You haven't used Veritas Volume Manager in a while, have you? Veritas has supported encapsulating your root volume for as long as I've been using it, since version 2.x. It also supports everything else you just mentioned, including resizing UFS partitions on the fly, without even taking down the box.
In regards to networking, the feature you're talking about is called IPMP, or IP Multipathing. It's really great because you can have two NICs going to the same ethernet segment and if one of them goes down the other interface automatically spoofs it's IP and MAC address and data flow continues. It also load balances outbound traffic, which is great for those large database servers where you need multiple gigabit NICs.
Nice comment man. Very informative. Too bad I don't have any mod points or I'd mod you up. It amazes me all the Lintel zealots out there that will moderate a comment complaining about the price of Unix up to 5, and yours only gets modded up to 3.
People that have worked on enterprise level Unix systems might be amazed if they try browsing at level 0 on this thread and see how many informative comments got moderated down to 0.
I know this is somewhat off-topic, but I think the Slashdot community would gain a lot more credibility by learning from Unix, rather than trying to discredit it or quash facts that make it look better than Linux.
Is Linux a great operating system? Yes. If you need to setup a low cost LAMP server or Samba server for your organization, or even if you need a highly scalable web farm with hundreds of 1U dual CPU boxes, Linux will definitely give you the best price/performance ratio.
On the other hand, if you need serious big iron for an Oracle database server sitting in front of 10 TB of SAN storage, Linux is not even a contender in that market. You want an HP-UX, Solaris, or AIX system.
The bottom line is this: Use the best tool for the job. Sometimes it's Linux, sometimes it's commercial Unix.
And a lot of smaller companies don't have the man power or the knowledge to manage an Active Directory either. I thought we were talking about large enterprises that need clustered directory solutions, not small mom-and-pops.
If you look at the Gflops number, and divide by the number of processors, you'll see that a 4 processor box gives you 0.7225 gflops per processor, and a 64 processor box still gives you 0.6846 gflops per processor. That is a very close difference and means that Solaris scales extremely well all the way from 1 up to however many processors you throw at it.
Trust me, Linux sucks on anything more than 4 processors. It has nothing to do with the hardware you put it on. The kernel needs to be optimized to take advantage of SMP better. I'm not a kernel developer, so I can't comment on what needs to happen, but if you are under the mistaken impression that Linux will run well on anything more than 4 processors, then you are simply mistaken.
Linear means if you graph out the number of processors by the performance of the system it goes diagonally up and to the right, linearly. If you do this on a Linux box it begins to flatline after you reach 4 processors.
Basically, this means that you get close to 100% performance out of every processor, as long as you're running multi-threaded applications like Oracle or Java.
Iplanet Directory Server (the most powerful and scalable LDAP implementation available) supports clustering and replication. It's available on Solaris, and you get a free license to use it for up to 200,000 entries with Solaris. Not only that, on every benchmark I've seen it outperforms OpenLDAP by about 10x.
Why would anyone want to lock themselves into Microsoft's proprietary extensions on LDAP when you could just have real LDAP, that works on any platform, not just Windows clients?
1. Binary Compatibility - Enterprise customers want to know that the app they wrote on Solaris 2.6 will still run on Solaris 10 when it's released next year, without recompiling it. Sun gives them that, with a binary compatibility promise. The Linux kernel and GNU libraries change too frequently for customers to have that guarantee.
2. Scalability - Linux needs to be able to scale linearly beyond 4 processors before it will be a serious contender in the Enterprise space.
3. 64-bit throughout - Sun has spent years removing all 32-bit bottlenecks from every piece of code that makes up the Solaris OE (operating environment). This takes a lot of effort and even when we see the AMD 64-bit processors this fall, it's still going to take at least a couple years of Linux development for all of the various 32-bit bottlenecks to be found and fixed. This is no trivial task, as developers at Sun found out during the move to the 64-bit product line in Solaris 7.
4. Enterprise Volume Management - I heard Veritas was releasing Volume Manager for Linux, but I'm not sure if it's out yet. When you're carving up 10 terabytes of disk space on your EMC storage, you definitely don't want to be using fdisk...:-) Also, Solaris has built in Solstice Disk Suite which is great for mirroring your root disks and comes free. In my experience it works much better than RAID on Linux.
5. Journaling File System - Native in the kernel and runs on top of standard UFS. While we're at it, why doesn't Linux make UFS their standard filesystem? If Linux's goal is to be a Unix workalike, why not go that route? It's been good enough for Unix for years, so why not use it? To add journaling support to a UFS volume all I have to do is add "logging" to the end of my mount entry in/etc/vfstab and remount the filesystem.
