I HATE straight pipes, and under the same regulations that outlawed glass packed pipes in the 60's and the regulations used to pull over the idiot with the thumping bass box they should be illegal in most jurisdictions. That said the only time I wished I had followed up on that urge to get a carry conceal permit was to defend a straight pipe riding idiot from some infinitly more idiotic SUV driving morons. The particular move that enraged me was that after exchanging words the SUV driving idiots swerved towards the guy on the bike who was in the break down lane in order to get out of their way. In my view that was attempted vehicular homicide and idiocy of the first degree.
Re:lack of pulsatile flow and coronary vessles
on
Living Without a Pulse
·
· Score: 3, Insightful
Uh, isn't the entire point of these pumps to REPLACE the damaged heart? I mean if you have a low maintenance pump with little chance of failure why are you relying at all on the failing/failed natural heart?
You missed his point. In a typical 8 drive RAID5 system only 7 drives are activly in the array, the eighth is left out as a hot spare. Then you add snapshot overhead (easily 20% of formatted capacity for only a couple generations of snapshots) and suddenly you are at the equivilant of about 3.5 drives worth of capacity usable for the live file system, or around 43%. If you assume a little more for the snapshots then you easily get down to his 40% figure. If you are using the array for purely archival purposes you might not need snapshotting but if you are like most normal people that wicked fast array will quickly become your primary storage site, so you will want snapshoting for when you overwrite that cv or for when you type rm -rf in the wrong window.
Uh, real latency numbers are about 3ns for modern ECC memory and about 15-60ms for Infiniband depending on message size, so not hundreds of times but rather 20 times worst case. That's still not good if you are message latency limited but you have to remember that crossbars and cross cabinet interconnects have latency as well.
In the worst case (very large messages) the latency between nodes in the VT cluster is about 3X worse than the latency in this SGI behemoth. Unless you are solely limited by message latency then you are usually better off with the VT style cluster since it is about half as much per flop. Btw basically ALL of the money going into researching problems that require traditional supercomputers is being done by governments, so all of the money is already coming from government work.
Volume Shadow Copy is a snapshotting technology, NOT an automatic versioning technolgy such as found in VMS. It's very cool but if you want to see it done right you have to look at a NetApp Filer, MS's implementation only allows you to keep the last x snapshots, NetApp's allows you to keep them based of date and time based rules.
I propose a new measure of chip perfomance Specfp_base_2000/Watt thermal load. That should give you a good idea of a chips potential for use in massive supercomputer applications. Sadly the Opteron does worse than both the Itanium 2 and the PPC970 in this measurement.
Another common method was to look for mono audio tracks. Most programming was in stereo whereas the ads were in mono. Advertisers caught onto this one when a couple VCR manufacturers came out with VCR's that used this trick to mark ads, so now almost all national ads are in stereo, even if it's really just a doubling of the mono signal.
Ugh, I can't stand paying for a cable box, besides I have PiP and so I would need TWO boxes with some way to keep them from responding to the same remote. I just go with basic non-digital cable and use the built in tuners. However I think you are right about the speakers, a good shielded center channel will be many times better than the built in speakers.
That's already been done. Some other "journalist" got people all over the net to sign up a gmail account to as many spam lists as possible, it took like 10 weeks to fill it with nothing but HTML and text spam.
hardly, try a Netapp with hourly snapshot's and snapmirror to a second unit across the WAN with backup to a robotic library. Yeah it costs a bit more but if the data is really worth $100K then it's worth protecting as such. It also works with databases or Exchange server stores. They have some unbeatable features (like in a cluster the systems can take over the personality of the other unit if it crashes and mount and share its disks, they can also use hot spares from the other part of the cluster).
I had always looked down on Hyundai's until about three years ago I drove a Sonata out west. Cruise controll at 90, AC cranked to the max, never broke a sweat which is much better than a bunch of the cheaper Japanese cars I've driven out there. Not quite as large as my Taurus but since I only plan to have two kids that shouldn't be a problem. Fit me fine and I'm 6'3 220lbs. The guy at the rental company said they have fewer problems with the Hyundai's but they use the cheap Taurus with the old Vulcan V6 so I'm not suprised.
Yeah well since the GPS watch is lighter than any other GPS on the market and I was hiking in the Alps and the Andes I didn't mind spending a bit, plus it had the altimeter which is a very good way to know of an impending storm front. When you get into extreme mountaneering you get used to paying for good equipment, that equipment has saved my life on more than one occasion so it was worth every penny.
