://www.top500.org/ .
BTW... RS/6000 is a product line, not a CPU. The RS/6000 line uses the PowerPC series (601,603,603e,604,604e) for interger math and Power, Power2, Power2sc, and Power3 processors for floating point math.