DeWitt, T., T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland: 1997, `ReMoS: A Resource Monitoring System for Network Aware Applications'. Technical Report CMU-CS-97-194, School of Computer Science, Carnegie Mellon University.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....at the user level, allowing history based data collected across the network to be used to generate the predictions needed by a particular user. Remos has been implemented and tested in a variety of networks and is in use in a number of di erent environments. 7] The API Remos started as an API [13]. Its goal is to de ne a portable, high level interface to network information. This comes down to two major abstractions: Flows: application level connections between pairs of computation nodes. Remos de nes three types of ows, according to bandwidth utilization. Fixed ows use a speci c, ....

T. DeWitt et. al. Remos: A resource monitoring system for network-aware applications. Technical report, Carnegie Mellon University, 1998.

....by connecting multiple consumer producer pipes. A number of groups have are now developing monitoring services based on the GMA architecture, such as R GMA [3] Relational GMA, so called because it uses a relational model for all data, organizes data about Grid entities in tables) REMOS [4], and TOPOMON [2] The OGSA notification service is very similar to GMA. The OGSA interface specifies a notification source and sink, which are very similar to a producer and consumer in the GMA. However the current OGSI specification does not provide an unsubscribe operation, or specify a ....

T. Dewitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkist, J. Subhlok and D. Sutherland, ReMoS: A resource monitoring system for network aware applications, Tech. Rep. CMU-CS-97-194, School of Computer Science, Carnegie Mellon University, December 1997

....the multiple resources used by applications. 1. Introduction Motivation. The need for flexible and extensible monitoring tools for the high performance computing environment, particularly for management and development of dynamic resource responsive applications, has been documented many times [15, 17, 3, 8, 1]. Shared memory architectures provide an easy way to achieve such monitoring, but distributed memory machines present a more difficult target. The management of a distributed system is based on accurate and up to date monitoring of the system s relevant observable quantities and events. Management ....

T. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. ReMoS: A Resource Monitoring System for Network-Aware Applications. Technical report, Carnegie Mellon School of Computer Science, CMU-CS-97-194, 1997.

....by di#erent CPU clock rates and the dynamic load behavior due to other processes on the nodes. Most issues related to the discovery of network topologies have been discussed and implemented in ReMoS [5] An algorithm for dynamically building clusters from network mapping information is given in [6]. To demonstrate the feasibility of large scale and accurate distributed folding simulation in this technical report, we quantify the increase of the quality of the folding simulations due to clustered workers and report our experience with clustered workers implemented on to the commercial ....

T. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. ReMoS: A Resource Monitoring System for Network-Aware Applications. Technical Report CMU-CS-97-194, Carnegie Mellon School of Computer Science, 1997.

....eciently and with high throughput, but such monitoring structures su er from high latency in data access. This is important when monitoring is used for online program tuning or steering[6, 22] Dproc could bene t from additional performance information captured at the network or switch levels[5, 23]. At this time, we are implementing network monitoring by inspection of kernel resident protocol stacks using the kernelresident libpcap portion of the well known tcpdump facility. Higher level services that interpret or analyze monitoring data[15, 21] are not the subject of this research, but ....

A. DeWitt, T. Gross, B. Lowekamp, N.Miller, P.Steenkiste, J. Subhlok, D. Sutherland, "ReMoS: A Resource Monitoring System for Network-Aware Applications". Carnegie Mellon School of Computer Science, CMU-CS-07-194.

....eciently and with high throughput, but such monitoring structures su er from high latency in data access. This is important when monitoring is used for online program tuning or steering[7, 25] Dproc could bene t from additional performance information captured at the network or switch levels[6, 28]. At this time, we are implementing network monitoring by inspection of kernel resident protocol stacks using the kernelresident libpcap portion of the well known tcpdump facility. Higher level services that interpret or analyze monitoring data[16, 24] are not the subject of this research, but ....

A. DeWitt, T. Gross, B. Lowekamp, N.Miller, P.Steenkiste, J. Subhlok, D. Sutherland, "ReMoS: A Resource Monitoring System for Network-Aware Applications". Carnegie Mellon School of Computer Science, CMU-CS-07-194.

....use empirical method to get their value; see Section 3.2 for details. 2.3 Network Module We use Remos in the Network Module to monitor and predict network performance. In the following, network performance means network bandwidth available for the application. Remos (Resource Monitoring System)[3, 4, 15] is a network performance middleware service developed at CMU. It provides a scalable, exible and portable network monitoring system for applications in distributed computing environments. Remos is composed of two parts: Modeler and Collector. The Modeler implements the Remos API, which enables ....

Tony DeWitt, Thomas Gross, Bruce Lowekamp, Nancy Miller, Peter Steenkiste, Jaspal Subhlok, Dean Sutherland. ReMoS: A Resource Monitoring System for Network-Aware Applications, Technical Report, CMU-CS-97-194.