Slashdot has been 500 and 530'ing all over itself ever since the code "upgrade" last week. I think it's mostly the DB backend blowing up (go MySQL) but it might be bad code causing the Apache/Pearl process to blow up.
Altavista still has a feature that Google lacks, the NEAR keyword. This allows you to search for two terms only when they apear withing about two paragraphs of each other. This can be extremely usefull for searching for two semi-popular terms that would not normally be found in the same article but where such occourances might be of interest. Unfortunatly they have so little of the net indexed these days that their results suck. Add to that the placed rankings and the meta tag bombs and I generally don't bother, I just wish Google would add the same structure to their queries.
requires deep pockets??? My first PC cost $1500 without the printer and that was 1993 dollars. Hell my non-geeky watches usually cost me around $300 for a decent one. My GPS, altimiter, electronic compass one was closer to $500 and that doesn't include email =)
1) If the stock pays dividends (Google won't, AFAIK), the more stock you have, the more dividends you get, so 100 shares of stock at $10 would be better than 10 shares of stock at $100.
Bullocks, the dividend is calculated by the board deciding what the gross profit is and deciding how much of that they wish to return to the investors. That lump of money is then devided by the number of outstanding shares. Microsoft didn't decide they were going to give out their $72B by deciding on a dividend per share then working backwards to the total valuation of the dividend, they calculated what they could afford to pay back then devided it by the number of outstanding shares and broke it up further by spreading the payments over multiple quarters with a large initial payout to reward the investors of record at the time of the announcement.
P.S. the S&P's P/E is VERY high historically, the 50 year historical average is only 16.48 linky
Try historic P/E. Even then it won't always give you the true story. For instance most auto companies historically trade at around an 18, but when they are bleeding money their P/E is N/A because they aren't making money, then you have to look at burn rate vs stockpile and potentially market share movement (since auto manufacturing is a volume game significant losses in market share can lead to increased cost of capital investments per unit which the company might never fully recover from). Aside from any of this info the real answer is a tracking mutual fund, if you need to ask such basic questions then you will NEVER beat the market over a statistically significant period of time so why not just ride the market in a vehicle which has low overhead costs.
Well the market cap isn't going to be $36.2B, that's the valuation of the company. The 24.6M shares are only an ~10% equity in the company, much of the company is held by employees and venture capitalists. The whole reason for the IPO is that the number of preivate shareholders in the company had grown large enough that SEC rules were going to force Google to open their books in the same manner as a publicly traded company so there was no reason NOT to IPO and allow some of the stockholders to convert some of their holding to liquid assets and raise some additional capital for the company. The reason not to dilute the shares more than offering a 10% equity stake is that the company does not need more capital than they are asking for and more shares would just dilute the value of the shares already privatly held.
Actually, yes Google DOES hold their entire dataset in ram. The HDD's are just there to repopulate ram should they lose power. If IDE HDD's weren't basically free they would probably just net boot the boxes and read the data from central storage. Each Google box is responsible for a dataset that is exactly the size of its ram minus the minimal overhead for the OS. They get redundancy by holding the entire database in chunks each of which is held on several machines at each datacenter and each datacenter has an entire copy of the whole dataset. Hell AFAIk there IS no central copy of the database, just lots of automagically updated data chunks at multiple datacenters worldwide.
storage review says about 64MB/s on the outside of the platter for the best performing 15K RPM drive (which is a 74GB drive not a 140GB one). So, to swamp an U320 bus with sustained transfers you will need at least 6 drives, not the 3 that some people keep spouting around here. So if you need 6 drives to saturate the bus, why have a few high capacity drives when more drives gives you lower latency and gets you to max sustained transfer.
HP high end servers DO in fact use redundant systems to check every calculation and every data transaction. Every calculation is done by two cpu's and if the results aren't identical it is run again, if they still aren't the same both CPU's are taken offline and the thread is migrated to another part of the system. Likewise all data path's are redundant and ECC'd. This costs LOTS of money, but if you are paranoid about data corruption/loss then that's what you pay for. Btw controllers going crazy isn't all that uncommon, every field tech I've met has at least one story of a crazy RAID controller necessitating going to tape, often from weeks before the repair call since the controller had been silently failing for some time before the corruption was noticed. Luckily I've never heard of it happening on a DB server, that would be one heck of a mess.
