Here is a solution that would be possible. run a Hardware RAID box (I dont have any names, but they use a SCSI to SCSI converter). and run the SCSI chain (U2Wide would be plenty) to 6 seperate servers. mount the partision(s) that hold the videos read only, and they can just blindly send the data out. on the sixth server, mount it read write and when you need to add a new movie, do it through the server that isnt serving anything (the one that has the partision mounted read write).
This sounds like the most likley senario. some quick calculations. at full buss speed (80 Meg/sec) = 640 Megabits/sec. that would serve 533 1.2 megabit/sec video feeds. in reality, lets say that you have the five servers all on there own 100 megabit/sec subnet, that would serve about 450 cliants. also, this would provide fault tolerance. you would have RAID for the storage and 5 servers. Linux would do this easily.
Re:Dual Celery Faster... I think not
on
Carmack on the K7
·
· Score: 1
I dont know about you, but I can instantly tell if I am sitting at a celeron system or a full blown PII/PIII system. it isnt nearly as responsive. for SMP work, the size of the cache is a lot more important. I am talking about things like Photoshop, Apache, etc. as far as pure number crunching like what you would do with RC-5 or even SETI@Home, there is little diffrence. also as many peaple have stated, going SMP relies a lot on the application. Most high end CAD or Graphics programs will work faster, but dont expect to go any faster with word perfect or something. with these two pitfalls of a dual celeron vs a single K7, I would go with the K7 any day (It would be much more cool anyways). If you notice, noone is marketing dual celeron systems for high end workstations or servers, even though there are dual celeron main boards out there. Do you think that there is a reason for this ???
Slashdot Mirror
Here is a solution that would be possible. run a Hardware RAID box (I dont have any names, but they use a SCSI to SCSI converter). and run the SCSI chain (U2Wide would be plenty) to 6 seperate servers. mount the partision(s) that hold the videos read only, and they can just blindly send the data out. on the sixth server, mount it read write and when you need to add a new movie, do it through the server that isnt serving anything (the one that has the partision mounted read write).
This sounds like the most likley senario. some quick calculations. at full buss speed (80 Meg/sec) = 640 Megabits/sec. that would serve 533 1.2 megabit/sec video feeds. in reality, lets say that you have the five servers all on there own 100 megabit/sec subnet, that would serve about 450 cliants. also, this would provide fault tolerance. you would have RAID for the storage and 5 servers. Linux would do this easily.
I dont know about you, but I can instantly tell if I am sitting at a celeron system or a full blown PII/PIII system. it isnt nearly as responsive. for SMP work, the size of the cache is a lot more important. I am talking about things like Photoshop, Apache, etc. as far as pure number crunching like what you would do with RC-5 or even SETI@Home, there is little diffrence. also as many peaple have stated, going SMP relies a lot on the application. Most high end CAD or Graphics programs will work faster, but dont expect to go any faster with word perfect or something. with these two pitfalls of a dual celeron vs a single K7, I would go with the K7 any day (It would be much more cool anyways). If you notice, noone is marketing dual celeron systems for high end workstations or servers, even though there are dual celeron main boards out there. Do you think that there is a reason for this ???