NetProfiler: Profiling Wide-Area Networks Using

Slides:



Advertisements
Similar presentations
Omniran TG 1 Cooperation for OmniRAN P802.1CF Max Riegel, NSN (Chair OmniRAN TG)
Advertisements

Technical Architectures
Web Caching Schemes1 A Survey of Web Caching Schemes for the Internet Jia Wang.
1 In VINI Veritas: Realistic and Controlled Network Experimentation Jennifer Rexford with Andy Bavier, Nick Feamster, Mark Huang, and Larry Peterson
15-441: Computer Networking Lecture 26: Networking Future.
Aleksandar Kuzmanovic & Edward W. Knightly A Performance vs. Trust Perspective in the Design of End-Point Congestion Control Protocols.
Kyushu University Graduate School of Information Science and Electrical Engineering Department of Advanced Information Technology Supervisor: Professor.
SESSION 9 THE INTERNET AND THE NEW INFORMATION NEW INFORMATIONTECHNOLOGYINFRASTRUCTURE.
Faten Yahya Ismael.  It is technology creates a network that is physically public, but virtually it’s private.  A virtual private network (VPN) is a.
FIREWALL TECHNOLOGIES Tahani al jehani. Firewall benefits  A firewall functions as a choke point – all traffic in and out must pass through this single.
1 Content Distribution Networks. 2 Replication Issues Request distribution: how to transparently distribute requests for content among replication servers.
P2P File Sharing Systems
Introduction Widespread unstructured P2P network
1 The SpaceWire Internet Tunnel and the Advantages It Provides For Spacecraft Integration Stuart Mills, Steve Parkes Space Technology Centre University.
Chapter 4. After completion of this chapter, you should be able to: Explain “what is the Internet? And how we connect to the Internet using an ISP. Explain.
Remote Access Chapter 4. Learning Objectives Understand implications of IEEE 802.1x and how it is used Understand VPN technology and its uses for securing.
ACM 511 Chapter 2. Communication Communicating the Messages The best approach is to divide the data into smaller, more manageable pieces to send over.
Service Architecture of Grid Faults Diagnosis Expert System Based on Web Service Wang Mingzan, Zhang ziye Northeastern University, Shenyang, China.
Professor OKAMURA Laboratory. Othman Othman M.M. 1.
Lecture#1 on Internet. Internet Addressing IP address: pattern of 32 or 128 bits often represented in dotted decimal notation IP address: pattern of 32.
2: Application Layer1 Chapter 2 outline r 2.1 Principles of app layer protocols r 2.2 Web and HTTP r 2.3 FTP r 2.4 Electronic Mail r 2.5 DNS r 2.6 Socket.
An Efficient Approach for Content Delivery in Overlay Networks Mohammad Malli Chadi Barakat, Walid Dabbous Planete Project To appear in proceedings of.
A Routing Underlay for Overlay Networks Akihiro Nakao Larry Peterson Andy Bavier SIGCOMM’03 Reviewer: Jing lu.
Tony McGregor RIPE NCC Visiting Researcher The University of Waikato DAR Active measurement in the large.
Addressing Issues David Conrad Internet Software Consortium.
Othman Othman M.M., Koji Okamura Kyushu University 1.
Kiew-Hong Chua a.k.a Francis Computer Network Presentation 12/5/00.
Content-oriented Networking Platform: A Focus on DDoS Countermeasure ( In incremental deployment perspective) Authors: Junho Suh, Hoon-gyu Choi, Wonjun.
Networking Fundamentals. Basics Network – collection of nodes and links that cooperate for communication Nodes – computer systems –Internal (routers,
1 Capacity Dimensioning Based on Traffic Measurement in the Internet Kazumine Osaka University Shingo Ata (Osaka City Univ.)
Setup and Management for the CacheRaQ. Confidential, Page 2 Cache Installation Outline – Setup & Wizard – Cache Configurations –ICP.
ITGS Network Architecture. ITGS Network architecture –The way computers are logically organized on a network, and the role each takes. Client/server network.
정하경 MMLAB Fundamentals of Internet Measurement: a Tutorial Nevil Brownlee, Chris Lossley, “Fundamentals of Internet Measurement: a Tutorial,” CMG journal.
CS 6401 Overlay Networks Outline Overlay networks overview Routing overlays Resilient Overlay Networks Content Distribution Networks.
1 NetProfiler: Profiling Networks From the Edge Venkat Padmanabhan Microsoft Research June 2005 With Sharad Agarwal (MSR), Jitu Padhye (MSR), Dilip Joseph.
Introduction to Active Directory
KYUNG-HWA KIM HENNING SCHULZRINNE 12/09/2008 INTERNET REAL-TIME LAB, COLUMBIA UNIVERSITY DYSWIS.
Network Architecture IS250 Spring 2010 John Chuang
Wireless sensor and actor networks: research challenges Ian. F. Akyildiz, Ismail H. Kasimoglu
Fall, 2001CS 6401 Switching and Routing Outline Routing overview Store-and-Forward switches Virtual circuits vs. Datagram switching.
Network Virtualization Ben Pfaff Nicira Networks, Inc.
Chapter 1 Characterization of Distributed Systems
Date: April. 13, Monday Evening.
The Intranet.
CIIT-Human Computer Interaction-CSC456-Fall-2015-Mr
Architecture and Algorithms for an IEEE 802
CONNECTING TO THE INTERNET
4.1.5 multi-homing.
Reddy Mainampati Udit Parikh Alex Kardomateas
Internet and Intranet.
PROTEAN: A Scalable Architecture for Active Networks
Mohammad Malli Chadi Barakat, Walid Dabbous Alcatel meeting
CHAPTER 3 Architectures for Distributed Systems
Vocabulary Prototype: A preliminary sketch of an idea or model for something new. It’s the original drawing from which something real might be built or.
Introduction to client/server architecture
Client-Server Interaction
Internet and Intranet.
CS222 Web Programming Course Outline
Transport Layer Unit 5.
Overlay Networking Overview.
The Future of Transport
Practical Issues for Commercial Networks
Lecture 2: Overview of TCP/IP protocol
Lecture 3: Secure Network Architecture
Network+ Guide to Networks, Fourth Edition
SAMANVITHA RAMAYANAM 18TH FEBRUARY 2010 CPE 691
EE 122: Lecture 22 (Overlay Networks)
Database System Architectures
An Empirical Evaluation of Wide-Area Internet Bottlenecks
Dong Xuan*, Sriram Chellappan*, Xun Wang* and Shengquan Wang+
Presentation transcript:

