If a large group of scientists are measuring thrust from this thing, I think it is worthwhile to figure out why. Even if it is "nonsense", it may lead people to come up with a new/different space drive that is more efficient than what we have today. For instance, it may, in some novel way, just be pushing against Earth's magnetic field in a way that wasn't considered before. Or perhaps when it is tested dynamically instead of statically, the thrust falls off because the enclosure moved. It may lead us to consider certain laws of physics more from an engineering perspective than a pure physics one.
And you actually don't know what you are talking about. The Alpha's memory management is pretty similar to the x86 memory management. Yes, the x86 will not let you access memory that isn't allocated, and sometimes the OS handles this by BSOD, and in rare cases it *might* be able to recover. This would happen if that OS was running on x86, Alpha or PowerPC.
My experience with TCP/IP under Windows indicates that it is very difficult to get more that 150 MBytes/Second from a single box (and I was using a specialized app to get that speed). I don't know what kind of delays the link imposes, but if you are using TCP/IP for the benchmark, be sure to enable scalable windows (search for RFC1323 on microsoft.com).
Another way to look at this is that PCI-Express links run at 2.5 Gbits/second. If you built something that ramped PCI-Express packets to/from your link, you could have an interesting I/O technology (comparable to Infiniband).
As others have pointed out, this sort of bandwidth is comparable/greater than most disk drive subsystems. Interesting demonstration is to use something like iSCSI to talk to disk drive that are on the other side of the link, and perhaps run some disk benchmarks, etc. Current (affordable) iSCSI controllers can push 60-80 MBytes/sec, so you could put 4 or 5 workstations on one side, and multiple iSCSI/iFiber routers on the other (with a bunch of iFiber disks).
This is just goes to show how little people pay attention to books and libraries. Libraries have had an *open* standard, called Z39.50, to access databases containing information about books, for years. It is a standard based on MARC records, and just about every commercial library cataloging program is based on it, and can read/write these records. The Library of Congress provides full access via the Z39.50 interface to two databases, one for people to test with, and the other containing the entire LOC.
Slashdot Mirror
If a large group of scientists are measuring thrust from this thing, I think it is worthwhile to figure out why. Even if it is "nonsense", it may lead people to come up with a new/different space drive that is more efficient than what we have today. For instance, it may, in some novel way, just be pushing against Earth's magnetic field in a way that wasn't considered before. Or perhaps when it is tested dynamically instead of statically, the thrust falls off because the enclosure moved. It may lead us to consider certain laws of physics more from an engineering perspective than a pure physics one.
And you actually don't know what you are talking about. The Alpha's memory management is pretty similar to the x86 memory management. Yes, the x86 will not let you access memory that isn't allocated, and sometimes the OS handles this by BSOD, and in rare cases it *might* be able to recover. This would happen if that OS was running on x86, Alpha or PowerPC.
(Ironic quote you have there).
ClearPath Plus Server Libra Model 580 & 590 Level Epsilon Architecture Support Reference Manual
Available at http://www.app1.unisys.com/bookstore/
Yes, it lives!
My experience with TCP/IP under Windows indicates that it is very difficult to get more that 150 MBytes/Second from a single box (and I was using a specialized app to get that speed). I don't know what kind of delays the link imposes, but if you are using TCP/IP for the benchmark, be sure to enable scalable windows (search for RFC1323 on microsoft.com).
Another way to look at this is that PCI-Express links run at 2.5 Gbits/second. If you built something that ramped PCI-Express packets to/from your link, you could have an interesting I/O technology (comparable to Infiniband).
As others have pointed out, this sort of bandwidth is comparable/greater than most disk drive subsystems. Interesting demonstration is to use something like iSCSI to talk to disk drive that are on the other side of the link, and perhaps run some disk benchmarks, etc. Current (affordable) iSCSI controllers can push 60-80 MBytes/sec, so you could put 4 or 5 workstations on one side, and multiple iSCSI/iFiber routers on the other (with a bunch of iFiber disks).
Damn good movie, lots of blood and guts.
This is just goes to show how little people pay attention to books and libraries. Libraries have had an *open* standard, called Z39.50, to access databases containing information about books, for years. It is a standard based on MARC records, and just about every commercial library cataloging program is based on it, and can read/write these records. The Library of Congress provides full access via the Z39.50 interface to two databases, one for people to test with, and the other containing the entire LOC.