Bytemobile, Inc.

University of Illinois

Sungwon Ha

Upamanyu Madhow

Constantine Polychronopoulos

Sung-wook Han

Tae-eun Kim

Vaduvur Bharghavan

Sung-Wook Han

Kannan Ramchandran

David Craig

Hwangnam Kim

In the present scalable system routing method, received packets are associating with threads for processing the received packets. While a previously received packet is being processed, arrival of an interrupt is checked. If there is an interrupt, a thread is created associating the interrupt is created. Then, a determination of whether the thread associated with the interrupt has a priority that is higher than the priority of a thread associated with the previously received packet is made. If the thread associated with the interrupt has a higher priority than the previously received packet, the thread associated with the previously received packet is saved in a Shared Arena storage area. However, if the thread associated with the interrupt does not have a higher priority than the previously received packet, the thread associated with the interrupt is queued. Because threads are attached to the packets, the threads can now be suspended and resumed without having to disable interrupts, which includes periods during a context switch. As a result, a more flexible and efficient scheduling routing method can be implemented.

Thread based scalable routing for an active router

Improved quality of service management for multiple connections between a sender and a receiver may be achieved by allocating a host-level transmission rate among the multiple connections based on a ratio of a weight associated with each connection and a sum of the weights associated the connections. Data packets associated with the connections may then be selectively transmitted to the receiver such that data packets having a highest difference between the allocated transmission rate and an actual transmission rate are transmitted first. The data packets transmitted to the sender may also be clocked using a transmission timer having period corresponding to the host-level transmission rate.

Quality of service management for multiple connections within a network communication system

Improved data transport and management within a network communication system may be achieved by utilizing a transmit timer incorporated within the sender device and exploiting host-level statistics for a plurality of connections between a sender and receiver. The period of the transmit timer may be periodically adjusted based on a ratio of the smoothed round-trip time and the smoothed congestion window, thereby reducing or eliminating bursty data transmission commonly associated with conventional TCP architectures. For applications having a plurality of connections between a sender and a receiver that share a common channel, such as web applications, the congestion window and smoothed round trip time estimates for all active connections may be used to initialize new connections and allocate bandwidth among existing connections. This aspect of the present invention may reduce the destructive interference that may occur as different connections compete with one another to maximize the bandwidth of each connection without regard to other connections serving the same application. Error recovery may also be improved by incorporating a short timer and a long timer that are configured to reduce the size of the congestion window and the corresponding transmission rate in response to a second packet loss with a predefined time period in order to increase resilience to random packet loss.

Data transport acceleration and management within a network communication system

Growing community of inventors

San Jose, CA, United States of America

Sungwon Ha