NetProfiler: Profiling Wide-Area Networks Using Peer Cooperation Venkata N. Padmanabhan (Microsoft Research) Sriram Ramabadhrah (UCSD) Jitendra Padhye (Microsoft Research) IPTPS 2005 No Notes University of Cincinnati, 2005

Motivation Internet ISP2 ISP3 ISP1 Conventional tools do not ping traceroute Conventional tools do not allow us to view the network from multiple vantage points. ISP1 University of Cincinnati, 2005

Problems Operators/ISPs: cant really see what kind of network experience there users are having. Users: cant do much if there are network problems. Current tools like ping and traceroute capture information only from the perspective of one user These tools focus only on the IP-level path. Problems maybe with higher layer entities like proxies and servers. University of Cincinnati, 2005

More problems and a solution Tools like ping and traceroute rely on “active probing”. Active probing creates overheads especially if too many hosts start probing simultaneously. Active probing doesn’t always disambiguate the cause of the problem. Proposed solution: Hosts accumulate performance information passively, from multiple vantage points, across all layers (not only the IP level path), and the information is aggregated using a p2p network. University of Cincinnati, 2005

Tasks To acquire performance data from end-user perspective. To aggregate this data in meaningful ways and share it among cooperating peers. To use this data for fault analysis and diagnosis. University of Cincinnati, 2005

Outline NetProfiler Task 1: Acquire data Experimental Results Architecture Data Acquisition TcpScope WebScope Normalization Data Aggregation Attributes Aggregation Tree Filtering Existing Systems Analysis and Diagnosis Distributed Blame Allocation Comparative Analysis N/W Engg. and Analysis N/W Health Reporting Experimental Results Discussion Conclusion Task 1: Acquire data

NetProfiler NetProfiler is the p2p system that is suggested which addresses the issues and problems outlined. University of Cincinnati, 2005