6. Enterprise Level Support - This is something that unfortunately nobody has provided for Linux yet. Sure, you can get great support from Dell on the hardware, but if there's a problem with the OS forget about it. If you have talented staff Linux isn't that hard to support, but what if your staff leaves or the brilliant Linux geek that architected your system steps in front of a bus tomorrow? If you have a platinum contract with Sun, you can have 2-hour response time 24/7, to anywhere in the world. That means if I have a remote box somewhere in a closet out in Timbuktu and it goes down, all I have to do is call a 1-800 number and a Sun guy will be onsite, have the problem resolved, and have your box back up within 2-hours. They'll even re-install the OS for you if necessary, not to mention they have a really great backline support team that knows how to analyze core files and can trace back through the stack to help you find the root cause of the crash. I once had a backline Sun kernel engineer that was actually able to tell me that the box had crashed because an admin logged in as root had done a "kill -9" on a process he shouldn't have. How many Linux vendors have support departments that good?
7. VOS (Veritas Oracle Sun) Alliance - This is an alliance between these three companies that allows them to share support resources. For really high-end OLTP systems like banks, telcos, financial markets, etc., they have a single group to contact for support on their high-end database cluster. These companies literally bleed millions if they are down for even a few hours, so it's important to have one group to point a finger at and get a solution as quick as possible.
Those are the only ones I could think of right now off the top of my head. I'd love to hear someone with more info than me give me dates that these things could be implemented by.
Solaris scales linearly all the way up to 106 processors on an Enterprise 15000. This is one thing that Sun has been working on for years and has nailed very well. There are a whole bunch of issues that you have to deal with when you scale beyond a single board system. Now you have to have an interconnect (basically, similar to a SAN fabric or network switch) that let's a CPU on one system board access memory that might be on a different system board entirely. With features like a whopping 8MB cache for each processor, patented cache coherency technology, and a blazing fast interconnect, Sun is probably at least 5-10 years ahead of any Dell or Compaq solution. The Sun Fireplane interconnect gives you 9.6GB (that's gigabytes, not gigabits) a second sustained transfer between any two system boards, and 33.6GB a second sustained aggregate throughput.
Basically this is all done in hardware, so if a Linux kernel was written properly to take advantage of all the Sun proprietary memory interfaces, you could theoretically get the same level of performance by installing Linux on your Enterprise 15000, but why would you want to (other than the geek factor and the fact that it would be cool to say: I have Linux 2.5beta running on my E15K.. w00t!)? Besides, Sun isn't about to hand over all of their trade secrets in interconnect design to a bunch of open source developers. That would be giving away the family jewels.
I usually deal with flapping connections by "administratively" shutting down the interface in Cisco IOS. Just connect to the interface and issue a "shut", or "shutdown" command.
Of course this is difficult if you don't have physical access. Most large core routers should have a console server attached to them with a different physical circuit attached to it, so you should still be able to console into it even if your primary net connections are down.
I think you misinterpreted the numbers. PCI-X is a point-to-point, rather than bus technology. A connection is 1 to 32 bits wide @ 2Gbps per bit.
So if you're using a measly 2 bit connection sure you might only get 250 MB a second, but if you read the article you'll see that Intel is planning on using a 16-bit wide connection for the graphics card. This would give you 32Gbps, or roughly the same as AGP 8X. There is potential to go much faster by using a 32-bit PCI-X connection. In comparison, the other PCI-X slots on the motherboard for peripherals will be much slower, probably only using 2 bits.
I fully expect server boards to have multiple 32-bit PCI-X slots for maximum I/O throughput. In addition, we will probably see new technologies for clustering that utilize a PCI-X expansion card as a high speed server-to-server bus.
Did you read the article? (I know stupid question, this is Slashdot)
One thing you might notice if you read the article is that on the Unix side he quotes 2x Sun Fire V1280s, each with 12 processors and 32 GB of RAM. That is plenty of horsepower to have 500 users simultaneously running Mozilla, OpenOffice, KDE or Gnome, and whatever other applications they want to run.
In other words, he's already budgeted the RAM and processing power you would need for this scenario.
I've found that the 3dmark is a good benchmark for comparing your system with others that have the same graphics card or processor speed that you have, to see if you're getting the performance you should be out of your setup.
I've used it several times and after the benchmark has completed, you can go online and compare your score with others and see how well you stack up against them.