....conditions so applications can make an educated choice when selecting a server. More recently, the idea of a measurement infrastructure that provides network distance information to applications became quite popular. Some of the current e orts in the area include ID Maps [3] PingER [1] Remos [4], and the Network Weather Service [11] As we point out in Section I, the most closely related e orts to this paper are the studies conducted by Crovella et al. 2] McManus [6] and the PingER work by Cottrell et al. Our work addresses the more general issue of what metrics should be used to ....

T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. Remos: A resource monitoring system for network-aware applications. Carnegie Mellon School of Computer Science, CMU-CS-97-194, 1997.

....to the available disks of a typical computing environment (e.g. 10s of gigabytes) and very large compared to any single disk (e.g. hundreds of 10 gigabytes) 2. 2 Network Aware Applications In order to efficiently use high speed wide area networks, applications will need to be network aware [6]. Network aware applications attempt to adjust their demands in response to changes in resource availability. For example, emerging QoS services will allow network aware applications to participate in resource management, so that network resources are applied in a way that is most effective for ....

....very nice approach, it will currently only help applications running on hosts with a customized version of the NetBSD operating system. Our approach can be used by any distributed application on any Unix system. Other network aware application or middleware projects include the Remulac project [6][22] which is also doing network monitoring and application adaptation based on the current network conditions. This project focuses on designing a middleware API for applications to use; we focus instead on issues for data intensive computing. The goals of our network monitoring system are ....

DeWitt, T. Gross, T. Lowekamp, B. Miller, N. Steenkiste, P. Subhlok, J. Sutherland, D., "ReMoS: A Resource Monitoring System for Network-Aware Applications" Carnegie Mellon School of Computer Science, CMU-CS-97-194. http://www.cs.cmu.edu/afs/cs/project/cmcl/www/remulac /remos.html

....and its analysis depends on the nature of the problem, and the necessary monitoring data may not be available when it is needed. 2 In addition to the ability to locate performance problems, in order to efficiently use NGI networks this new class of applications will need to be network aware [10]. Network aware applications attempt to adjust their resource demands in response to changes in resource availability. Emerging QoS services will allow the application to participate in resource management, so that network resources are applied in a way that is most effective for the application. ....

DeWitt, T. Gross, T. Lowekamp, B. Miller, N. Steenkiste, P. Subhlok, J. Sutherland, D., "ReMoS: A Resource Monitoring System for Network-Aware Applications" Carnegie Mellon School of Computer Science, CMU-CS-97-194. http://www.cs.cmu.edu/afs/cs/project/cmcl/www/remulac/remos.html

....configuration, and execution. A GUI allows library routines or user developed routines to be combined into an application task graph. The task graph is then interpreted and configured to execute on currently available resources. At Carnegie Mellon, the ReMoS (Resource Monitoring System) project [7] is developing a portable and systemindependent API that allows applications to obtain informa tion about network status and capabilities. Most architectures generate information about the network hardware and software in a system specific format. ReMoS provides a standard interface format that ....

....applications can query the directory service through an Application Programming Interface. For example, the Metacomputing Directory Service (MDS) in Globus [8] provides current information on start up costs and end to end bandwidths between every pair of processors. The ReMoS API, developed at CMU [7], is an example of an API that is independent of the details of network hardware. Table 1 and 2 are examples of information provided by the directory service in GUSTO, which is a testbed of Globus. The directory provides current values of end to end network latency and bandwidth between any pair ....

T. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. ReMoS: A resource monitoring system for network-aware applications. TechnicalReport CMU-CS-97-194, School of Computer Science, Carnegie Mellon University, Dec 1997.

....rst implementation uses ping s estimates of network latency as proximity metric. More recently, the idea of a measurement infrastructure that provides network distance information to applications became quite popular. Some of the current e orts in the area include ID Maps [3] PingER [1] Remos [4], and the Network Weather Service [11] The most closely related e orts to this paper are the studies conducted by Crovella et al. 2] McManus [7] and the PingER work by Cottrell et al. Our work addresses the more general issue of what metrics should be used to measure server proximity. The ....

T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. Remos: A resource monitoring system for network-aware applications. Carnegie Mellon School of Computer Science, CMU-CS-97-194, 1997.

.... over a wide range of system conditions [13] Andresen et al. study compositional models and performance trade offs for distributed data intensive applications in [4] Performance monitoring and forecasting of wide area networks is discussed in works such as the NWS [36, 37] GloPerf [19] ReMoS [12] and [7, 10] The Netlogger system [33] presents a profiling framework for distributed storage systems. Many open problems remain. Some preliminary investigation and experimental results indicate that the aggregation of 64 kilobyte network samples collected at short term intervals increases the ....

T. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. ReMoS: A resource monitoring system for network-aware applications. Technical Report CMU-CS-97-194-REVISED, School of Computer Science, Carnegie Mellon University, December 1998.

No context found.

DeWitt, T., T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland: 1997, `ReMoS: A Resource Monitoring System for Network Aware Applications'. Technical Report CMU-CS-97-194, School of Computer Science, Carnegie Mellon University.

