Richard Black and Simon Crosby. Experience and results from the implementation of an ATM socket family. In Proceedings USENIX Winter Technical Conference, San Francisco, January 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....QoS parameters that are not only passed into the kernel but also need to be negotiated with the peer endpoint and then passed back to the application. If QoS parameters are passed as part of the address structure or as a socket option, this would make QoS negotiation cumbersome (see Reference [4]) Further, we wanted to allow an application on any host with IP connectivity to a router to be able to request a channel from the signaling entity. Since we had already decided to place signaling in user space, given these two requirements, we decided that applications should communicate with ....

....based applications. Finally, unlike our design, they do not allow applications on machines without ATM host interfaces to access their API and send encapsulated cells into their ATM LAN. Black and Crosby at the University of Cambridge describe support for the MSNA ATM protocol stack in Reference [4]. A key design decision that they made was to use a special socket for communication between the signaling entity and the kernel. This leads to numerous problems with concurrent accesses to per VCI data structures. Further, the kernel needs to be aware of signaling messages interspersed with ....

R. Black and S. Crosby, Experience and Results from the Implementation of an ATM Socket Family, Proc. USENIX '94, 1994.

....work, including high resolution clocks and hardware time stamping of cells. Each port is equipped with 2K cell buffers, 4 MB of RAM and an ARM RISC processor, which implements all policy specific traffic control functions, using a flexible control plane which exports RPC style interfaces [5, 33] for signalling and management. This enabled the implementation of a wide range of measurement facilities to accurately monitor queue behaviour at the cell level resolution. Many manufacturers of commercial ATM equipment base their switch designs on an effectively non blocking (output buffered) ....

Richard Black and Simon Crosby. Experience and results from the implementation of an ATM socket family. In Proceedings USENIX Winter Technical Conference, San Francisco, January 1994.

....5 can be used for inter operation between all types of equipment. A detailed discussion of multiplexing techniques and the advantages and disadvantages of ATM networks may be found in [McAuley90] The author s experience of ATM includes the implementation of an ATM protocol stack within Unix [Black94c]. 2.1.2 Fairisle The Fairisle project (on which the author was employed for a period of a year before beginning research) has built a switch based ATM network for the local area [Leslie91, Black94d] Since this network was to be used to investigate the management of quality of service in a local ....

....protocol processing being performed at the lower interrupt priority splnet. This leads to performance loss due to the context switches (dispatching overhead and register file saves and loads) and reconfiguration of device interrupt masks. One discussion of this performance cost can be found in [Black94c]. A related cost is due to the effect of live lock, where excessive data arriving can hinder the progress of requests already in the system. Live lock in Unix has been reported in [Burrows88] and [Mogul] This particular live lock problem is now so severe with high performance network devices that ....

R. Black and S. Crosby. Experience and Results from the Implementation of an ATM Socket Family. In USENIX Winter 1994 Conference, pages 143--152, January 1994. (pp 6, 19)

....of cells. Fairisle can be configured as either an input or output buffered switch. Each port is equipped with 2K cell buffers, 4 MB of RAM and an ARM RISC processor, which implements all policy specific traffic control functions, using a flexible control plane which exports RPC style interfaces [2, 19] for signalling and management. This has enabled us to implement a wide range of measurement facilities to accurately monitor network performance, and to experiment with non FCFS queueing disciplines which attempt to provide guarantees to different types of traffic. 3 Experimental Network ....

R. Black and S. Crosby. Experience and results from the implementation of an ATM socket family. In Proceedings USENIX Winter Technical Conference, San Francisco, January 1994.

....IP over ATM and raw ATM channels. A key feature of our implementation is that only the data paths are implemented within the kernel; all the signalling, socket control, and management is implemented in a generic and portable user space daemon. Details of this work have been previously presented in [2]. 5 Performance The performance measurements presented in this section were aimed at verifying that our switch design (which included, at some considerable complexity, the ability to easily reprogram and experiment with its major components) provides a realistic platform on which to perform ....

R. Black and S. Crosby. Experience and results from implementation of an atm socket family. In Usenix Winter 1994 Conference. USENIX, January 1994.