It's not just a dick-measuring contest, I've actually found my system performing 25% slower than other systems using the same graphics card, and it turned out I just had to upgrade my ATI drivers to the latest and greatest and my performance jumped 25%. This is a valuable tool to see if your system isn't configured properly.
Here's another scenario for you: How many non-geeks know that if you enable memory interleaving in your BIOS and have 2 or more DIMMS, you can essentially double your memory throughput and all of your games/apps will run much faster? (interleaving is like RAID-0 for memory)
I've also found that 3dmark is a good benchmark to run after you've overclocked your graphics card. 3dmark seems to test parts of the card that most games don't, and I've seen several times where an overclocked card will run UT2003 stable, but you throw 3dmark on there and start to get artifacts. So it is valuable to see if maybe you're overclocking a little too much and pushing your card beyond a stable level.
You might just find that by making a simple BIOS setting change or updating your drivers all of a sudden your system is twice as fast as it used to be.
3dmark is good for that, but you do have to take it with a grain of salt and run several other benchmarks to see how your card really stacks up.
Just a suggestion: You might look at getting a faster motherboard and processor combination. I see you have a decent graphics card, the GF4Ti4600 used to be top of the line before the FX came out, but you are really bottlenecking on the processor. A newer Athlon XP 2400+ or P4 2.4Ghz should do the trick for you.
I really like the way Sun handles patches, they have a much more intelligent system that doesn't rely on invading your privacy. Here's how it works:
1. You download the patchdiag.xref file from Sunsolve. This file is updated daily and contains a list of all patches available for all versions of Solaris. It's currently about 1.4 megabytes in size. You only need to download this once, throw it up on an NFS server and all of your Solaris hosts can use it.
2. You execute a Perl script called patchk.pl that compares your currently installed patches with what's available for your OS and generates an HTML page that is automatically opened in Netscape. The HTML page simply lists every patch you need and has check-boxes, a lot like Windows Update.
3. Check all the boxes for patches you need and click a button at the bottom of the page and Sunsolve generates a tarball of all your patches for you.
4. Download tarball and install from single user mode.
That is the proper way to do it, and it seems like Windows Update used to do that in previous versions but the xref file got to be too big for every single client to download every time. MS should provide an xref file that Windows administrators can download and run Windows Update across their enterprise using the xref file, not sending any information to Microsoft.
Sun has been selling systems to three letter governement agencies for quite some time that would never even consider purchasing a product that "phoned-home". If Microsoft wants to play in that ball-game they need to pull their head out of their ass and provide real enterprise level patch management.
P.S. The ability to roll-back a failed Windows Update would be nice too...
Yes, Sonar/Cakewalk is Windows only. I've tried several versions of it and I always thought it sucked. It's mostly a low-end hobbyist type package trying to masquerade as a high-end studio package.
My favorite package right now is Cubase VST with Reason 2.0 and shitloads of soft-synths. I don't have a Mac yet, but I can dream... I think I might pick up one of those new 17" Powerbooks pretty soon. My main reason for not getting a Mac was that they didn't support DDR memory. That reason is no longer valid, and with OS X they are looking better and better.
I appreciate the fact that you replied to my message. I used to be a really big on Intel based hardware and purchased quite a number of Dell PowerEdge servers at past sysadmin jobs I worked at. Then I got a job at Sun and found out what makes their servers so good.
SPEC benchmarks are not the best way to show system performance. They only measure raw CPU performance, which is not very important for most applications.
You really should be looking at memory bandwidth and I/O bandwidth for a true measure of performance.
How much memory bandwidth do you get with a P4? Well, PC800 RDRAM gives you 3.2GB/s. I think with PC1066 RDRAM and a newer 533mhz. FSB you might be able to push close to 5 GB/s.
How much memory bandwidth do you get with a SunFire V480? Here's a quote from Sun: "The high performance system interconnect, the Sun[tm] Fireplane Interconnect, delivers 9.6 GB/sec. sustained bandwidth and has an aggregate bandwidth of 33.6 GB/sec."
I guess having a 576-bit data path for memory sure is a lot better than that measly little 64 or 128-bit path you get with an Intel architecture... Even given the fact that Sun doesn't use DDR memory.
Now let's talk I/O bandwidth.
Every Intel based chipset I've seen only has 1 PCI bus. Some of the newer ones actually use 64-bit slots, but only 33mhz., so you're getting essentially 256MB/second throughput. If you've only got 32-bit PCI slots, you're only getting 128MB/second throughput. This is shared between all PCI slots, because they're on the same bus.
