It does well in a single portion of the server benchmark space, and the real question is, how long will it continue to hold that lead when Itanium's future dev looks relatively bleak in contrast to AMD's push?
Multi-cores and NUMA on the desktop? Desktop pricing pressures on server features? That price point drops like a rock, can Intel still sell Itanium at anything approaching a reasonable price in 1 year that customers will pay that allows intel to reap any profit whatsoever?
I'll predict it here - Intel has lost this battle, and in 2 years, Itanium will be a footnote in the processor space, much like alpha and MIPS. Not only that, but Intel's going to be much smaller in the CPU space unless they can come up with a rabbit to sell in the 64 bit space. I don't think rumors of the P6 dual core driven processors is going to cut it either, in light of AMD's dual core opterons...
Some of us tried and tried to support alternative software way back when. MS killed all chances of doing so. I was running OS/2 with wordperfect, several other OS/2 apps including compilers and my own code. All for naught when MS Office 97 came out and had '0' backwards compatibility and the execs had copies of 97....
Same issue here - have problems with internal site being written by some IE only dimwits, and what's IE only? Why, DHTML menus with drop downs, which are perfectly possible in Firefox as well, with proper coding. Oh well, life goes on, I suppose.
You're surprised? You give the rival's website as the only example? You'd expect a site that depends most likely on security holes, err, features, in IE to work with a real browser?
Mine looks nothing like that. I'm running PR1.0 now, and have been running Firefox since the 0.8 beta, with Mozilla all the way back to 1.1 days. Mozilla had some issues, but they mostly dissappeared around versions 1.3/1.4. I too would love to see a page that's
a real site that's not someone's screwed up homepage that doesn't know HTML from an abaqus
And screws up something other than DHTML/Flash that's especially broken for IE only
to log into the client to access anything at all on the laptop (can be gotten around, I know)
the VPN client not having the capability to save the password or the user name. (Quite annoying if you have a long name as the basis for a 20 character username....)
a 5 digit PIN required as a prefix or suffix to your SecureID number. (varied by account)
if you failed to log in 5 times, the VPN account was locked
Yep, seems like a breeze to break into a VPN with those restrictions....
SecureID just seems like the next logical step. I used one for 3 years, and, once you get used to not attempting to log into your VPN when only the last bar is showing (there's a countdown bar indicating how much time is left before the number changes) it's really not so bad.
They appear to run on pseudo random number generators, and are synched up with the server with a known seed. I imagine they'd be very difficult to crack, as our system was configured to only allow 1 login attempt per number, if you typed in the wrong password/SecureID number, you had to wait until the next number came along. Annoying, but definitely better than the 3 (or 5) attempts and get a system admin to unlock your account.
Hey, don't run them on my XP box either, and it runs a whole lot better without all that crap. However, let's be fair and say that it is behind a firewall on a private network, and has no MS software installed on top of the OS. So - no macro viruses via office documents, no web bugs/viruses via IE, and no virus of the day via Outlook. Amazing, that. Also, all executable downloads are run through a virus checker first, on another machine.
Actually, I have had 2K and XP here at work. Since "upgrading" to XP, I've had to reboot at least every 4-6 days, because really odd things start happening. BTW, these are IT managed systems, so I would assume they're equally managed, as it's the same folks managing them.
Now, I have a home XP machine that's up and running generally around 4-5 months between reboots. And that's even playing a couple of games on those (not done @ work of course). I should mention that autoupdate is turned off, among other things, and it lives isolated behind a firewall and runs no MS software.:) The 2K machine stays up continuously, running as a webserver with apache.
Hey, I just dug back down to the desk level view here over the past few weeks. Quite an effort, and will probably result in the offloading of a bunch of crap to ebay....
Those retail prices for SATA disks are pulled from Pricewatch. Yes, you can find special sales, sales with rebates, etc, that drop the price, but people putting together RAID systems generally don't want to wait over a period of 5 months to get all the drives they need for their RAID systems.
