Explanation of the multi-session TCP transfer mechanism in the Pixel Plow agent

One of the more difficult challenges facing render farm services is how to move large quantities of data to and from their customers. Pixel Plow has implemented a fully automatic multi-session TCP transfer mechanism to address this problem. The result of this mechanism is that the customer Internet connection is fully saturated when moving large files, no matter where the customer is in the world.

A bit more specifically, the agent starts connecting to our farm using a few simultaneous TCP sessions, 4 at the time of this writing. While data is being transferred, it adds a single new session every minute until it detects that doing so does not result in an increase in throughput. This allows the agent to take advantage of higher session counts if the customer network, firewall, and Internet connection can sustain them. It does this without the overhead of establishing more TCP sessions than would help in overall data transfer performance.

There are rare occasions when customers utilize firewall software/hardware in their premise that interfere with the agent’s ability to use multiple TCP streams. Usually, such firewalls treat our agent as though it were a Bittorrent or other P2P transfer program, due to the multiple session count. If that functionality can be disabled in your firewall, or at least bypassed for our agent, we recommend you do so. If it can not, please contact support for other options. The connection count can be fixed to a specified number, thereby disabling the dynamic nature of the session addition algorithm.