Now, take that SunFire V480, it has 2 PCI busses, each one has all 64-bit slots, some 66mhz., some 33mhz. You can put 1 fibre HBA in each PCI bus in a 64-bit 66mhz. slot and get 512MB/second throughput through each bus! Try doing that with an Intel box. Having 1 GB/Second out to your SAN is what is needed for enterprise apps, not to mention the internal memory bandwidth to handle all of those transactions.
While I appreciate the fact that you use statistics to back up your argument that Intel chips are faster, I don't think you're taking into account all the other components that might make a system perform better overall.
I've found in general that the Slashdot crowd is just uneducated on a lot of the complexities that make up enterprise systems and why it is that Sun boxes cost so much. Sun is not all about doing everything at blinding fast speeds. They are much more about doing it reliably, even if processor speeds need to be slowed down. You can see this throughout their entire product line. I don't think you'll find any Intel vendor out there that puts ECC across all data paths. Sun is much more concerned with making sure all of the data gets where it's supposed to go, and not necessarily as concerned with how fast it gets there.
Sorry, it's not an outright lie. You will never pay retail price for a Sun server. You need to compare "actual Sun price" quote from a reseller in order to see what you're really getting.
Also, according to the Dell web page when I add 4 Xeon processors which only have 2MB cache, and 4 GB of RAM, the price is now over $25k.
The V480 uses 1.2Ghz UltraSparc III processors with a massive 8MB of onboard cache. I'm sorry, but your wimpy little Xeon will not keep up with these processors.
Also, keep in mind a major selling point of Sun servers: ECC across all data paths. Don't expect Dell to ever give you that. You might have ECC memory, but what about the memory bus that connects to that memory? No ECC error checking/correction. This is a major differentiator between Sun and Intel systems, and one that unfortunately the Slashdot crowd doesn't understand at all.
I think that one thing people fail to realize when they are saying "Sun should convert to Linux" is that the argument for Solaris vs. Linux isn't so cut and dried.
Let's just say Linux worked out all of the scalability and SMP performance issues and was able to run on Sun's Big Iron (TM) just as fast, or even faster than Solaris. This would be great for Linux, however, I still don't think Sun would adopt Linux on a wide scale for their larger SMP boxes.
Why, you might ask? Binary Compatibility. The number one reason why a lot of enterprise customers use Solaris for their homegrown applications is that they can be confident that if they develop an application on Solaris 8 right now, when Solaris 10 comes out next year it will still run without even needing to be recompiled.
This is a huge plus for enterprise customers that spend millions developing custom in-house applications and don't want to have to worry about the FSF "breaking" glibc on every release. They also don't want to worry if Linus Torvalds' latest kernel will break their app.
Linux has been able to make a lot of headway in a short amount of time, and I'm frankly amazed at how quickly the functionality has surpassed Solaris and commercial Unices, however, that rapid growth and feature bloat comes at a price... Backwards compatibility... I would be seriously amazed if you could take an app compiled with gcc on a Redhat system from 1997 and run it on a Redhat 8.0 system from today without recompiling it.
Wow, I just checked and you are right. Logic Audio 5 was the last version for PC. Logic Audio 6 is only mac... It's hard to believe that Emagic left the Windows market so fast after being acquired by Apple... I guess Cubase is now the only decent MIDI/Audio package for PC.
At least get your facts straight. NBC is owned by General Electric, which is one of the largest defense contractors to the U.S. government around. In short, they make big bucks whenever we go to war.
How do you expect their news to be biased towards the "left", when most voices on the left are strongly opposed to war?
Sure, they may have editors and a large number of staff that are democrats, but when the voice comes down from on high at GE corporate, you can damn well bet that they're going to be about as pro-war as they come.
This is a huge part of the media bias problem in America. Almost every media company is part of a corporate conglomerate that has it's own profit interests. Too many times the profit interests take precedence over unbiased reporting...
While I agree with most of what you just said, I have to comment that Emagic Logic Audio is definitely NOT mac only. It has been available for Windows for quite some time now, and seems to have the same features on both platforms.
Slashdot Mirror
You haven't used Veritas Volume Manager in a while, have you? Veritas has supported encapsulating your root volume for as long as I've been using it, since version 2.x. It also supports everything else you just mentioned, including resizing UFS partitions on the fly, without even taking down the box.