Check out this quick review where premium U-320 4 drive array spanks SATA in every test, and this was merely for RAID 0. That's no extra processing overhead. Tom's Hardware has a more thorough article, but is heavily slanted in SATA's favor. I especially love this quote:
As incredible as this may sound, SATA has performance advantages over Ultra320 - provided it's used correctly and in conjunction with a sufficiently fast interface with the system. That is because each SATA hard drive communicates with the controller via its own fast (150 MB/s) point-to-point connection while the SCSI bus is used jointly by all devices.
Yes, it sounds incredible to anyone that knows anything about how to setup SCSI RAID. First, there's not a drive made that comes anywhere near 150 MB/s continuous transfer speed, only the data in the buffer could theoretically be transferred at that speed, so that red herring is pretty smelly already.
Second, the configuration issue is definitely something to be considered with SCSI. Proper setup will actually have SCSI hitting near its theoretical peaks data transfer more consistently than not. There are also ways of utilizing multiple channels on a single controller to gain significant speed advantages. To hit the 66MHz 64 bit limit of 512MB/s data transfer would take roughly 9 striped drives for a single call. For a SCSI U-320 bus, it takes about 5-6 drives to fill up a single channel. (Note, this is for a single read/write operation, multiple concurrent read/writes add additional complexity to the tests)
Lasltly, this article utilizes the most expensive SCSI 320 products in a test that much lower hardware could have smoked with ease. Any 4 10K or 15K SCSI drives could have gained the same performance advantages over SATA as the uber SCSI drives, and at a fraction of the cost. What you gain with U-320 is not speed in the sense that the drives are faster, but the max number of drives on a single channel before reaching saturation along with slightly faster messaging speeds.
So, basically, SATA appears to be fine for most single-user desktops, but for real servers, use SCSI. One of these days, I'll actually do a real benchmark test of my own, maybe.... Except I don't have SATA RAID hardware to compare against.
How about a regular program? ie, I don't want to feed Big Brother, so I run some program that spoofs compliance by listing completely arbitrary programs running - like BeOS and nothing else.
Using series to approximate the solution of differntial equations is taught in class. Heck, go a little further in mathematics and you'll conjure up polynomials functions as the solution to a set of partial differential equations, known as the Galerkin Method
So in what way is the above news? (Hint, take a look at the link and what's stated there.)
It should be the job of the programming manager to create the programming. This would include adding new material to the list, as that would be part of the job of managing the damn playlist.
Pushing it off on the listeners is a cop out. If that were the case, we'd just get Britney Spears latest 24 hours a day (who else other than 12 y-olds have time to dial into the radio station day in and day out?) Not to mention that if it becomes a listener voting system, in an altruistic world only those listening to a station would vote for their music, and the majority rule would generally drive the minority away, decreasing diversity even further, as the minority can leave, as there are many other options, especially outside of radio regarding music. (In the real world, small bands of script kiddies would work on 0wnX0r a local station)
Music was around before radio, and will survive in spite of radio. The RIAA's spiel of P2P killing the music biz is true - it's killing the biz part, but the music's just fine, thank you. The biz was there to capitalize on a short coming - the lack of a distribution mechanism available to artists. That's no longer true, and will become less true as time goes on. As the value of the biz's service decreases, we can only hope that both the current monopolistic practices, stranglehold, and inflated prices will all fail.
I'd agree 100% with you - ATA or SATA RAID is the consumer desktop version of RAID. It has no business in production systems (production to me means 24/7/365, anything less is pretty much experimental or in test mode until it becomes "serious";). When setting up production hardware, you spec the hardware to the software you're going to use, not the other way around, and certainly not to be able to state "we're using SATA RAID, we're better than those old SCSI RAID setups, so we're worth the extra $10 you spent". BTW, I've noticed that a real SATA setup will cost as much or more than an equivalent enterprise SCSI RAID setup in size, while being less capable from high performance standpoint.
I'll clarify the above statement: SATA on a price/performance standpoint will deliver less total performance for the dollar, while it might deliver more storage for the dollar.
