This paper analyzes the performance of a reliable multicast transpoft protocol and discusses experimental test results. The Reliable Multicast Transport Protocol has been proposed to support "reliable" information delivery from a server to thousands of receivers over unreliable networks via IP-multicast. The protocol provides high-performance for most receivers through the advantage of IP multicast while also supporting temporarily unavailable or performance impaired receivers. lts applicability to large scale delivery is examined using an experimental network and the backoff time algorithm which avoids ACK implosion. The two types of flow control with the protocol are also exarnined. Separate retransmission is used to offset the local performance decline limited to a small nurober of receivers. Monitor-based rate control is used to offset the global performance declines due to causes such as network congestion.

No abstract

Multicore processors are widely used for various types of computers. In order to achieve high-performance on such multicore systems, it is necessary to extract coarse grain task parallelism from a target program in addition to loop parallelism. Regarding the development of parallel programs, Java or a Java-extension language represents an attractive choice recently, thanks to its performance improvement as well as its platform independence. Therefore, this paper proposes a parallel Java code generation scheme that realizes coarse grain task parallel processing with layerunified execution control. In this parallel processing, coarse grain tasks of all layers are collectively managed through a dynamic scheduler. In addition, we have developed a prototype parallelizing compiler for Java programs with directives. In performance evaluations, the compiler-generated parallel Java code was confirmed to attain high performance. Concretely, we obtained 7.82 times faster speed-up for the Jacobi program, 7.38 times faster speed-up for the Turb3d program, 6.54 times faster speed-up for the Crypt program, and 6.15 times faster speed-up for the MolDyn program on eight cores of Xeon E5-2660.

No abstract

This paper proposes Monitor-based Flow Control (MBFC) to realize flow/congestion control needed for one-to-many bulk reliable multicast (RM) protocols. Bulk RM on top of IP multicast requires flow/congestion control because 1) it needs to adjust to the effective bandwidth to min:imize packet losses and retransmission, and 2) to share the link bandwidth with other legacy traffic such as TCP so that RM does not override them aggressively. MBFC is a generic mechanism to implement such flow/congestion control. Thus we think it is very important to provide an effective flow and congestion control mechanism that enables RM traffic to coexist with legacy TCP traffic in Internet. MBFC is based on rate flow control and involves a monitor mechanism to adjust its sending rate. In order to realize coexistence with TCP traffic it mimics TCP's flow/congestion control algorithm, that is, additive increase/multiplicative decrease. To investigate its effectiveness MBFC was implemented in our RM protocol, and a series of simulations were performed in which simultaneous bulk RM and TCP flow were poured. The simulation results showed that our rate control policy effectively achieved bandwidth sharing between bulk RM and bulk TCP transfer traffic. We also conclude that 1) Setting the transmission rate to the worst receiver leads to an extremely unfair bandwidth sharing occupied by TCP flow, 2) If RM only experiences a few loss it should not give up its bandwidth, otherwise TCP monopolizes it. 3) It is more effective for MBFC to introduce RED gateways to prevent RM from an aggressive TCP traffic.