Exploring the Use of a Reliable IP Multicast to Distribute BaBar's Online Event Processing and Filter Software to a Large Number of Farm Machines
- SLAC
Currently, the problem at hand is in distributing identical copies of OEP and filter software to a large number of farm nodes. One of the common methods used to transfer these softwares is through unicast. Unicast protocol faces the problem of repetitiously sending the same data over the network. Since the sending rate is limited, this process poses to be a bottleneck. Therefore, one possible solution to this problem lies in creating a reliable multicast protocol. A specific type of multicast protocol is the Bulk Multicast Protocol [4]. This system consists of one sender distributing data to many receivers. The sender delivers data at a given rate of data packets. In response to that, the receiver replies to the sender with a status packet which contains information about the packet loss in terms of Negative Acknowledgment. The probability of the status packet sent back to the sender is+, where N is the number of receivers. The protocol is designed to have approximately 1 status packet for each data packet sent. In this project, we were able to show that the time taken for the complete transfer of a file to multiple receivers was about 12 times faster with multicast than by the use of unicast. The implementation of this experimental protocol shows remarkable improvement in mass data transfer to a large number of farm machines.
- Research Organization:
- Stanford Linear Accelerator Center, Menlo Park, CA (US)
- Sponsoring Organization:
- USDOE Office of Science (US)
- DOE Contract Number:
- AC03-76SF00515
- OSTI ID:
- 826446
- Report Number(s):
- SLAC-PUB-9387
- Country of Publication:
- United States
- Language:
- English
Similar Records
PVM and IP multicast
Multicast communication in distributed algorithms