300GB SATA drives start at ~$235 (7200 rpm)
~150GB SCSI drives start at ~$250 (10K)
73GB SCSI drive starts at $115 (10K)
36GB SATA drive starts at... $115. (10K)
73GB SCSI drives are available at 15K starting at ~$350
SATA drives max at 10K rpm, and 74GB at that max (starting price $178)
So, looking at the above, unless you've got a large single file storage system or low disk concurrent disk usage application, SCSI is still the most appropriate system to use for server apps. (not to mention I don't know if SATA finally really addresses the achilles heel of ATA's concurrent access lockout). There is also the issue of space, of course, but if you're not trading performance for large storage, the space requirements are the same regarding number of disks, and SCSI will beat SATA hands down on both price and resulting performance.
You can, there's more to it than that though - for many higher end RAID controllers, you can save and restore the array configuration to... a floppy. This is from the RAID array BIOS program.
Why, you ask? Well, RAID BIOS can talk to the int 13 device, but not a CDROM or the like. Therefore - no USB, no CD-xxx, no DVD+-xx, just good ole floppies. (There's no OS running at this point, so no drivers to load - your OS is on that RAID controlled stripe set anyways)
And if you're ever playing around with array configurations and screw something up, or the gods of electricity hiccup during a configuration change, you'll be very happy that you backed up that config on a floppy.
It's a easy, quick, cheap way to boot/repair/replace configs on those non-essential things like SCSI controllers, especially those low-dollar RAID controllers. (Necessary through at least 2002's batch of RAID controllers, if they've progressed beyond the floppy for config since then, I don't know, as I don't own any newer RAID controllers.)
Speaking of which, this is one area that IDE RAID controllers, at least those cheaper ones, absolutely fail on. Lose your config on those, you may as well flip a coin as to whether you'll recover that striped set or not.
And OLE was based on several patents from IBM, to which they were granted licensing rights during the smooze fest prior to MS striking out on their own.
because they're very very specific functions. GPUs on nVidia's latest cards have 16 pipelines that run concurrently IIRC. That's a 16 fold increase in gigaflops possible over current CPUs, which only process a single pipeline.
More choices is good, especialy for a lightweight word processor. This article at least makes me want to look at this word processor, although I found the article itself a little light (no real criticisms, which I find peculiar because of the nature of word processors, which always have quirks/issues).
Slashdot Mirror
It does well in a single portion of the server benchmark space, and the real question is, how long will it continue to hold that lead when Itanium's future dev looks relatively bleak in contrast to AMD's push?
Multi-cores and NUMA on the desktop? Desktop pricing pressures on server features? That price point drops like a rock, can Intel still sell Itanium at anything approaching a reasonable price in 1 year that customers will pay that allows intel to reap any profit whatsoever?
I'll predict it here - Intel has lost this battle, and in 2 years, Itanium will be a footnote in the processor space, much like alpha and MIPS. Not only that, but Intel's going to be much smaller in the CPU space unless they can come up with a rabbit to sell in the 64 bit space. I don't think rumors of the P6 dual core driven processors is going to cut it either, in light of AMD's dual core opterons...
Some of us tried and tried to support alternative software way back when. MS killed all chances of doing so. I was running OS/2 with wordperfect, several other OS/2 apps including compilers and my own code. All for naught when MS Office 97 came out and had '0' backwards compatibility and the execs had copies of 97....
If a graphics card = a mortgage payment, you're either buying one hell of a graphics card, or I want your mortgage payment!
Same issue here - have problems with internal site being written by some IE only dimwits, and what's IE only? Why, DHTML menus with drop downs, which are perfectly possible in Firefox as well, with proper coding. Oh well, life goes on, I suppose.
Give me a break.
You're surprised? You give the rival's website as the only example? You'd expect a site that depends most likely on security holes, err, features, in IE to work with a real browser?
- to log into the client to access anything at all on the laptop (can be gotten around, I know)
- the VPN client not having the capability to save the password or the user name. (Quite annoying if you have a long name as the basis for a 20 character username....)
- a 5 digit PIN required as a prefix or suffix to your SecureID number. (varied by account)
- if you failed to log in 5 times, the VPN account was locked
Yep, seems like a breeze to break into a VPN with those restrictions....SecureID just seems like the next logical step. I used one for 3 years, and, once you get used to not attempting to log into your VPN when only the last bar is showing (there's a countdown bar indicating how much time is left before the number changes) it's really not so bad.
