Granted, the single-assignment policy won't help mainstream OSes and applications, but I am pretty sure high-performance routers probably do it all the time (i.e. pass pointers instead of copying, and garbage-collect buffers after the data has been sent and ack'ed).
Good point. I'd rather keep the mechanical mounting tight in order to improve heat dissipation, though. 7200RPM is 120Hz, so a 10s beat period would require a 0.1% design tolerance on that (i.e. pretty loose, therefore unlikely even across brands, as you suggest). But the beating could also occur between harmonics (e.g. 120Hz times the number of poles in the motors) and vibration modes probably depend on mechanical construction (material densities, shapes, etc). Hopefully different brands would have enough differences to prevent any noticeable resonance.
My system used to generate a very low frequency humming that would easily propagate through walls and floors, peaking every 10s or so. I found out that although each of my disks was amazingly silent, when mounted together they would vibrate at a beat frequency due to minor differences in rotation speed.
1. Rack-mounted all the 24x7 stuff in a closet. 2. Reengineered the airflow (you'd think front-to-back is a pretty easy rule to follow, but most manufacturers ruin it with flashy vents on all sides). 3. Did the math: 40CFM will gain 8degC when extracting 200W, and will flow at 2m/s through a 10cmx10cm hole. 4. Replaced the small fans with two slow, large blowers for the whole enclosure. 5. Enjoy diskless, fanless workstations.
I went through the paper and it suggests that whenever one of the two Skype clients is behind a firewall, traffic will flow through a supernode.
So, no fancy firewall-piercing here. And that's not even optimal: there are other cases where traffic could be direct (and only signalling would need to go through a SN).
however, his firewall will now be expecting return traffic originating from the port that his computer attempted to connect to. therefore, the second connection attempt, from the other computer, will succeed
Interesting, but I wouldn't expect this technique to defeat anything else than trivial "personal firewalls". Any NAT between the two computers would alter the calling ports and break your assumptions.
Slashdot Mirror
Also it looks like the BSD zero-copy sockets already use the MMU copy-on-write trick I mentioned in order to preserve the userspace semantics:h tml
http://www.cs.duke.edu/ari/trapeze/freenix/node6.
> If the application changes the data directly after the send()-call this should not affect what is sent.
So just don't let the application change the data (hint: single-assignment programming languages).
> This means that the OS has to copy the data into kernel memory,
Either that, or you could improve support for copy-on-write in the MMU (which might benefit other tasks than just networking).
Sometimes changing the assumptions is the proper way to solve the problem.
Good point. I'd rather keep the mechanical mounting tight in order to improve heat dissipation, though.
7200RPM is 120Hz, so a 10s beat period would require a 0.1% design tolerance on that (i.e. pretty loose, therefore unlikely even across brands, as you suggest). But the beating could also occur between harmonics (e.g. 120Hz times the number of poles in the motors) and vibration modes probably depend on mechanical construction (material densities, shapes, etc). Hopefully different brands would have enough differences to prevent any noticeable resonance.
My system used to generate a very low frequency humming that would easily propagate through walls and floors, peaking every 10s or so.
I found out that although each of my disks was amazingly silent, when mounted together they would vibrate at a beat frequency due to minor differences in rotation speed.
See this thread.
I would recommend mixing disks from different manufacturers or different models, even for a RAID setup, in order to avoid this.
1. Rack-mounted all the 24x7 stuff in a closet.
2. Reengineered the airflow (you'd think front-to-back is a pretty easy rule to follow, but most manufacturers ruin it with flashy vents on all sides).
3. Did the math: 40CFM will gain 8degC when extracting 200W, and will flow at 2m/s through a 10cmx10cm hole.
4. Replaced the small fans with two slow, large blowers for the whole enclosure.
5. Enjoy diskless, fanless workstations.
I went through the paper and it suggests that whenever one of the two Skype clients is behind a firewall, traffic will flow through a supernode.
So, no fancy firewall-piercing here. And that's not even optimal: there are other cases where traffic could be direct (and only signalling would need to go through a SN).
Interesting, but I wouldn't expect this technique to defeat anything else than trivial "personal firewalls". Any NAT between the two computers would alter the calling ports and break your assumptions.
Has anyone actually tried this ?