NetProfiler - Architecture The architecture and design of NetProfiler can be described with respect to the following activities: Data Acquisition Normalization Data Aggregation Analysis and Diagnosis University of Cincinnati, 2005

NetProfiler – Architecture - Data Acquisition Performed by ‘sensors’. Sensors are software modules residing on end hosts. Primarily perform passive observation. Sensors could be developed for various protocols and applications. Two sensors were developed TcpScope (for tcp) WebScope (for http) University of Cincinnati, 2005

NetProfiler – Architecture - Data Acquisition - TcpScope Designed to work at the transport layer. Passive sensor which listens for TCP transfers. Parameters that are measured include (but are not limited to): Round Trip Time (RTT) Congestion Window of sender (cwnd) Bottleneck bandwidth University of Cincinnati, 2005

NetProfiler – Architecture - Data Acquisition - WebScope If a caching proxy is used, the user loses visibility at the transport layer beyond the proxy. To get around this problem, another sensor called WebScope was developed. This gives visibility into an end-to-end web transaction. Cache busting techniques are employed to make sure content is not served from the cache University of Cincinnati, 2005

NetProfiler – Architecture - Data Acquisition Data Acquisition - Summary Data is acquired at the hosts by the use of sensors (software modules) They do not actively probe but listen passively. Two have been implemented – TcpScope, WebScope. Measures include – RTT, cwnd, bottleneck bandwidth University of Cincinnati, 2005

Outline NetProfiler Experimental Results Architecture Data Acquisition TcpScope WebScope Normalization Data Aggregation Attributes Aggregation Tree Filtering Existing Systems Analysis and Diagnosis Distributed Blame Allocation Comparative Analysis N/W Engg. and Analysis N/W Health Reporting Experimental Results Discussion Conclusion Task 2: Aggregate and share data

NetProfiler – Architecture - Normalization Once the data is acquired through sensors, it needs to be aggregated and shared with peers. But, before that, it must be normalized. This is required because every host is operating with different parameters such as bandwidth and data transfer rate. University of Cincinnati, 2005

NetProfiler – Architecture – Data Aggregation Sensors are used to gather performance data at each host. For the system to be meaningful, an efficient aggregation scheme must be devised. The authors have proposed an aggregation based on a “set of client attributes”. University of Cincinnati, 2005

NetProfiler – Architecture – Data Aggregation - Attributes Hierarchical – impose a natural hierarchy on the hosts along which data can be aggregated. Example - Geographical Location Neighborhood -> City -> Region -> Country Non – hierarchical – used for filtering Example – Bandwidth class University of Cincinnati, 2005

NetProfiler – Architecture – Data Aggregation – Aggregt’n Tree Aggregation Tree A tree which mirrors the hierarchy imposed by the attribute. Data can be aggregated along this tree. University of Cincinnati, 2005

NetProfiler – Architecture – Data Aggregation – Filtering Hierarchical attributes are used to logically organize the data aggregation. But as data travels from node to node, it is separated into classes / types based on non-hierarchical attributes of the host. University of Cincinnati, 2005

NetProfiler – Architecture – Data Aggregation – Existing Systems Existing Systems for distributed monitoring, aggregating and querying on the Internet SDIMS Ganglia Slicestat IrisNet PIER Sophia Astrolabe The idea here is to leverage these systems to aggregate data. The goal is to develop a system for fault-diagnosis and analysis rather than to design a new data aggregation system. University of Cincinnati, 2005

Outline NetProfiler Experimental Results Discussion Conclusion Architecture Data Acquisition TcpScope WebScope Normalization Data Aggregation Attributes Aggregation Tree Filtering Existing Systems Analysis and Diagnosis Distributed Blame Allocation Comparative Analysis N/W Engg. and Analysis N/W Health Reporting Experimental Results Discussion Conclusion Task 3: Use data for analysis and diagnosis

NetProfiler – Architecture – Analysis and Diagnosis So far, the acquisition and aggregation of data has been discussed. How is the aggregated data used to analyze the system and diagnose commonly occurring problems ? University of Cincinnati, 2005

NetProfiler – Architecture – Analysis and Diagnosis – Dist NetProfiler – Architecture – Analysis and Diagnosis – Dist. Blame Allocation Distributed Blame Allocation A client experiencing bad performance will tend to blame the entities involved in the end-to-end transaction. The idea behind distributed blame allocation is to use data supplied by sensors and other nodes to locate trouble spots. University of Cincinnati, 2005