They appear to run on pseudo random number generators, and are synched up with the server with a known seed. I imagine they'd be very difficult to crack, as our system was configured to only allow 1 login attempt per number, if you typed in the wrong password/SecureID number, you had to wait until the next number came along. Annoying, but definitely better than the 3 (or 5) attempts and get a system admin to unlock your account.
Hey, don't run them on my XP box either, and it runs a whole lot better without all that crap. However, let's be fair and say that it is behind a firewall on a private network, and has no MS software installed on top of the OS. So - no macro viruses via office documents, no web bugs/viruses via IE, and no virus of the day via Outlook. Amazing, that. Also, all executable downloads are run through a virus checker first, on another machine.
Actually, I have had 2K and XP here at work. Since "upgrading" to XP, I've had to reboot at least every 4-6 days, because really odd things start happening. BTW, these are IT managed systems, so I would assume they're equally managed, as it's the same folks managing them.
Now, I have a home XP machine that's up and running generally around 4-5 months between reboots. And that's even playing a couple of games on those (not done @ work of course). I should mention that autoupdate is turned off, among other things, and it lives isolated behind a firewall and runs no MS software. :) The 2K machine stays up continuously, running as a webserver with apache.
Hey, I just dug back down to the desk level view here over the past few weeks. Quite an effort, and will probably result in the offloading of a bunch of crap to ebay....
If true - we'd have any game worth playing on Linux or Macs, and life would be good, most likely, too good to be true.... :(
Those retail prices for SATA disks are pulled from Pricewatch. Yes, you can find special sales, sales with rebates, etc, that drop the price, but people putting together RAID systems generally don't want to wait over a period of 5 months to get all the drives they need for their RAID systems.
Check out this quick review where premium U-320 4 drive array spanks SATA in every test, and this was merely for RAID 0. That's no extra processing overhead. Tom's Hardware has a more thorough article, but is heavily slanted in SATA's favor. I especially love this quote:
Yes, it sounds incredible to anyone that knows anything about how to setup SCSI RAID. First, there's not a drive made that comes anywhere near 150 MB/s continuous transfer speed, only the data in the buffer could theoretically be transferred at that speed, so that red herring is pretty smelly already.
Second, the configuration issue is definitely something to be considered with SCSI. Proper setup will actually have SCSI hitting near its theoretical peaks data transfer more consistently than not. There are also ways of utilizing multiple channels on a single controller to gain significant speed advantages. To hit the 66MHz 64 bit limit of 512MB/s data transfer would take roughly 9 striped drives for a single call. For a SCSI U-320 bus, it takes about 5-6 drives to fill up a single channel. (Note, this is for a single read/write operation, multiple concurrent read/writes add additional complexity to the tests)
Lasltly, this article utilizes the most expensive SCSI 320 products in a test that much lower hardware could have smoked with ease. Any 4 10K or 15K SCSI drives could have gained the same performance advantages over SATA as the uber SCSI drives, and at a fraction of the cost. What you gain with U-320 is not speed in the sense that the drives are faster, but the max number of drives on a single channel before reaching saturation along with slightly faster messaging speeds.
So, basically, SATA appears to be fine for most single-user desktops, but for real servers, use SCSI. One of these days, I'll actually do a real benchmark test of my own, maybe.... Except I don't have SATA RAID hardware to compare against.
How about a regular program? ie, I don't want to feed Big Brother, so I run some program that spoofs compliance by listing completely arbitrary programs running - like BeOS and nothing else.
Using series to approximate the solution of differntial equations is taught in class. Heck, go a little further in mathematics and you'll conjure up polynomials functions as the solution to a set of partial differential equations, known as the Galerkin Method
So in what way is the above news? (Hint, take a look at the link and what's stated there.)
It should be the job of the programming manager to create the programming. This would include adding new material to the list, as that would be part of the job of managing the damn playlist.