In regards to networking, the feature you're talking about is called IPMP, or IP Multipathing. It's really great because you can have two NICs going to the same ethernet segment and if one of them goes down the other interface automatically spoofs it's IP and MAC address and data flow continues. It also load balances outbound traffic, which is great for those large database servers where you need multiple gigabit NICs.
Nice comment man. Very informative. Too bad I don't have any mod points or I'd mod you up. It amazes me all the Lintel zealots out there that will moderate a comment complaining about the price of Unix up to 5, and yours only gets modded up to 3.
I guess kids have to learn somedays...
Gnu is not unix...
People that have worked on enterprise level Unix systems might be amazed if they try browsing at level 0 on this thread and see how many informative comments got moderated down to 0.
I know this is somewhat off-topic, but I think the Slashdot community would gain a lot more credibility by learning from Unix, rather than trying to discredit it or quash facts that make it look better than Linux.
Is Linux a great operating system? Yes. If you need to setup a low cost LAMP server or Samba server for your organization, or even if you need a highly scalable web farm with hundreds of 1U dual CPU boxes, Linux will definitely give you the best price/performance ratio.
On the other hand, if you need serious big iron for an Oracle database server sitting in front of 10 TB of SAN storage, Linux is not even a contender in that market. You want an HP-UX, Solaris, or AIX system.
The bottom line is this: Use the best tool for the job. Sometimes it's Linux, sometimes it's commercial Unix.
Funny how anything informative in this thread gets moderated down... Time to medamoderate and whack some of these Sun haters down to size.
And a lot of smaller companies don't have the man power or the knowledge to manage an Active Directory either. I thought we were talking about large enterprises that need clustered directory solutions, not small mom-and-pops.
If you look at the Gflops number, and divide by the number of processors, you'll see that a 4 processor box gives you 0.7225 gflops per processor, and a 64 processor box still gives you 0.6846 gflops per processor. That is a very close difference and means that Solaris scales extremely well all the way from 1 up to however many processors you throw at it.
Trust me, Linux sucks on anything more than 4 processors. It has nothing to do with the hardware you put it on. The kernel needs to be optimized to take advantage of SMP better. I'm not a kernel developer, so I can't comment on what needs to happen, but if you are under the mistaken impression that Linux will run well on anything more than 4 processors, then you are simply mistaken.
Linear means if you graph out the number of processors by the performance of the system it goes diagonally up and to the right, linearly. If you do this on a Linux box it begins to flatline after you reach 4 processors.
Basically, this means that you get close to 100% performance out of every processor, as long as you're running multi-threaded applications like Oracle or Java.
Iplanet Directory Server (the most powerful and scalable LDAP implementation available) supports clustering and replication. It's available on Solaris, and you get a free license to use it for up to 200,000 entries with Solaris. Not only that, on every benchmark I've seen it outperforms OpenLDAP by about 10x.
Why would anyone want to lock themselves into Microsoft's proprietary extensions on LDAP when you could just have real LDAP, that works on any platform, not just Windows clients?
1. Binary Compatibility - Enterprise customers want to know that the app they wrote on Solaris 2.6 will still run on Solaris 10 when it's released next year, without recompiling it. Sun gives them that, with a binary compatibility promise. The Linux kernel and GNU libraries change too frequently for customers to have that guarantee.
:-) Also, Solaris has built in Solstice Disk Suite which is great for mirroring your root disks and comes free. In my experience it works much better than RAID on Linux.
/etc/vfstab and remount the filesystem.
2. Scalability - Linux needs to be able to scale linearly beyond 4 processors before it will be a serious contender in the Enterprise space.
3. 64-bit throughout - Sun has spent years removing all 32-bit bottlenecks from every piece of code that makes up the Solaris OE (operating environment). This takes a lot of effort and even when we see the AMD 64-bit processors this fall, it's still going to take at least a couple years of Linux development for all of the various 32-bit bottlenecks to be found and fixed. This is no trivial task, as developers at Sun found out during the move to the 64-bit product line in Solaris 7.
4. Enterprise Volume Management - I heard Veritas was releasing Volume Manager for Linux, but I'm not sure if it's out yet. When you're carving up 10 terabytes of disk space on your EMC storage, you definitely don't want to be using fdisk...
