TY - JOUR
T1 - Fast RPC on the SHRIMP virtual memory mapped network interface
AU - Bilas, Angelos
AU - Felten, Edward W.
N1 - Funding Information:
We are grateful to Cezary Dubnicki for his expert advice on the SHRIMP simulator [11], and for his implementation of the SHRIMP system software. We also thank Liviu Iftode for his implementation of an early mini version of the system software. We thank the members of the SHRIMP team for many useful discussions during the course of this work. This project is sponsored in part by ARPA under Contract N00014-91-1-1144, by NSF under Grant MIP-9420653, by Digital Equipment Corporation, and by Intel Corporation. Felten is supported in part by an NSF National Young Investigator Award.
PY - 1997/1/10
Y1 - 1997/1/10
N2 - The emergence of new network interface technology is enabling new approaches to the development of communications software. This paper evaluates the SHRIMP virtual memory mapped network interface by using it to build two fast implementations of remote procedure call (RPC). Our first implementations, called vRPC, is fully compatible with the SunRPC standard. We change the RPC runtime library, the operating system kernel is unchanged, and only a minimal change was needed in the stub generator to create a new protocol identifier. Despite these restrictions, our vRPC implementation is several times faster than existing SunRPC implementations. A round-trip null RPC with no arguments and results under vRPC takes about 33 μs. Our second implementation, called ShrimpRPC, is not compatible with SunRPC but offers much better performance. ShrimpRPC specializes the stub generator and runtime library to take full advantage of SHRIMP's features. The result is a round-trip null RPC latency of 9.5 μs, which is about 1 μs above the hardware minimum.
AB - The emergence of new network interface technology is enabling new approaches to the development of communications software. This paper evaluates the SHRIMP virtual memory mapped network interface by using it to build two fast implementations of remote procedure call (RPC). Our first implementations, called vRPC, is fully compatible with the SunRPC standard. We change the RPC runtime library, the operating system kernel is unchanged, and only a minimal change was needed in the stub generator to create a new protocol identifier. Despite these restrictions, our vRPC implementation is several times faster than existing SunRPC implementations. A round-trip null RPC with no arguments and results under vRPC takes about 33 μs. Our second implementation, called ShrimpRPC, is not compatible with SunRPC but offers much better performance. ShrimpRPC specializes the stub generator and runtime library to take full advantage of SHRIMP's features. The result is a round-trip null RPC latency of 9.5 μs, which is about 1 μs above the hardware minimum.
UR - http://www.scopus.com/inward/record.url?scp=0001756980&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0001756980&partnerID=8YFLogxK
U2 - 10.1006/jpdc.1996.1272
DO - 10.1006/jpdc.1996.1272
M3 - Article
AN - SCOPUS:0001756980
SN - 0743-7315
VL - 40
SP - 138
EP - 146
JO - Journal of Parallel and Distributed Computing
JF - Journal of Parallel and Distributed Computing
IS - 1
ER -