No context found.

A. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland: "ReMoS: A Resource Monitoring System for Network-Aware Applications" Carnegie Mellon School of Computer Science, CMU-CS-97-194.

....is then possible to derive the value of that parameter that will optimize performance, as a function of the other parameters. Applications that use model based adaptation collect information about system characteristics when they start up, either by querying the system through a special interface [4], or by running a small set of benchmarks. The system parameters are then plugged into the system model, which yields the method of distribution (e.g. the number of nodes) that optimizes performance for this particular system configuration. This method has been used successfully for a small but ....

Tony DeWitt, Thomas Gross, Bruce Lowekamp, Nancy Miller, Peter Steenkiste, and Jaspal Subhlok. ReMoS: A Resource Monitoring System for Network Aware Applications. Technical Report CMU-CS-97-194, Carnegie Mellon University, December 1997.

....history, current network conditions, or an estimate of the future availability. The local area implementation of Remos is based on SNMP processes on network nodes and entails a very low overhead. More details of Remos design and implementation are available in [15] and a full API is described in [4]. 3 Node selection procedure We first describe the structure of the network topology graph, which is the basis for node selection. We present a set of algorithms to solve the simple cases of node selection, and then discuss the extensions and restrictions in the context of more general network ....

DEWITT, T., GROSS, T., LOWEKAMP, B., MILLER, N., STEENKISTE, P., SUBHLOK, J., AND SUTHERLAND, D. Remos: A resource monitoring system for networkaware applications. Tech. Rep. CMU-CS-97-194, Carnegie Mellon University, Dec 1997.

....can easily be done using the logical topology [1] but are difficult to resolve using, for example, pair wise information. Note that applications can obtain pair wise information by issuing separate queries for each pair of endpoints. More details on the Remos API can be found elsewhere [2], 1] 3] III. REMOS DESIGN We present the design of the CMU Remos prototype. A. High level Remos architecture Our Remos implementation has two layers: a collector and a modeler (Figure 2) they are responsible for network oriented and application oriented functionality, respectively. The ....

Tony DeWitt, Thomas Gross, Bruce Lowekamp, Nancy Miller, Peter Steenkiste, and Jaspal Subhlok, "ReMoS: A Resource Monitoring System for Network Aware Applications," Tech. Rep. CMU-CS-97-194, Carnegie Mellon University, December 1997.

....dimensions has been picked. Implicit feedback is also like to give more accurate and timely information (in a narrower part of the operating space) than explicit feedback. 3 Example: Remos The Remos API provides a query based interface that allows clients to obtain best effort information [3] on network conditions. The applications specifies the kind of information it needs, and Remos supplies the best available information. To limit the scope of the query, the application must select network parameters and parts of a larger network that are of interest. In this section we briefly ....

....and Remos supplies the best available information. To limit the scope of the query, the application must select network parameters and parts of a larger network that are of interest. In this section we briefly describe the main Remos features. A more detailed description can be found elsewhere [3, 6]. 3.1 Level of abstraction To accommodate the diverse application needs, the Remos API provides two levels of abstraction: high level flowbased queries and lower level topology based queries. Remos supports flow based queries. A flow is an application level connection between a pair of ....

T. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. Remos: A resource monitoring system for network-aware applications. Technical Report CMU-CS-97-194, Carnegie Mellon School of Computer Science, Dec 1997.

No context found.

T. De Witt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland, "ReMoS a resource monitoring system for network aware applications," School Comput. Sci., Carnegie Mellon Univ., Pittsburgh, PA, Tech. Rep. CMU-CS-97-194, 1997.

No context found.

Tony DeWitt, Thomas Gross, Bruce Lowekamp, Nancy Miller, Dean Sutherland, and Jaspal Subhlok. Remos: A resource monitoring system for network-aware applications. Technical Report CMU-CS-97-194, Carnegie Mellon School of Computer Science, 1997.

No context found.

A. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, J. Subhlok, and D. Sutherland. Remos: A resource monitoring system for network-aware applications. Technical Report CMU-CS-97-194, Carnegie Mellon School of Computer Science, 1997.

No context found.

T. DeWitt, T. Gross, B. Lowekamp, N. Miller, P. Steenkiste, and J. Subhlok, ReMoS: A Resource Monitoring System for Network Aware Applications, Technical Report, Carnegie Mellon University, CMU-CS97 -194 (1997).

No context found.

A. DeWitt, et al., "ReMoS: A Resource Monitoring System for Network-Aware Applications," CMU-CS-97-194, Dec. 1997.

No context found.

Tony DeWitt, Thomas Gross, Bruce Lowekamp, Nancy Miller, Peter Steenkiste, Jaspal Subhlok, and Dean Sutherland. ReMoS: A resource monitoring system for network-aware applications. Technical Report CMU-CS-97-194, Carnegie Mellon, 1997.

First 50 documents

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