Slashdot Mirror
I HATE straight pipes, and under the same regulations that outlawed glass packed pipes in the 60's and the regulations used to pull over the idiot with the thumping bass box they should be illegal in most jurisdictions. That said the only time I wished I had followed up on that urge to get a carry conceal permit was to defend a straight pipe riding idiot from some infinitly more idiotic SUV driving morons. The particular move that enraged me was that after exchanging words the SUV driving idiots swerved towards the guy on the bike who was in the break down lane in order to get out of their way. In my view that was attempted vehicular homicide and idiocy of the first degree.
Uh, isn't the entire point of these pumps to REPLACE the damaged heart? I mean if you have a low maintenance pump with little chance of failure why are you relying at all on the failing/failed natural heart?
You missed his point. In a typical 8 drive RAID5 system only 7 drives are activly in the array, the eighth is left out as a hot spare. Then you add snapshot overhead (easily 20% of formatted capacity for only a couple generations of snapshots) and suddenly you are at the equivilant of about 3.5 drives worth of capacity usable for the live file system, or around 43%. If you assume a little more for the snapshots then you easily get down to his 40% figure. If you are using the array for purely archival purposes you might not need snapshotting but if you are like most normal people that wicked fast array will quickly become your primary storage site, so you will want snapshoting for when you overwrite that cv or for when you type rm -rf in the wrong window.
Uh, real latency numbers are about 3ns for modern ECC memory and about 15-60ms for Infiniband depending on message size, so not hundreds of times but rather 20 times worst case. That's still not good if you are message latency limited but you have to remember that crossbars and cross cabinet interconnects have latency as well.
In the worst case (very large messages) the latency between nodes in the VT cluster is about 3X worse than the latency in this SGI behemoth. Unless you are solely limited by message latency then you are usually better off with the VT style cluster since it is about half as much per flop. Btw basically ALL of the money going into researching problems that require traditional supercomputers is being done by governments, so all of the money is already coming from government work.
Volume Shadow Copy is a snapshotting technology, NOT an automatic versioning technolgy such as found in VMS. It's very cool but if you want to see it done right you have to look at a NetApp Filer, MS's implementation only allows you to keep the last x snapshots, NetApp's allows you to keep them based of date and time based rules.
I propose a new measure of chip perfomance Specfp_base_2000/Watt thermal load. That should give you a good idea of a chips potential for use in massive supercomputer applications. Sadly the Opteron does worse than both the Itanium 2 and the PPC970 in this measurement.
Another common method was to look for mono audio tracks. Most programming was in stereo whereas the ads were in mono. Advertisers caught onto this one when a couple VCR manufacturers came out with VCR's that used this trick to mark ads, so now almost all national ads are in stereo, even if it's really just a doubling of the mono signal.
Ugh, I can't stand paying for a cable box, besides I have PiP and so I would need TWO boxes with some way to keep them from responding to the same remote. I just go with basic non-digital cable and use the built in tuners. However I think you are right about the speakers, a good shielded center channel will be many times better than the built in speakers.
That's already been done. Some other "journalist" got people all over the net to sign up a gmail account to as many spam lists as possible, it took like 10 weeks to fill it with nothing but HTML and text spam.
hardly, try a Netapp with hourly snapshot's and snapmirror to a second unit across the WAN with backup to a robotic library. Yeah it costs a bit more but if the data is really worth $100K then it's worth protecting as such. It also works with databases or Exchange server stores. They have some unbeatable features (like in a cluster the systems can take over the personality of the other unit if it crashes and mount and share its disks, they can also use hot spares from the other part of the cluster).
I had always looked down on Hyundai's until about three years ago I drove a Sonata out west. Cruise controll at 90, AC cranked to the max, never broke a sweat which is much better than a bunch of the cheaper Japanese cars I've driven out there. Not quite as large as my Taurus but since I only plan to have two kids that shouldn't be a problem. Fit me fine and I'm 6'3 220lbs. The guy at the rental company said they have fewer problems with the Hyundai's but they use the cheap Taurus with the old Vulcan V6 so I'm not suprised.
Yeah well since the GPS watch is lighter than any other GPS on the market and I was hiking in the Alps and the Andes I didn't mind spending a bit, plus it had the altimeter which is a very good way to know of an impending storm front. When you get into extreme mountaneering you get used to paying for good equipment, that equipment has saved my life on more than one occasion so it was worth every penny.
Slashdot has been 500 and 530'ing all over itself ever since the code "upgrade" last week. I think it's mostly the DB backend blowing up (go MySQL) but it might be bad code causing the Apache/Pearl process to blow up.
