NDT:
Desktop Troubleshooting for All Users
The decentralized design of the Internet hampers the provision of services that must be coordinated across operating domains, such as end-to-end path analysis. The average user does not know what tools to run or what the results indicate. With the development of the NDT, end-users have a user-friendly tool that can quickly tell them if several common problems have been found in their path.
What is NDT? It is the Network Diagnostic Tool, developed and expanded under several government grants. For access to an NDT server, see the Testing to Abilene link off the NDT home page (http://e2epi.internet2.edu/ndt/) or http://netspeed.stanford.edu/.
NDT is a client/server program that provides network configuration and performance testing to a desktop or laptop computer. The system is composed of a client program (command line or Java applet) and a pair of server programs (a web server and a testing/analysis engine). Both command line and web-based clients communicate with a Web100-enhanced server to perform these diagnostic functions. Multi-level results allow novice and expert users to view and understand the test results.
Several studies have shown that the majority of network performance problems occur in or near the users’ desktop or laptop computer (the “last mile”). These problems include, but are not limited to: duplex mismatch conditions on Ethernet or FastEthernet links, incorrectly set TCP buffers in the user’s computer, or problems with the local network infrastructure.
NDT is designed to quickly and easily identify a specific set of conditions that are known to affect network performance. A multi-level series of plain language messages (suitable for novice users) and detailed test results (suitable for network engineers) are generated and available to the user. The test results may easily be emailed to the appropriate administrator to assist in the problem resolution phase.
The NDT consists of several components. Both client and server processes are used to perform a specific set of tests. The server processes include a basic web server (fakewww) to handle incoming web-based client requests and a test engine (web100srv) to perform the specific tests needed to determine what problems, if any, exist. The web100srv process also analyzes the test results and returns these results to the client.
Both command line (web100clt) and web-based (Java applet) clients are included in the NDT package. The command line client (web100clt) can be compiled and manually downloaded onto numerous client computers. The web-based client uses a Java applet to automate the testing process. This applet is downloaded when the user accesses the web server, removing the restriction to manually pre-load the software onto the client computer. This approach allows system administrators to pre-load the command line client onto their main server class computers while allowing the vast majority of users to automatically load the web-based client on an as needed basis. Features include:
- A simple web server process, which eliminates the need for a full web server;
- Throughput measurements that are reported in both directions;
- Web100 data extraction, used to analyze the test results;
- A web-based Java applet, which allows testing from any client;
- A command line tool, which allows testing from remote locations; and
- Multi-level results, tailored to novice and expert users.
Running NDT
The Java Applet (Tcpbw100.jar) is downloaded automatically when you access a specific NDT server. Once the download is complete, the applet is ready for testing. The user must press the ‘Start’ button to begin a
test. Once a test is complete, the ‘Statistics’, ‘More Details’ and ‘Report Problem’ buttons are activated.
At the conclusion of the test, the main applet page shows the basic throughput results, both to and from the client, and a simple statement of what link technology may limit the throughput. Link types range from Dial-up to 10 Gbps. See the web100srv.1 man page for more details. If warranted, ‘Warning’ or ‘Information’ messages may be printed.
The ‘Statistics’ page provides intermediate level details, including:
- The results of five specific configuration tests performed by NDT;
- The client’s OS and Java information;
- Several performance details including round trip time (RTT), packet size (MSS), and loss statistics (retransmissions, duplicate ACKs, SACKs, and out-of-order delivery);
- The percentage of time the server was receive-buffer limited, send-buffer limited, or congestion-window limited;
- The various TCP options that were negotiated between the client and server (SACK, Nagle, ECN, time stamps, window scaling); and
- The results of the middlebox tests (MSS modification and NAT detection).
- The raw web100 variables gathered on the server,
- The analysis results calculated by the server,
- The bandwidth*delay product for the send/receive/congestion window buffers,
- The variables and conditions used by specific detection algorithms, and
- The theoretical throughput available to this client based on the Mathis et. al. formula.
% web100clt server.domain.name
To see all available options:
% web100clt –h
To keep informed on the latest tools and techniques for problem-solving end-to-end difficulties, see the End-to-End Performance Initiative (E2Epi) website (http://e2epi.internet2.edu) or join the E2E-Perf Interest Group.