5. Journaling File System - Native in the kernel and runs on top of standard UFS. While we're at it, why doesn't Linux make UFS their standard filesystem? If Linux's goal is to be a Unix workalike, why not go that route? It's been good enough for Unix for years, so why not use it? To add journaling support to a UFS volume all I have to do is add "logging" to the end of my mount entry in
6. Enterprise Level Support - This is something that unfortunately nobody has provided for Linux yet. Sure, you can get great support from Dell on the hardware, but if there's a problem with the OS forget about it. If you have talented staff Linux isn't that hard to support, but what if your staff leaves or the brilliant Linux geek that architected your system steps in front of a bus tomorrow? If you have a platinum contract with Sun, you can have 2-hour response time 24/7, to anywhere in the world. That means if I have a remote box somewhere in a closet out in Timbuktu and it goes down, all I have to do is call a 1-800 number and a Sun guy will be onsite, have the problem resolved, and have your box back up within 2-hours. They'll even re-install the OS for you if necessary, not to mention they have a really great backline support team that knows how to analyze core files and can trace back through the stack to help you find the root cause of the crash. I once had a backline Sun kernel engineer that was actually able to tell me that the box had crashed because an admin logged in as root had done a "kill -9" on a process he shouldn't have. How many Linux vendors have support departments that good?
7. VOS (Veritas Oracle Sun) Alliance - This is an alliance between these three companies that allows them to share support resources. For really high-end OLTP systems like banks, telcos, financial markets, etc., they have a single group to contact for support on their high-end database cluster. These companies literally bleed millions if they are down for even a few hours, so it's important to have one group to point a finger at and get a solution as quick as possible.
Those are the only ones I could think of right now off the top of my head. I'd love to hear someone with more info than me give me dates that these things could be implemented by.
Bzzt! Wrong...
Solaris scales linearly all the way up to 106 processors on an Enterprise 15000. This is one thing that Sun has been working on for years and has nailed very well. There are a whole bunch of issues that you have to deal with when you scale beyond a single board system. Now you have to have an interconnect (basically, similar to a SAN fabric or network switch) that let's a CPU on one system board access memory that might be on a different system board entirely. With features like a whopping 8MB cache for each processor, patented cache coherency technology, and a blazing fast interconnect, Sun is probably at least 5-10 years ahead of any Dell or Compaq solution. The Sun Fireplane interconnect gives you 9.6GB (that's gigabytes, not gigabits) a second sustained transfer between any two system boards, and 33.6GB a second sustained aggregate throughput.
Basically this is all done in hardware, so if a Linux kernel was written properly to take advantage of all the Sun proprietary memory interfaces, you could theoretically get the same level of performance by installing Linux on your Enterprise 15000, but why would you want to (other than the geek factor and the fact that it would be cool to say: I have Linux 2.5beta running on my E15K.. w00t!)? Besides, Sun isn't about to hand over all of their trade secrets in interconnect design to a bunch of open source developers. That would be giving away the family jewels.
I usually deal with flapping connections by "administratively" shutting down the interface in Cisco IOS. Just connect to the interface and issue a "shut", or "shutdown" command.
Of course this is difficult if you don't have physical access. Most large core routers should have a console server attached to them with a different physical circuit attached to it, so you should still be able to console into it even if your primary net connections are down.
Hey man, where did you get that great GWB quote? Do you have a link to the article it was in?
Thanks!
I think you misinterpreted the numbers. PCI-X is a point-to-point, rather than bus technology. A connection is 1 to 32 bits wide @ 2Gbps per bit.
So if you're using a measly 2 bit connection sure you might only get 250 MB a second, but if you read the article you'll see that Intel is planning on using a 16-bit wide connection for the graphics card. This would give you 32Gbps, or roughly the same as AGP 8X. There is potential to go much faster by using a 32-bit PCI-X connection. In comparison, the other PCI-X slots on the motherboard for peripherals will be much slower, probably only using 2 bits.
I fully expect server boards to have multiple 32-bit PCI-X slots for maximum I/O throughput. In addition, we will probably see new technologies for clustering that utilize a PCI-X expansion card as a high speed server-to-server bus.
Off Topic:
Hey, I tried typing about:Mozilla in my address bar of IE6 and I get a plain blue screen in my browser. Am I supposed to see something different?
Thanks!
Did you read the article? (I know stupid question, this is Slashdot)
One thing you might notice if you read the article is that on the Unix side he quotes 2x Sun Fire V1280s, each with 12 processors and 32 GB of RAM. That is plenty of horsepower to have 500 users simultaneously running Mozilla, OpenOffice, KDE or Gnome, and whatever other applications they want to run.
In other words, he's already budgeted the RAM and processing power you would need for this scenario.