NetProfiler – Architecture – Analysis and Diagnosis – Comparative Analysis This is an attractive feature of the profiling system. Clients have information about the performance experienced by neighbors. This is used to make decisions such as switching to a higher type of service or switching ISPs. University of Cincinnati, 2005

NetProfiler – Architecture – Analysis and Diagnosis – Comparative Analysis Comparative Analysis - Examples Am I being treated fairly ? A client continues to experience worse performance than its neighbors even though they are served by the same ISP…time to change ISPs. Without the aid of NetProfiler, this kind of information cannot be made available. Should I get a better connection ? A client has information about the performance experienced by neighbors who are in a higher bandwidth class. This aids in the decision making process. University of Cincinnati, 2005

NetProfiler – Architecture – Analysis and Diagnosis – N/W Engg NetProfiler – Architecture – Analysis and Diagnosis – N/W Engg. & Analysis Network Engineering and Analysis Network engineering and analysis is facilitated by NetProfiler performance data. Example Network operators may want to choose from a set of alternatives for a company network. Like – leased lines or VPN Usually, performance data is based on link and router monitoring. NetProfiler provides and end-user perspective of network performance from multiple vantage points. University of Cincinnati, 2005

NetProfiler – Architecture – Analysis and Diagnosis – N/W Health Reporting Network Health Reporting The information gathered by NetProfiler can be used to generate reports about the general health of a wide-area-network. These reports are available through other means but the advantage of using NetProfiler are greater coverage, lower cost, flexibility and contribution based operation. University of Cincinnati, 2005

Outline NetProfiler Experimental Results Discussion Conclusion Architecture Data Acquisition TcpScope WebScope Normalization Data Aggregation Attributes Aggregation Tree Filtering Existing Systems Analysis and Diagnosis Distributed Blame Allocation Comparative Analysis N/W Engg. and Analysis N/W Health Reporting Experimental Results Discussion Conclusion

NetProfiler – Experimental Results - Setup Heterogeneous set of clients 147 PlanetLab nodes, dialup hosts, connected to 26 Pops on MSN network and 5 connected to the Microsoft Corporate Network. Repeatedly downloading from a diverse set of 70 websites University of Cincinnati, 2005

NetProfiler – Architecture – Experimental Results - Findings Access to a website failed at almost every client. Suggests a server-side problem. Differences in failure rates among seemingly “equivalent” clients. A group of nodes sharing an attribute had higher failure rates than all others. This suggests a common cause of the problem affecting these nodes. [details] University of Cincinnati, 2005

NetProfiler – Experimental Results - Findings Sometimes, problems occurred between a client-server pair. This suggests that there is some problem in the path between these two nodes. [details] University of Cincinnati, 2005

NetProfiler – Discussion Deployment Models Coordinated Organic University of Cincinnati, 2005

NetProfiler – Discussion Organic Deployment - Challenges Security Clients may not want to reveal IP or other info. They may be allowed to report at a “coarser” level, eg. At ISP Data integrity – malicious users sending incorrect performance data Bootstrapping – chicken-and-egg Proposed solution is to allow some active probing in the initial stages University of Cincinnati, 2005

Conclusions NetProfiler is a system that enables monitoring and diagnosis of network problems. It is meant for wide area networks and relies on heavy participation. It can potentially benefit both ISPs and end users. Ongoing work includes detection of surreptitious activity to predict attacks.

End http://geocities.com/alisaif_uc/ppts.htm Saif Ali ECECS University of Cincinnati April 2005

Details of experimental findings System wide failures Most nodes Could not reach the sites www.technion.ac.il and www.hku.hk Discrepancy among seemingly equivalent nodes Among MSN nodes connected to PoPs, those using ICG as the upstream provider experienced lower failure rates than those using other providers like QWEST or UUNET. Group of clients sharing an attribute All 5 hosts on the Microsoft Corporate Network experience high failure rates in accessing a site www.royal.gov.uk whereas failure for other hosts is negligible. Particular Client-Server Pair A Microsoft Corporate Network node in China can never access the site www.nmt.edu whereas others do not have this problem. Back