Traffic Scheduler for Multithread
Posted: Thu Jan 26, 2017 3:04 pm
Building from:
Multithread (http://www.vpnusers.com/viewtopic.php?f=7&t=5163)
and
Maximum Throughput Site to Site VPN Solution (http://www.vpnusers.com/viewtopic.php?f=7&t=7270)
Is there an internal scheduler to SoftEther that could balance traffic across multiple internal instances to make use of all available cores (multithread)?
Peplink has this feature calling is "SpeedFusion" where sessions are spread across multiple tunnels. For Peplink the tunnels are per hardware link technology, but how about SoftEther doing a similar thing via spawning multiple "tunnels" between points A and B and then scheduling sessions across the tunnels to maximize throughput?
The point here is to limit it to sessions so as to avoid packet re-ordering.
More and more hardware have increasing cpu core count but SoftEther appears to be limited by cpu core speed. The proposed method here of SoftEther spawning one tunnel connection per core (as a user defined setting) and scheduling across them all (even a crude round robin approach) should, in theory, increase the overall SoftEther throughput.
Is this reasonably possible with the current SoftEther software architecture, please?
Thank you!
Multithread (http://www.vpnusers.com/viewtopic.php?f=7&t=5163)
and
Maximum Throughput Site to Site VPN Solution (http://www.vpnusers.com/viewtopic.php?f=7&t=7270)
Is there an internal scheduler to SoftEther that could balance traffic across multiple internal instances to make use of all available cores (multithread)?
Peplink has this feature calling is "SpeedFusion" where sessions are spread across multiple tunnels. For Peplink the tunnels are per hardware link technology, but how about SoftEther doing a similar thing via spawning multiple "tunnels" between points A and B and then scheduling sessions across the tunnels to maximize throughput?
The point here is to limit it to sessions so as to avoid packet re-ordering.
More and more hardware have increasing cpu core count but SoftEther appears to be limited by cpu core speed. The proposed method here of SoftEther spawning one tunnel connection per core (as a user defined setting) and scheduling across them all (even a crude round robin approach) should, in theory, increase the overall SoftEther throughput.
Is this reasonably possible with the current SoftEther software architecture, please?
Thank you!