Pushing it off on the listeners is a cop out. If that were the case, we'd just get Britney Spears latest 24 hours a day (who else other than 12 y-olds have time to dial into the radio station day in and day out?) Not to mention that if it becomes a listener voting system, in an altruistic world only those listening to a station would vote for their music, and the majority rule would generally drive the minority away, decreasing diversity even further, as the minority can leave, as there are many other options, especially outside of radio regarding music. (In the real world, small bands of script kiddies would work on 0wnX0r a local station)
Music was around before radio, and will survive in spite of radio. The RIAA's spiel of P2P killing the music biz is true - it's killing the biz part, but the music's just fine, thank you. The biz was there to capitalize on a short coming - the lack of a distribution mechanism available to artists. That's no longer true, and will become less true as time goes on. As the value of the biz's service decreases, we can only hope that both the current monopolistic practices, stranglehold, and inflated prices will all fail.
I'd agree 100% with you - ATA or SATA RAID is the consumer desktop version of RAID. It has no business in production systems (production to me means 24/7/365, anything less is pretty much experimental or in test mode until it becomes "serious";). When setting up production hardware, you spec the hardware to the software you're going to use, not the other way around, and certainly not to be able to state "we're using SATA RAID, we're better than those old SCSI RAID setups, so we're worth the extra $10 you spent". BTW, I've noticed that a real SATA setup will cost as much or more than an equivalent enterprise SCSI RAID setup in size, while being less capable from high performance standpoint.
I'll clarify the above statement: SATA on a price/performance standpoint will deliver less total performance for the dollar, while it might deliver more storage for the dollar.
- 300GB SATA drives start at ~$235 (7200 rpm)
- ~150GB SCSI drives start at ~$250 (10K)
- 73GB SCSI drive starts at $115 (10K)
- 36GB SATA drive starts at
... $115. (10K)
- 73GB SCSI drives are available at 15K starting at ~$350
- SATA drives max at 10K rpm, and 74GB at that max (starting price $178)
So, looking at the above, unless you've got a large single file storage system or low disk concurrent disk usage application, SCSI is still the most appropriate system to use for server apps. (not to mention I don't know if SATA finally really addresses the achilles heel of ATA's concurrent access lockout). There is also the issue of space, of course, but if you're not trading performance for large storage, the space requirements are the same regarding number of disks, and SCSI will beat SATA hands down on both price and resulting performance.You can, there's more to it than that though - for many higher end RAID controllers, you can save and restore the array configuration to... a floppy. This is from the RAID array BIOS program.
Why, you ask? Well, RAID BIOS can talk to the int 13 device, but not a CDROM or the like. Therefore - no USB, no CD-xxx, no DVD+-xx, just good ole floppies. (There's no OS running at this point, so no drivers to load - your OS is on that RAID controlled stripe set anyways)
And if you're ever playing around with array configurations and screw something up, or the gods of electricity hiccup during a configuration change, you'll be very happy that you backed up that config on a floppy.
It's a easy, quick, cheap way to boot/repair/replace configs on those non-essential things like SCSI controllers, especially those low-dollar RAID controllers. (Necessary through at least 2002's batch of RAID controllers, if they've progressed beyond the floppy for config since then, I don't know, as I don't own any newer RAID controllers.)
Speaking of which, this is one area that IDE RAID controllers, at least those cheaper ones, absolutely fail on. Lose your config on those, you may as well flip a coin as to whether you'll recover that striped set or not.
And OLE was based on several patents from IBM, to which they were granted licensing rights during the smooze fest prior to MS striking out on their own.
"I can see the colors of the sounds I hear"...
because they're very very specific functions. GPUs on nVidia's latest cards have 16 pipelines that run concurrently IIRC. That's a 16 fold increase in gigaflops possible over current CPUs, which only process a single pipeline.
At least, I'm guessing that's how that goes.
It doesn't have GHz after it, so it must be slow!
More choices is good, especialy for a lightweight word processor. This article at least makes me want to look at this word processor, although I found the article itself a little light (no real criticisms, which I find peculiar because of the nature of word processors, which always have quirks/issues).