Nice system... :)
I've found that the 3dmark is a good benchmark for comparing your system with others that have the same graphics card or processor speed that you have, to see if you're getting the performance you should be out of your setup.
I've used it several times and after the benchmark has completed, you can go online and compare your score with others and see how well you stack up against them.
It's not just a dick-measuring contest, I've actually found my system performing 25% slower than other systems using the same graphics card, and it turned out I just had to upgrade my ATI drivers to the latest and greatest and my performance jumped 25%. This is a valuable tool to see if your system isn't configured properly.
Here's another scenario for you: How many non-geeks know that if you enable memory interleaving in your BIOS and have 2 or more DIMMS, you can essentially double your memory throughput and all of your games/apps will run much faster? (interleaving is like RAID-0 for memory)
I've also found that 3dmark is a good benchmark to run after you've overclocked your graphics card. 3dmark seems to test parts of the card that most games don't, and I've seen several times where an overclocked card will run UT2003 stable, but you throw 3dmark on there and start to get artifacts. So it is valuable to see if maybe you're overclocking a little too much and pushing your card beyond a stable level.
You might just find that by making a simple BIOS setting change or updating your drivers all of a sudden your system is twice as fast as it used to be.
3dmark is good for that, but you do have to take it with a grain of salt and run several other benchmarks to see how your card really stacks up.
Just a suggestion: You might look at getting a faster motherboard and processor combination. I see you have a decent graphics card, the GF4Ti4600 used to be top of the line before the FX came out, but you are really bottlenecking on the processor. A newer Athlon XP 2400+ or P4 2.4Ghz should do the trick for you.
I really like the way Sun handles patches, they have a much more intelligent system that doesn't rely on invading your privacy. Here's how it works:
1. You download the patchdiag.xref file from Sunsolve. This file is updated daily and contains a list of all patches available for all versions of Solaris. It's currently about 1.4 megabytes in size. You only need to download this once, throw it up on an NFS server and all of your Solaris hosts can use it.
2. You execute a Perl script called patchk.pl that compares your currently installed patches with what's available for your OS and generates an HTML page that is automatically opened in Netscape. The HTML page simply lists every patch you need and has check-boxes, a lot like Windows Update.
3. Check all the boxes for patches you need and click a button at the bottom of the page and Sunsolve generates a tarball of all your patches for you.
4. Download tarball and install from single user mode.
That is the proper way to do it, and it seems like Windows Update used to do that in previous versions but the xref file got to be too big for every single client to download every time. MS should provide an xref file that Windows administrators can download and run Windows Update across their enterprise using the xref file, not sending any information to Microsoft.
Sun has been selling systems to three letter governement agencies for quite some time that would never even consider purchasing a product that "phoned-home". If Microsoft wants to play in that ball-game they need to pull their head out of their ass and provide real enterprise level patch management.
P.S. The ability to roll-back a failed Windows Update would be nice too...
Yes, Sonar/Cakewalk is Windows only. I've tried several versions of it and I always thought it sucked. It's mostly a low-end hobbyist type package trying to masquerade as a high-end studio package.
My favorite package right now is Cubase VST with Reason 2.0 and shitloads of soft-synths. I don't have a Mac yet, but I can dream... I think I might pick up one of those new 17" Powerbooks pretty soon. My main reason for not getting a Mac was that they didn't support DDR memory. That reason is no longer valid, and with OS X they are looking better and better.
I appreciate the fact that you replied to my message. I used to be a really big on Intel based hardware and purchased quite a number of Dell PowerEdge servers at past sysadmin jobs I worked at. Then I got a job at Sun and found out what makes their servers so good.
SPEC benchmarks are not the best way to show system performance. They only measure raw CPU performance, which is not very important for most applications.
You really should be looking at memory bandwidth and I/O bandwidth for a true measure of performance.
How much memory bandwidth do you get with a P4? Well, PC800 RDRAM gives you 3.2GB/s. I think with PC1066 RDRAM and a newer 533mhz. FSB you might be able to push close to 5 GB/s.
How much memory bandwidth do you get with a SunFire V480? Here's a quote from Sun: "The high performance system interconnect, the Sun[tm] Fireplane Interconnect, delivers 9.6 GB/sec. sustained bandwidth and has an aggregate bandwidth of 33.6 GB/sec."
I guess having a 576-bit data path for memory sure is a lot better than that measly little 64 or 128-bit path you get with an Intel architecture... Even given the fact that Sun doesn't use DDR memory.
Now let's talk I/O bandwidth.