....of cells. Fairisle can be configured as either an input or output buffered switch. Each port is equipped with 2K cell buffers, 4 MB of RAM and an ARM RISC processor, which implements all policy specific traffic control functions, using a flexible control plane which exports RPC style interfaces [8, 9] for signalling and management. This has enabled us to implement a wide range of measurement facilities to accurately monitor network performance. In the experiments reported in this paper, Fairisle was configured as an output buffered switch with FIFO queueing. 2 Measuring the rate function ....

R. Black and S. Crosby. Experience and results from the implementation of an ATM socket family. In Proceedings USENIX Winter Technical Conference, San Francisco, January 1994.

....of cells. Fairisle can be configured as either an input or output buffered switch. Each port is equipped with 2K cell buffers, 4 MB of RAM and an ARM RISC processor, which implements all policy specific traffic control functions, using a flexible control plane which exports RPC style interfaces [7, 8] for signalling and management. This has enabled us to implement a wide range of measurement facilities to accurately monitor network performance. In the experiments reported in this paper, Fairisle was configured as an output buffered switch with FIFO queueing. 2 Measuring the Rate Function In ....

R. Black and S. Crosby. Experience and results from the implementation of an ATM socket family. In Proceedings USENIX Winter Technical Conference, San Francisco, January 1994.

....over ATM and raw ATM channels. A key feature of this implementation is that only the data paths are implemented within the kernel; all the signalling, socket control, and management is implemented in a generic and portable user space daemon. Details of this work have been previously presented in [Black94]. 5 Performance The performance measurements presented in this section were aimed at verifying that our switch design (which included, at some considerable complexity, the ability to easily reprogram and experiment with its major components) provides a realistic platform on which to perform ....

R. Black and S. Crosby. Experience and Results from Implementation of an ATM Socket Family. In Usenix Winter 1994 Conference. USENIX, January 1994. (p 9--10)

....appendices A and B give some sample code for accessing AAL5 PDUs from a user program on Windows NT and Linux respectively. 6 1 Chapter 2 An EATM driver for Windows NT 2. 1 Introduction The author has previously implemented an ATM protocol family in other operating systems including Unix [Black94a]. The Windows NT operating system was expected to form a difficult platform for the implementation of a new network protocol due to Microsoft s history of failing to adopt concepts proven by the research community, and this indeed turned out to be the case. The documentation available for NT is ....

....one would find on Unix. There are also a lot of similarities with serial line device drivers for NT since they are also accessed asynchronously and with unformatted data. 2. 3 Overall structure The overall structure of the ATM protocol family was originally planned to be in the general form of [Black94a] as shown in figure 2.1, but experience with NT forced this design to be altered. Instead the structure is now intended to be as shown in figure 2.2. This is because the kernel device driver is not in a position to relay the signalling requirements to a signalling system since the NT system makes ....

[Article contains additional citation context not shown here]

R. Black and S. Crosby. Experience and Results from the Implementation of an ATM Socket Family. In USENIX Winter 1994 Conference, pages 143--152, January 1994. (pp 6--2, 6--3, 6--7)

....of cells. Fairisle can be configured as either an input or output buffered switch. Each port is equipped with 2K cell buffers, 4 MB of RAM and an ARM RISC processor, which implements all policy specific traffic control functions, using a flexible control plane which exports RPC style interfaces [2, 3] for signalling and management. This has enabled us to implement a wide range of measurement facilities to accurately monitor network performance. The switch at Telia is an 16 port output buffered switch operating at a line rate of 155 Mb s. The test bed incorporates high resolution measurement ....

R. Black and S. Crosby. Experience and results from the implementation of an ATM socket family. In Proceedings USENIX Winter Technical Conference, San Francisco, January 1994.

No context found.

Richard Black and Simon Crosby. Experience and results from the implementation of an ATM socket family. In Proceedings USENIX Winter Technical Conference, San Francisco, January 1994.

No context found.

R. Black and S. Crosby, "Experience and Results from the Implementation of an ATM Socket Family," Proceedings of the USENIX Winter Technical Conference, San Francisco, January 1994.

No context found.

Black, R., and S. Crosby, "Experience and Results from the Implementation of an ATM Socket Family", ATM Document Collection 3 (The Blue Book, Cambridge Computing Labs, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC