End-to-end arguments in system design
ACM Transactions on Computer Systems (TOCS)
An end-to-end approach to globally scalable network storage
Proceedings of the 2002 conference on Applications, technologies, architectures, and protocols for computer communications
Squid: The Definitive Guide
A study of bandwidth-sharing mechanisms in connection-oriented networks
A study of bandwidth-sharing mechanisms in connection-oriented networks
On using circuit-switched networks for file transfers
On using circuit-switched networks for file transfers
Flow labelled IP: a connectionless approach to ATM
INFOCOM'96 Proceedings of the Fifteenth annual joint conference of the IEEE computer and communications societies conference on The conference on computer communications - Volume 3
Dynamic information-based scalable hashing on a cluster of web cache servers
Concurrency and Computation: Practice & Experience
Hi-index | 0.00 |
A new type of gateway, called a ConnectionLess Circuit (CLC) gateway is proposed to internetwork connectionless (CL) networks with dynamically shared circuit/virtual networks. In contrast with IP routers, CLC gateways implement (a) a signaling function to receive circuit setup/release requests triggered by end hosts and initiate dynamic circuit setup/release through their attached circuit subnetworks, (b) transport-layer functionality to terminate TCP connections across CL subnetworks and Circuit-TCP (CTCP) connections across circuits, and (c) disk butTering to decouple transfers across CL subnetworks from transfers across high-speed circuit subnetworks. Specifically, this internetworking architecture is proposed for large files transfers with a goal of increasing end-to-end transfer rates. A path for introducing this architecture and services into the current Internet, involving programmable patch panels in addition to CLC gateways and circuit subnetworks, is presented. A CLC gateway using web proxy Squid software is implemented. Experiments with this gateway demonstrate the need for decoupling through disk butTering and TCP/CTCP terminations. Measurements on a rate-mismatched path, with a 155Mbps circuit between CLC gateways, and a IMbps bottleneck link rate on a CL subnetwork, show that a 100MB file transfer without disk butTering resulted in a 0.6% circuit utilization, tying up the circuit for 843.8sec, while with disk butTering, utilization improves to 90.8% while holding the circuit for only 5.96sec.