anderson@charming.Berkeley.EDU (David Anderson) (05/03/89)
[ Also available for ftp in the archive on Midgard.UCSC.EDU --DL ] %A David P. Anderson %A Domenico Ferrari %T The DASH Project %J ACM SIGOPS Workshop on Distributed Systems %C Amsterdam %D SEP 1986 %X Very early position paper. Superceded. %A David P. Anderson %A Domenico Ferrari %A P. Venkat Rangan %A Shin-Yuan Tzou %T The DASH Project: Issues in the Design of Very Large Distributed Systems %R Technical Report 87/338 %I UC Berkeley CS Division %D JAN 1987 %X Early position paper. Main points: 1) technological trends point to large scale (global) and high network performance; 2) distributed system research should anticipate and study the new opportunities (multimedia communication, massive distributed parallelism) and problems (security, high network delay, etc.) arising from this change; 3) optimal solutions often involve integration between system levels (network design, process scheduling, VM), which may justify discarding existing standards and starting over. %A David P. Anderson %A Domenico Ferrari %A P. Venkat Rangan %T Subtransport Level: The Right Place for End-to-End Security Mechanisms %R Technical Report 87/346 %I UC Berkeley CS Division %D MAR 1987 %X Cleaned-up version of TR 87/328. %A David P. Anderson %A P. Venkat Rangan %T A Basis for Secure Communication in Large Distributed Systems %J IEEE Symposium on Security and Privacy %D APR 1987 %O Also Technical Report 87/328, UC Berkeley CS Division %X Describes the DASH scheme for network security. Security principals (hosts, owners, etc.) are authenticated by PKE, and public keys are distributed by a hierarchical name server. Host-to-host channels (which use SKE) are established as needed. %X Owners are authenticated (using PKE signatures) across these channels, and kernels maintain "authentication caches" to avoid repeating this procedure. It is argued that this "subtransport-level" security mechanism is more efficient than higher-level mechanisms. %X The basic idea of the design is used in the current DASH system, but many of the details have changed. %A David P. Anderson %A Domenico Ferrari %A P. Venkat Rangan %A Bruno Sartirana %T The Empirical Evaluation of a Security-Oriented Datagram Protocol %J IFIP Performance '87 %C Brussels %D DEC 1987 %O Also Technical Report 87/350, UC Berkeley CS Division %X Similar to TR 87/328, with additional performance measurements. %A David P. Anderson %A P. Venkat Rangan %T High-Performance Interface Architectures for Cryptographic Hardware %J Eurocrypt '87 %C Amsterdam %D APR 1987 %X Assuming that a fast encryption chip is available, where is the best place to put it in a system architecture? The goals are to reduce CPU overhead and I/O bus bandwidth. %A David P. Anderson %A Domenico Ferrari %A P. Venkat Rangan %A Bruno Sartirana %T A Protocol for Secure Communication and its Performance %J Proc. 7th ICDCS %D SEP 87 %X Similar to TR 87/328, with additional performance measurements. %A Shin-Yuan Tzou %A David P. Anderson %A G. Scott Graham %T Efficient Local Data Movement in Shared-Memory Multiprocessor Systems %J Technical Report 87/385 %I UC Berkeley CS Division %D DEC 1987 %X Early description of the DASH facility for fast data movement using VM remapping. Performance is increased using various techniques: 1) allowing temporary inconsistency between PTE and data structures; 2) lazy mapping; 3) no change of virtual address when pages are moved. Mostly superceded by TR 88/452 %A David P. Anderson %A Domenico Ferrari %T The DASH Project: An Overview %R Technical Report 88/405 %I UC Berkeley CS Division %D FEB 1988 %X Snapshot of the project as of 1/88. The basic design of the system (real-time kernel, VM system, channel-based IPC architecture) is described. Somewhat sketchy, since implementation was still at an early stage. %A David P. Anderson %T A Software Architecture for Network Communication %J Proc. 8th ICDCS %D June 1988 %O Also Technical Report 87/386, UC Berkeley CS Division %X Also UCBTR 87/386 Explores the idea of basing a communication architecture on "channels" with meaningful performance, reliability, security parameters. Calls them "real-time message streams (RMS)"; later changed to "channels". Superceded by TR 89/498. %A David P. Anderson %A Shin-Yuan Tzou %T The DASH Local Kernel Structure %R Technical Report 88/463 %I UC Berkeley CS Division %D NOV 1988 %X Design document for the DASH kernel. Describes process scheduling (multiprocessor deadline scheduling), kernel message-passing, timers, synchronization, user program interface (message-passing, object references, system calls). Network communication and the VM system are described in two other TR's. All three design documents just present and describe the C++ interfaces; no discussion or comparison with other systems. %A David P. Anderson %A Robert Wahbe %T The DASH Network Communication Architecture %R Technical Report 88/462 %I UC Berkeley CS Division %D NOV 1988 %X Design document for the DASH network communication architecture. This is based on the abstraction of simplex "channels" with performance, reliability and security parameters. The abstraction is used at several levels; at higher levels the delay parameters include CPU processing time. A "subtransport layer" manages local resources, does multiplexing and caching, and does security and reliability functions. Slightly out-of-date; channel parameters have been changed, and the management protocol used by the ST layer now recursively calls a higher-level RPC facility. %A David P. Anderson %A Shin-Yuan Tzou %A G. Scott Graham %T The DASH Virtual Memory System %R Technical Report 88/461 %I UC Berkeley CS Division %D NOV 1988 %X Design document for the DASH VM system. Basic ideas: use different mechanisms (and different fixed regions of each VAS) for the different VM functions: private read/write data, shared read-only data, and IPC buffers. %A Shin-Yuan Tzou %A David P. Anderson %T A Performance Evaluation of the DASH Message-Passing System %R Technical Report 88/452 %I UC Berkeley CS Division %D NOV 1988 %X Detailed performance study of the integrated message-passing/VM system. Microsecond-level timing breakdown of an MP operation into about 40 pieces. Moving a page between two VAS's takes between 87 and 254 microseconds, depending on whether it is accessed. %A D. Ferrari %T Guaranteeing Performance for Real-Time Communications in Wide-Area Networks %R Technical Report 89/485 %I UC Berkeley CS Division %D JAN 1989 %X An algorithm for implementing channels (w/ various types of performance guarantees) in an Internetwork. %A David P. Anderson %A Robert Wahbe %T A Framework for Multimedia Communication in a General-Purpose Distributed System %J Technical Report 89/498 %I UC Berkeley CS Division %D MAR 1989 %X Motivates the design given in TR 88/462, gives some comparisons, and discusses implications for protocol and local system design. Description of channel parameters supercedes TR 88/462.