Tierney, B., W. Johnston, G. Hoo, J. Lee, Performance Analysis in High-Speed Wide Area ATM Networks: Top-to-bottom end-to-end Monitoring, IEEE Network, Vol. 10, no. 3, May/June 1996, LBL-38246.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Berkeley National Laboratory WEJohnston lbl.gov 1 June 8, 1997 10:31 pm [NGI.HENP.paper.fm] High Speed Distributed Data Handling for High Energy and Nuclear Physics 1 William E. Johnston 2 , William Greiman, Brian Tierney, Arie Shoshani, Craig Tull Information and Computing Sciences Division Douglas Olson, Nuclear Science Division Ernest Orlando Lawrence Berkeley National Laboratory University of California 1.0 Introduction The advent (and promise) of shared, widely available, high speed networks provides the ....

Tierney "Performance Analysis in High-Speed Wide Area ATM Networks: Top-to-bottom end-to-end Monitoring", B. Tierney, W. Johnston, J. Lee, G. Hoo. IEEE Networking, May 1996.

....its usefulness in a variety of contexts, but most frequently in loosely coupled client server architectures. We began developing NetLogger in 1994, and in previous work we have shown that detailed application monitoring is vital for both performance analysis and application debugging [3] 17] 18][19]. This paper summarizes all NetLogger components, provides details on some recent enhancements, and provides some case studies. NetLogger has also proven to be a very useful tool for debugging multi threaded programs, allowing the developer to easily visualize interactions between threaded ....

....event log. In historical mode, the user can change the position in the log file, change the scale of the graph, zoom in and out interactively, choose a subset of events to look at, and so on. 4. 0 Recent NetLogger Enhancements We first developed NetLogger in 1994, and first published it in 1996 [19]. Over several years experience, we discovered a few missing features that we needed to add to make NetLogger more useful in a distributed environment, and to be suited for instrumenting middleware as well as client and server code. In this section we summarize some recent NetLogger enhancements, ....

Tierney, B., W. Johnston, G. Hoo, J. Lee, Performance Analysis in High-Speed Wide Area ATM Networks: Top-to-bottom end-to-end Monitoring, IEEE Network, Vol. 10, no. 3, May/June 1996, LBL-38246.

....common logging format. At the same time, various operating system and network parameters may be logged in the same format. This is accomplished by the Netlogger monitoring system [10] which has been used to analyze several network generated problems that showed up in the distributed applications [11]. 4.0 Other DPSS Applications We have conducted a set of high speed, network based, data intensive computing experiments between Lawrence Berkeley National Laboratory (LBNL) in Berkeley, Calif. and the Stanford Linear Accelerator (SLAC) in Palo Alto, Calif. The results of this experiment were ....

Tierney, B., W. Johnston, J. Lee, and G. Hoo, "Performance Analysis in High-Speed Wide Area ATM Networks: Top-to-bottom end-to-end Monitoring, " IEEE Networking, May 1996. (Available at http://www-itg.lbl.gov/DPSS/papers.)

....system in which we are developing, implementing, and testing these ideas. In most configurations, the DPSS is used as a network based, high performance, random access logical block server designed to supply and consume high speed data streams for other network based processing systems. See [21]. The DPSS is a logical block server whose functional components are distributed across a wide area network. See Figure 9. The DPSS is fundamentally a random access logical block server: There is no inherent organization to the blocks, and in particular, they would never be organized ....

Tierney, B., W. Johnston, G. Hoo, J. Lee, "Performance Analysis in High-Speed Wide-Area ATM Networks: Top-to-Bottom End-to-End Monitoring", IEEE Network, May, 1996.

....infrastructure use condition certificates object management (persistence, metadata mgmt, storage mgmt) 5) 5) Figure 8. The distributed Large Data Object overall architecture and data flow. 8 William Johnston, et al. speed data streams for other network based processing systems. See [21]. The DPSS is a logical block server whose functional components are distributed across a wide area network. See Figure 9. The DPSS is fundamentally a random access logical block server: There is no inherent organization to the blocks, and in particular, they would never be organized ....

Tierney, B., W. Johnston, G. Hoo, J. Lee, Performance Analysis in High-Speed Wide-Area ATM Networks: Top-to-Bottom End-to-End Monitoring, IEEE Network, May,

....of the DPSS is illustrated in Figure 2. Other papers describing the DPSS in more detail include [20] which describes how the DPSS was used to provide high speed access to remote data for a terrain visualization application, 21] which describes the basic architecture and implementation, and [22], which describes how the instrumentation abilities in the DPSS were used to help track down a wide area network problem. The application interface to the DPSS cache supports a variety of I O semantics, including Unixlike I O semantics, through an easy to use client API library (e.g. dpssOpen( ....