Every Intel based chipset I've seen only has 1 PCI bus. Some of the newer ones actually use 64-bit slots, but only 33mhz., so you're getting essentially 256MB/second throughput. If you've only got 32-bit PCI slots, you're only getting 128MB/second throughput. This is shared between all PCI slots, because they're on the same bus.
Now, take that SunFire V480, it has 2 PCI busses, each one has all 64-bit slots, some 66mhz., some 33mhz. You can put 1 fibre HBA in each PCI bus in a 64-bit 66mhz. slot and get 512MB/second throughput through each bus! Try doing that with an Intel box. Having 1 GB/Second out to your SAN is what is needed for enterprise apps, not to mention the internal memory bandwidth to handle all of those transactions.
While I appreciate the fact that you use statistics to back up your argument that Intel chips are faster, I don't think you're taking into account all the other components that might make a system perform better overall.
I've found in general that the Slashdot crowd is just uneducated on a lot of the complexities that make up enterprise systems and why it is that Sun boxes cost so much. Sun is not all about doing everything at blinding fast speeds. They are much more about doing it reliably, even if processor speeds need to be slowed down. You can see this throughout their entire product line. I don't think you'll find any Intel vendor out there that puts ECC across all data paths. Sun is much more concerned with making sure all of the data gets where it's supposed to go, and not necessarily as concerned with how fast it gets there.
Thanks for reading this long reply.
Sorry, it's not an outright lie. You will never pay retail price for a Sun server. You need to compare "actual Sun price" quote from a reseller in order to see what you're really getting.
Also, according to the Dell web page when I add 4 Xeon processors which only have 2MB cache, and 4 GB of RAM, the price is now over $25k.
The V480 uses 1.2Ghz UltraSparc III processors with a massive 8MB of onboard cache. I'm sorry, but your wimpy little Xeon will not keep up with these processors.
Also, keep in mind a major selling point of Sun servers: ECC across all data paths. Don't expect Dell to ever give you that. You might have ECC memory, but what about the memory bus that connects to that memory? No ECC error checking/correction. This is a major differentiator between Sun and Intel systems, and one that unfortunately the Slashdot crowd doesn't understand at all.
I think that one thing people fail to realize when they are saying "Sun should convert to Linux" is that the argument for Solaris vs. Linux isn't so cut and dried.
Let's just say Linux worked out all of the scalability and SMP performance issues and was able to run on Sun's Big Iron (TM) just as fast, or even faster than Solaris. This would be great for Linux, however, I still don't think Sun would adopt Linux on a wide scale for their larger SMP boxes.
Why, you might ask? Binary Compatibility. The number one reason why a lot of enterprise customers use Solaris for their homegrown applications is that they can be confident that if they develop an application on Solaris 8 right now, when Solaris 10 comes out next year it will still run without even needing to be recompiled.
This is a huge plus for enterprise customers that spend millions developing custom in-house applications and don't want to have to worry about the FSF "breaking" glibc on every release. They also don't want to worry if Linus Torvalds' latest kernel will break their app.
Linux has been able to make a lot of headway in a short amount of time, and I'm frankly amazed at how quickly the functionality has surpassed Solaris and commercial Unices, however, that rapid growth and feature bloat comes at a price... Backwards compatibility... I would be seriously amazed if you could take an app compiled with gcc on a Redhat system from 1997 and run it on a Redhat 8.0 system from today without recompiling it.
Anyway, here is a link to Sun's Binary Compatibility Promise.
Wow, I just checked and you are right. Logic Audio 5 was the last version for PC. Logic Audio 6 is only mac... It's hard to believe that Emagic left the Windows market so fast after being acquired by Apple... I guess Cubase is now the only decent MIDI/Audio package for PC.
At least get your facts straight. NBC is owned by General Electric, which is one of the largest defense contractors to the U.S. government around. In short, they make big bucks whenever we go to war.
How do you expect their news to be biased towards the "left", when most voices on the left are strongly opposed to war?
Sure, they may have editors and a large number of staff that are democrats, but when the voice comes down from on high at GE corporate, you can damn well bet that they're going to be about as pro-war as they come.
This is a huge part of the media bias problem in America. Almost every media company is part of a corporate conglomerate that has it's own profit interests. Too many times the profit interests take precedence over unbiased reporting...
While I agree with most of what you just said, I have to comment that Emagic Logic Audio is definitely NOT mac only. It has been available for Windows for quite some time now, and seems to have the same features on both platforms.