Altavista still has a feature that Google lacks, the NEAR keyword. This allows you to search for two terms only when they apear withing about two paragraphs of each other. This can be extremely usefull for searching for two semi-popular terms that would not normally be found in the same article but where such occourances might be of interest. Unfortunatly they have so little of the net indexed these days that their results suck. Add to that the placed rankings and the meta tag bombs and I generally don't bother, I just wish Google would add the same structure to their queries.
requires deep pockets??? My first PC cost $1500 without the printer and that was 1993 dollars. Hell my non-geeky watches usually cost me around $300 for a decent one. My GPS, altimiter, electronic compass one was closer to $500 and that doesn't include email =)
1) If the stock pays dividends (Google won't, AFAIK), the more stock you have, the more dividends you get, so 100 shares of stock at $10 would be better than 10 shares of stock at $100.
Bullocks, the dividend is calculated by the board deciding what the gross profit is and deciding how much of that they wish to return to the investors. That lump of money is then devided by the number of outstanding shares. Microsoft didn't decide they were going to give out their $72B by deciding on a dividend per share then working backwards to the total valuation of the dividend, they calculated what they could afford to pay back then devided it by the number of outstanding shares and broke it up further by spreading the payments over multiple quarters with a large initial payout to reward the investors of record at the time of the announcement.
P.S. the S&P's P/E is VERY high historically, the 50 year historical average is only 16.48 linky
Try historic P/E. Even then it won't always give you the true story. For instance most auto companies historically trade at around an 18, but when they are bleeding money their P/E is N/A because they aren't making money, then you have to look at burn rate vs stockpile and potentially market share movement (since auto manufacturing is a volume game significant losses in market share can lead to increased cost of capital investments per unit which the company might never fully recover from). Aside from any of this info the real answer is a tracking mutual fund, if you need to ask such basic questions then you will NEVER beat the market over a statistically significant period of time so why not just ride the market in a vehicle which has low overhead costs.
Well the market cap isn't going to be $36.2B, that's the valuation of the company. The 24.6M shares are only an ~10% equity in the company, much of the company is held by employees and venture capitalists. The whole reason for the IPO is that the number of preivate shareholders in the company had grown large enough that SEC rules were going to force Google to open their books in the same manner as a publicly traded company so there was no reason NOT to IPO and allow some of the stockholders to convert some of their holding to liquid assets and raise some additional capital for the company. The reason not to dilute the shares more than offering a 10% equity stake is that the company does not need more capital than they are asking for and more shares would just dilute the value of the shares already privatly held.
Actually, yes Google DOES hold their entire dataset in ram. The HDD's are just there to repopulate ram should they lose power. If IDE HDD's weren't basically free they would probably just net boot the boxes and read the data from central storage. Each Google box is responsible for a dataset that is exactly the size of its ram minus the minimal overhead for the OS. They get redundancy by holding the entire database in chunks each of which is held on several machines at each datacenter and each datacenter has an entire copy of the whole dataset. Hell AFAIk there IS no central copy of the database, just lots of automagically updated data chunks at multiple datacenters worldwide.
storage review says about 64MB/s on the outside of the platter for the best performing 15K RPM drive (which is a 74GB drive not a 140GB one). So, to swamp an U320 bus with sustained transfers you will need at least 6 drives, not the 3 that some people keep spouting around here. So if you need 6 drives to saturate the bus, why have a few high capacity drives when more drives gives you lower latency and gets you to max sustained transfer.
Well since ALL of the shares to be sold to the public are for less than 50% equity in the company I don't think you have to worry about that.
HP high end servers DO in fact use redundant systems to check every calculation and every data transaction. Every calculation is done by two cpu's and if the results aren't identical it is run again, if they still aren't the same both CPU's are taken offline and the thread is migrated to another part of the system. Likewise all data path's are redundant and ECC'd. This costs LOTS of money, but if you are paranoid about data corruption/loss then that's what you pay for. Btw controllers going crazy isn't all that uncommon, every field tech I've met has at least one story of a crazy RAID controller necessitating going to tape, often from weeks before the repair call since the controller had been silently failing for some time before the corruption was noticed. Luckily I've never heard of it happening on a DB server, that would be one heck of a mess.
Thanks for the link, there was some info there but it was the worst written article I think I have ever had the displeasure of reading.
Did anyone get the text of the 6315 story? I'm really interested in how it compares to the Treo 600.