Brian Tierney, W. Johnston, G. Hoo, J. Lee, "Performance Analysis in High-Speed Wide-Area ATM Networks: Top-to-Bottom End-to-End Monitoring," IEEE Network, May, 1996, Vol. 10, no. 3. LBL Report 38246, 1996. http://www-itg.lbl.gov/DPSS/papers.html

....bandwidth that the total throughput increased significantly. While this solved an immediate problem, it was not a general solution, so we went back and conducted a series of experiments attempting to pinpoint the specific cause of the problem. These experiments and their results are described in [17]; however, here we illustrate some of the analysis. Referring to Figure 10, first, let us analyze what the performance monitoring shows directly. If we look at the long delayed block life lines (emphasized in the figure) we see the characteristic behavior of a data block getting into the write ....

....that the network operators upgraded the switch at A. The new switch as 600 KBy of buffering, which allows TCP s congestion algorithms to work correctly, and throughput is now up to an average of 30 Mbit s per data path, as should be the case. For a more detailed analysis of this experiment, see [17]. 6.0 Conclusions In order to achieve high end to end performance in widely distributed applications, a great deal of analysis and tuning is needed. In the MAGIC testbed we are evolving a methodology that includes network wide precision time sources and extensive instrumentation for time, ....

Tierney, B., W. Johnston, G. Hoo, J. Lee, "Performance Analysis in High-Speed Wide-Area ATM Networks: Top-to-Bottom End-to-End Monitoring", IEEE Network, May, 1996, Vol. 10, no. 3. LBL Report 38246, 1996. (Also see http://www-itg.lbl.gov/DPSS/papers.html.)

....providing parallel access to 20 30 disks. Other papers describing the DPSS in more detail include [23] which describes how the DPSS was used to provide high speed access to remote data for a terrain visualization application, 24] which describes the basic architecture and implementation, and [25], which describes how the instrumentation abilities in the DPSS were used to help track down a wide area network problem. This paper focuses on how we were able to greatly improve total throughput to applications by making the DPSS network aware. The application interface to the DPSS cache ....

Brian Tierney, W.Johnston, G. Hoo, J. Lee, "Performance Analysis in High-Speed Wide-Area ATM Networks: Top-to-Bottom End-to-End Monitoring," IEEE Network, May, 1996, Vol. 10, no. 3. LBL Report 38246, 1996. http://www-itg.lbl.gov/DPSS/papers.html

....and IP routers. The first generation of ATM switches had a very small amount of buffering. For example, the FORE ASX 100 had only 9K of buffers per output port. Multiple TCP streams, or even a single bursty host, could easily overflow the buffer, causing lost ATM cells and hence lost IP packets [TJH96]. Several enhancements addressed this problem, including the addition of the ability to do cell pacing in the ATM device drivers, larger switch buffers, and adding early packet discard to the switches. With early packet discard, the ATM switch is smart enough to not just drop one cell, but to ....

Tierney, B., Johnston, W., Hoo, G., and Lee J., "Performance Analysis in High-Speed Wide Area ATM Networks: Top-to-bottom end-to-end Monitoring", IEEE Network, Vol. 10, no. 3, May/June 1996, LBL-38246.

....event logs that can be used to provide detailed end to end application and system level monitoring, and tools for visualizing log data to view the state of the distributed system in real time. NetLogger has proven to be invaluable for diagnosing problems in networks and in distributed systems code [30]. This approach is novel in that it combines network, host, and application level monitoring, providing a complete view of the entire system. NetLogger monitoring allows us to identify hardware and software problems, and to react dynamically to changes in the system. LBNL has also been developing ....

B. Tierney, W. Johnston, J. Lee, G. Hoo, "Performance Analysis in High-Speed Wide Area ATM Networks: Top-to-bottom end-to-end Monitoring," IEEE Networking, May 1996. http://www-itg.lbl.gov/DPSS/ papers.html

....of the DPSS is illustrated in Figure 2. Other papers describing the DPSS in more detail include [16] which describes how the DPSS was used to provide high speed access to remote data for a terrain visualization application, 17] which describes the basic architecture and implementation, and [18], which describes how the instrumentation abilities in the DPSS were used to help track down a wide area network problem. The application interface to the DPSS cache supports a variety of I O semantics, including Unix like I O semantics, through an easy to use client API library (e.g. dpssOpen( ....

Brian Tierney, W. Johnston, G. Hoo, J. Lee, "Performance Analysis in High-Speed Wide-Area ATM Networks: Top-to-Bottom End-to-End Monitoring," IEEE Network, May, 1996, Vol. 10, no. 3. LBL Report 38246, 1996. http://www-itg.lbl.gov/DPSS/papers.html

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