Search NASA⌕ Search

Engineering topics

Whetten, Brian

Publications and source records attributed to Whetten, Brian.

8 records
All sourcesNASA NTRSDOE OSTI
NASA NTRS1997

High-Performance, Reliable Multicasting: Foundations for Future Internet Groupware Applications

Network protocols that provide efficient, reliable, and totally-ordered message delivery to large numbers of users will be needed to support many future Internet applications. The Reliable Multicast Protocol (RMP) is implemented on top of IP multicast to facilitate reliable transfer of data for replicated databases and groupware applications that will emerge on the Internet over the next decade. This paper explores some of the basic questions and applications of reliable multicasting in the context of the development and analysis of RMP.

Callahan, John↗
NASA NTRS1996

Specification and Design of a Fault Recovery Model for the Reliable Multicast Protocol

The Reliable Multicast Protocol (RMP) provides a unique, group-based model for distributed programs that need to handle reconfiguration events at the application layer. This model, called membership views, provides an abstraction in which events such as site failures, network partitions, and normal join-leave events are viewed as group reformations. RMP provides access to this model through an application programming interface (API) that notifies an application when a group is reformed as the result of a some event. RMP provides applications with reliable delivery of messages using an underlying IP Multicast media to other group members in a distributed environment even in the case of reformations. A distributed application can use various Quality of Service (QoS) levels provided by RMP to tolerate group reformations. This paper explores the implementation details of the mechanisms in RMP that provide distributed applications with membership view information and fault recovery capabilities.

Montgomery, Todd↗
NASA NTRS1995

Fault recovery in the reliable multicast protocol

The Reliable Multicast Protocol (RMP) provides a unique, group-based model for distributed programs that need to handle reconfiguration events at the application layer. This model, called membership views, provides an abstraction in which events such as site failures, network partitions, and normal join-leave events are viewed as group reformations. RMP provides access to this model through an application programming interface (API) that notifies an application when a group is reformed as the result of a some event. RMP provides applications with reliable delivery of messages using an underlying IP Multicast (12, 5) media to other group members in a distributed environment even in the case of reformations. A distributed application can use various Quality of Service (QoS) levels provided by RMP to tolerate group reformations. This paper explores the implementation details of the mechanisms in RMP that provide distributed applications with membership view information and fault recovery capabilities.

Callahan, John R.↗
NASA NTRS1995

Reliable multicast protocol specifications protocol operations

This appendix contains the complete state tables for Reliable Multicast Protocol (RMP) Normal Operation, Multi-RPC Extensions, Membership Change Extensions, and Reformation Extensions. First the event types are presented. Afterwards, each RMP operation state, normal and extended, is presented individually and its events shown. Events in the RMP specification are one of several things: (1) arriving packets, (2) expired alarms, (3) user events, (4) exceptional conditions.

Callahan, John R.↗
NASA NTRS1995

Reliable multicast protocol specifications flow control and NACK policy

This appendix presents the flow and congestion control schemes recommended for RMP and a NACK policy based on the whiteboard tool. Because RMP uses a primarily NACK based error detection scheme, there is no direct feedback path through which receivers can signal losses through low buffer space or congestion. Reliable multicast protocols also suffer from the fact that throughput for a multicast group must be divided among the members of the group. This division is usually very dynamic in nature and therefore does not lend itself well to a priori determination. These facts have led the flow and congestion control schemes of RMP to be made completely orthogonal to the protocol specification. This allows several differing schemes to be used in different environments to produce the best results. As a default, a modified sliding window scheme based on previous algorithms are suggested and described below.

Callahan, John R.↗
NASA NTRS1995

A high performance totally ordered multicast protocol

This paper presents the Reliable Multicast Protocol (RMP). RMP provides a totally ordered, reliable, atomic multicast service on top of an unreliable multicast datagram service such as IP Multicasting. RMP is fully and symmetrically distributed so that no site bears un undue portion of the communication load. RMP provides a wide range of guarantees, from unreliable delivery to totally ordered delivery, to K-resilient, majority resilient, and totally resilient atomic delivery. These QoS guarantees are selectable on a per packet basis. RMP provides many communication options, including virtual synchrony, a publisher/subscriber model of message delivery, an implicit naming service, mutually exclusive handlers for messages, and mutually exclusive locks. It has commonly been held that a large performance penalty must be paid in order to implement total ordering -- RMP discounts this. On SparcStation 10's on a 1250 KB/sec Ethernet, RMP provides totally ordered packet delivery to one destination at 842 KB/sec throughput and with 3.1 ms packet latency. The performance stays roughly constant independent of the number of destinations. For two or more destinations on a LAN, RMP provides higher throughput than any protocol that does not use multicast or broadcast.

Montgomery, Todd↗