Stop posting stories about single drive failures, they are irrelevant, instead go to Storage Review and check out the reliability database. With the exception of the Deathstars and some problem models from other manufacturers most IDE drives have decent failure rates that generally seem to jibe with the manufacturers MTBF. And if you really care add all your current drives and all future purchases to the reliability survey and go back when they die or go out of service. I have a total of 10 Maxtor's in the reliability database from my personal systems and none of them are dead, still statistically irrelevant but it does help add to the data pool. I also had a batch of ~100 Hitachi laptop HDD's where nearly every one died within a year, now THAT was an obviously bad batch and not statistically insignificant, but most people aren't reporting those kinds of results.
After working at a number of Select liscensees and working for a couple MS partners I have come to the conclusion that MS liscensing is impossible to figure out. If you ask your select contact you will get a different answer from the sales exec which will be different from the select liscensing manager which would probably be different from the BSA audit's conclusions. MS liscensing contacts run to over 100 pages of small print legalese that isn't even necessarily internally constistant.
Trust me if they break CIFS there will be MAJOR problems for them. For one their large customers would have to coordinate upgrading ALL desktops systems and ALL servers. Beyond that almost no medium or larger corporations run a purely MS shop. SAMBA and other CIFS implementations like Netapp filers would be affected and MS would have so LOUD cries from many of their large enterprise customers.
How many times must this myth be dispelled? NFS is perfectly secure if you use secure transport and authentication both of which are available on NFSv3 and higher. Linux may or may not have a good implementations of them but Solaris and other Unix variants do. Using NFSv3 and yp+ we had a secure filesharing system with universal login and automouting of things like home drives and compiler paths.
The simple way to do it is to continuously pump the waste heat into the heating ducts during the winter and have the main heating system on a zoned thermostat system. That way your main heating system is only producing (n-m) btu's where n is its normal btu output and m is the waste heat output from the cluster. If you later remove m then the system still has the ability to produce n btu's. Besides that most datacenters have INCREASED in btu output and power consumption over the years not decreased. The old systems were large and hot but the new systems are small and REALLY hot. Trust me I have been in quite a few datacenters that have had their HVAC systems upgraded from the days of room sized mainframes.
No, ECC ram typically is just made with faster internals. As an example most ECC comodity ram is CAS2 latency whereas most generic ram is CAS3, so the ECC ram will perform exactly the same as the non-ECC ram. You can buy CAS2 non-ECC ram but it's nearly as expensive as the ECC ram. If you have a simple idiot check at the end of a complex calculation then saving the cost of going with ECC may be worth it but most clusters this large will be used on too many different projects to assume that all of them will have such checks. For an idea of how important ECC is read (a href="http://www.ibm.com/servers/eserver/pseries/c ampaigns/chipkill.pdf">This IBM whitepaper on their chipkill ECC scheme. Even normal SEC ECC ram (what most ECC ram is today) will have aproximately 900 failures per 10TB per three years. I think that IBM is right and that eventually all ram will be RAID-M, that is a RAID5 style array of redundant memory banks that are composed of ECC banks. At future densities this will be necessary because a single high energy particle will have the ability to scramble an entire memory word including it's ECC checking bits.
If you want a complete solution including site survey capabilities, security audits, etc then the best solution I have seen is airmagnet's handheld solution. It worked well even in Cisco Aironet's office which is possibly the most wifi busy location on the planet. I would definitly suggest you take a look.
You're talking about multi-thousand Watt radar installations vs 100mW max wifi cards. I worked in a building with hundreds of wifi cards and base stations for years and suffered no effects. Many of my clients had been working on wifi and its predecessors for decades and none of them suffered ill effects (out of 185 none suffered from cancers, statistically lower than the national average). I wouldn't advicate standing in front of radar stations as good for your health but I have seen ZERO scientifically valid reports or papers showing a linkage between wifi/cellphone/other low power RF sources and cancer. It's like the high voltage powerline scare a decade ago, one badly done statisticall sampling was disproved by several hundred followup scientific studies. Cancer is the modern boogyman, it's horrible and with the exception of smokers it is often unexpected so people look for something to blame. The fact is that human DNA is not meant to withstand the environment as long as we live today.
compare 2x21" monitors vs say a 35" plasma display and you will see a real benifit. When I was supporting engineers at Cisco almost every single one of them had 2x21" Trinitron's. Some of them ran CAD apps where they would have their work area on one monitor and the vis window on the other. Others had their windows stuff on one monitor and their Unix X displays on the other. The cost of getting a display that could do 4K*3K pixels is insane, but the cost of a couple of good 21" monitors is well under $1K, when you are paying an engineer 6 figures spending an extra couple hundred on making him more productive is a no brainer.
That's why good headhunters place people in a contract to hire capacity if possible. They get a minimal compensation when the employee is under temporary contract but typically get a percentage of the first year salaray if the employee is placed permenantly (paid by the employer). This is how I was placed through Manpower Professional.
Then Congress is applying the same constitutional test that the Supreme Court has used time and again. Commercial speech IS less protected than other forms such as political, religious, or private speech.
You must be one of the old geezers that hangs out around here occasionally. For kids Xmas is THE most important day of the year. Plus for many people the production and selling season leading up to the holidays is the time they are most likely to get overtime, also many companies give a holiday bonus. So yes, the holidays are an important time of year for consumers as well as retailers.
The parity calculations are computed as if there was one parity disk and then they are distributed across all the disks. The parity blocks are the same size as the data blocks and as long as only one disks dies the other disks will contain all but one of the data blocks plus the the parity block (on average, any particular file may be missing a data or parity block). There is also another newer form of RAID that some vendors called RAID-6 which writes two disk failures per array but at the price of losing two disks worth of capacity to parity data, in real world experience I would say that is overkill because most situations that would lead to more than one HDD per array dying will kill most/all of them. This page's illustration might make it a bit more clear for you.
In a VoIP house, the telecoms kick back, take their money, and only worry about the customers having a solid data connection from the main office to whatever endpoint the connection is going to.
That's basically all they do for any major customer. With a PBX system they just provide the T-1 circuit and setup the billing codes, everything else is automated. Basically you can get voice grade SLA's on your data lines, hell our SLA's were better than regulated voice line standards, we always had a telecom engineer out the same day, regulation is 2 days.
Sure, power the phones with PoE and keep the switching gear on UPS's and/or generators. That's basically how the POTS or PBX systems work anyways, your phone is powered by the voltage supplied from the central office or PBX switch. VoIP phones do suck about 4x as much power as dumb phones but so what?
Re:Implementation or spec?
on
Athlon 64 Debuts
·
· Score: 2, Informative
No, PAE does not allow a single program to access more than 4GB, only the OS. So while they were stupid to add the physical portion to the address space tag they were basically correct. How many people that run applications that need 4GB or more of ram run more than one at a time? Basically anyone who needs more than 4GB needs it available to just one app so PAE is worthless in most cases.
PC was never popularyly synonomous with the terms it is derived from, but rather short for IBM compatible PC. Go back through trade press and that is almost always the context it is used in. Sorry but the Mac lost out in the popular culture a LONG time ago. I love OSX and if I couldn't build my own PC's so cheaply I might even own an Apple to run it, as it is I just play with the one in the compatibility lab =)
Re:Threat to Athlon64: Prescott (not Pentium 4)
on
Athlon 64 Debuts
·
· Score: 5, Insightful
Actually the 64 bit instructions from Prescott ARE x86-64 which is AMD's extensions. Intel has a liscense for x86-64 and has been working on project Yamhill for some time to integrate the x86-64 instructions into the P4 core architecture. They did this when they Microsoft agreed to make Windows for x86-64 as well as IA64, MS will NOT keep three forks of their codebase so Intel's option was so ceed the low end to AMD or use their liscense, they of course chose the latter since almost noone has installed Itanium systems. Btw Intel chips have been faster than Sparc for some time, people don't buy Sun machines for number crunching they buy them for stability and scalability. Until there is an Intel machine that can scale like the 6800 and is as stable as Sun hardware there will be a niche for Sun. Of course Oracle with it's RAC initiative may reduce the scalability argument.
Typical cards have a max transmit power of 50mW and the good ones (like Cisco's) have a max of 100mW. I'm not sure if Poland's entry into the EU means that they fall under ETSI regulations, but if it does this most certainly was NOT legal as the ETSI standard limits transmit on cards to 50mW and point to point links are even more limited to than the US (which is 24dBi gain on a 100mW card).
This would not be legal in the US based just on the antenna which is 27dBi, the max gain you can have with a 100mW card is 24dBi. In addition they used a 500mW linear amplifier which makes the output power many times more than the FCC allowed limit and probably creates all sorts of nasty noise on adjacent frequencies.
Now I don't program stuff on this level, but isn't that kind of what IBM's MQ Series stuff is all about, making sure things talk nicely across platforms and between systems?
Obviously an EE in training =) Have you ever tried to route a complex multilayer PCB design? If you have then you will know that it would be basically impossible to guarentee all straight paths between the CPU and RAM, or any other component. Besides if you want fast ram you put it on or near the CPU die. Hence processors like the Xeon, Itanium, HP PA-8800, etc which derive most of their performance gains over their desktop competitors by having large L2 and huge L3 caches.
Tell that to companies like eBay, buy.com, Amazon, or Progressive. These companies absolutly rely on their websites, for them web development IS the enterprise application. And hacking together something in PHP, Perl, whatever just doesn't cut it.
Slashdot Mirror
Stop posting stories about single drive failures, they are irrelevant, instead go to Storage Review and check out the reliability database. With the exception of the Deathstars and some problem models from other manufacturers most IDE drives have decent failure rates that generally seem to jibe with the manufacturers MTBF. And if you really care add all your current drives and all future purchases to the reliability survey and go back when they die or go out of service. I have a total of 10 Maxtor's in the reliability database from my personal systems and none of them are dead, still statistically irrelevant but it does help add to the data pool. I also had a batch of ~100 Hitachi laptop HDD's where nearly every one died within a year, now THAT was an obviously bad batch and not statistically insignificant, but most people aren't reporting those kinds of results.
After working at a number of Select liscensees and working for a couple MS partners I have come to the conclusion that MS liscensing is impossible to figure out. If you ask your select contact you will get a different answer from the sales exec which will be different from the select liscensing manager which would probably be different from the BSA audit's conclusions. MS liscensing contacts run to over 100 pages of small print legalese that isn't even necessarily internally constistant.
Trust me if they break CIFS there will be MAJOR problems for them. For one their large customers would have to coordinate upgrading ALL desktops systems and ALL servers. Beyond that almost no medium or larger corporations run a purely MS shop. SAMBA and other CIFS implementations like Netapp filers would be affected and MS would have so LOUD cries from many of their large enterprise customers.
How many times must this myth be dispelled? NFS is perfectly secure if you use secure transport and authentication both of which are available on NFSv3 and higher. Linux may or may not have a good implementations of them but Solaris and other Unix variants do. Using NFSv3 and yp+ we had a secure filesharing system with universal login and automouting of things like home drives and compiler paths.
The simple way to do it is to continuously pump the waste heat into the heating ducts during the winter and have the main heating system on a zoned thermostat system. That way your main heating system is only producing (n-m) btu's where n is its normal btu output and m is the waste heat output from the cluster. If you later remove m then the system still has the ability to produce n btu's. Besides that most datacenters have INCREASED in btu output and power consumption over the years not decreased. The old systems were large and hot but the new systems are small and REALLY hot. Trust me I have been in quite a few datacenters that have had their HVAC systems upgraded from the days of room sized mainframes.
No, ECC ram typically is just made with faster internals. As an example most ECC comodity ram is CAS2 latency whereas most generic ram is CAS3, so the ECC ram will perform exactly the same as the non-ECC ram. You can buy CAS2 non-ECC ram but it's nearly as expensive as the ECC ram. If you have a simple idiot check at the end of a complex calculation then saving the cost of going with ECC may be worth it but most clusters this large will be used on too many different projects to assume that all of them will have such checks. For an idea of how important ECC is read (a href="http://www.ibm.com/servers/eserver/pseries/c ampaigns/chipkill.pdf">This IBM whitepaper on their chipkill ECC scheme. Even normal SEC ECC ram (what most ECC ram is today) will have aproximately 900 failures per 10TB per three years. I think that IBM is right and that eventually all ram will be RAID-M, that is a RAID5 style array of redundant memory banks that are composed of ECC banks. At future densities this will be necessary because a single high energy particle will have the ability to scramble an entire memory word including it's ECC checking bits.
If you want a complete solution including site survey capabilities, security audits, etc then the best solution I have seen is airmagnet's handheld solution. It worked well even in Cisco Aironet's office which is possibly the most wifi busy location on the planet. I would definitly suggest you take a look.
You're talking about multi-thousand Watt radar installations vs 100mW max wifi cards. I worked in a building with hundreds of wifi cards and base stations for years and suffered no effects. Many of my clients had been working on wifi and its predecessors for decades and none of them suffered ill effects (out of 185 none suffered from cancers, statistically lower than the national average). I wouldn't advicate standing in front of radar stations as good for your health but I have seen ZERO scientifically valid reports or papers showing a linkage between wifi/cellphone/other low power RF sources and cancer. It's like the high voltage powerline scare a decade ago, one badly done statisticall sampling was disproved by several hundred followup scientific studies. Cancer is the modern boogyman, it's horrible and with the exception of smokers it is often unexpected so people look for something to blame. The fact is that human DNA is not meant to withstand the environment as long as we live today.
compare 2x21" monitors vs say a 35" plasma display and you will see a real benifit. When I was supporting engineers at Cisco almost every single one of them had 2x21" Trinitron's. Some of them ran CAD apps where they would have their work area on one monitor and the vis window on the other. Others had their windows stuff on one monitor and their Unix X displays on the other. The cost of getting a display that could do 4K*3K pixels is insane, but the cost of a couple of good 21" monitors is well under $1K, when you are paying an engineer 6 figures spending an extra couple hundred on making him more productive is a no brainer.
That's why good headhunters place people in a contract to hire capacity if possible. They get a minimal compensation when the employee is under temporary contract but typically get a percentage of the first year salaray if the employee is placed permenantly (paid by the employer). This is how I was placed through Manpower Professional.
Then Congress is applying the same constitutional test that the Supreme Court has used time and again. Commercial speech IS less protected than other forms such as political, religious, or private speech.
You must be one of the old geezers that hangs out around here occasionally. For kids Xmas is THE most important day of the year. Plus for many people the production and selling season leading up to the holidays is the time they are most likely to get overtime, also many companies give a holiday bonus. So yes, the holidays are an important time of year for consumers as well as retailers.
Baydel call's it that, and I think I remember at least one other vendor doing the same.
The parity calculations are computed as if there was one parity disk and then they are distributed across all the disks. The parity blocks are the same size as the data blocks and as long as only one disks dies the other disks will contain all but one of the data blocks plus the the parity block (on average, any particular file may be missing a data or parity block). There is also another newer form of RAID that some vendors called RAID-6 which writes two disk failures per array but at the price of losing two disks worth of capacity to parity data, in real world experience I would say that is overkill because most situations that would lead to more than one HDD per array dying will kill most/all of them. This page's illustration might make it a bit more clear for you.
Wrong calculations, that's for a 3 drive RAID-5, for a 6 drive array it's 5/6ths or 1TB in this case. In general it is (N-1)*capacity.
In a VoIP house, the telecoms kick back, take their money, and only worry about the customers having a solid data connection from the main office to whatever endpoint the connection is going to.
That's basically all they do for any major customer. With a PBX system they just provide the T-1 circuit and setup the billing codes, everything else is automated. Basically you can get voice grade SLA's on your data lines, hell our SLA's were better than regulated voice line standards, we always had a telecom engineer out the same day, regulation is 2 days.
Sure, power the phones with PoE and keep the switching gear on UPS's and/or generators. That's basically how the POTS or PBX systems work anyways, your phone is powered by the voltage supplied from the central office or PBX switch. VoIP phones do suck about 4x as much power as dumb phones but so what?
No, PAE does not allow a single program to access more than 4GB, only the OS. So while they were stupid to add the physical portion to the address space tag they were basically correct. How many people that run applications that need 4GB or more of ram run more than one at a time? Basically anyone who needs more than 4GB needs it available to just one app so PAE is worthless in most cases.
PC was never popularyly synonomous with the terms it is derived from, but rather short for IBM compatible PC. Go back through trade press and that is almost always the context it is used in. Sorry but the Mac lost out in the popular culture a LONG time ago. I love OSX and if I couldn't build my own PC's so cheaply I might even own an Apple to run it, as it is I just play with the one in the compatibility lab =)
Actually the 64 bit instructions from Prescott ARE x86-64 which is AMD's extensions. Intel has a liscense for x86-64 and has been working on project Yamhill for some time to integrate the x86-64 instructions into the P4 core architecture. They did this when they Microsoft agreed to make Windows for x86-64 as well as IA64, MS will NOT keep three forks of their codebase so Intel's option was so ceed the low end to AMD or use their liscense, they of course chose the latter since almost noone has installed Itanium systems. Btw Intel chips have been faster than Sparc for some time, people don't buy Sun machines for number crunching they buy them for stability and scalability. Until there is an Intel machine that can scale like the 6800 and is as stable as Sun hardware there will be a niche for Sun. Of course Oracle with it's RAC initiative may reduce the scalability argument.
Typical cards have a max transmit power of 50mW and the good ones (like Cisco's) have a max of 100mW. I'm not sure if Poland's entry into the EU means that they fall under ETSI regulations, but if it does this most certainly was NOT legal as the ETSI standard limits transmit on cards to 50mW and point to point links are even more limited to than the US (which is 24dBi gain on a 100mW card).
This would not be legal in the US based just on the antenna which is 27dBi, the max gain you can have with a 100mW card is 24dBi. In addition they used a 500mW linear amplifier which makes the output power many times more than the FCC allowed limit and probably creates all sorts of nasty noise on adjacent frequencies.
Now I don't program stuff on this level, but isn't that kind of what IBM's MQ Series stuff is all about, making sure things talk nicely across platforms and between systems?
Obviously an EE in training =)
Have you ever tried to route a complex multilayer PCB design? If you have then you will know that it would be basically impossible to guarentee all straight paths between the CPU and RAM, or any other component. Besides if you want fast ram you put it on or near the CPU die. Hence processors like the Xeon, Itanium, HP PA-8800, etc which derive most of their performance gains over their desktop competitors by having large L2 and huge L3 caches.
Tell that to companies like eBay, buy.com, Amazon, or Progressive. These companies absolutly rely on their websites, for them web development IS the enterprise application. And hacking together something in PHP, Perl, whatever just doesn't cut it.