home *** CD-ROM | disk | FTP | other *** search
- Analyzing LAN/WAN Internets:
- Testing IPX Routes
- Using Novell's LANalyzer 3.0
-
- Tim Hayes
- Systems Engineering Manager
- Novell Network Analysis Products Group
- LANalyzer Products Division
-
-
- Abstract:
- This AppNote describes techniques used to make analysis and problem
- resolution simple for large LAN and WAN systems. It presents a method for
- analyzing internets that was applied to a real problem at Novell.
-
- Disclaimer
-
- Novell, Inc. makes no representations or warranties with respect to the
- contents or use of these Application Notes (AppNotes) or of any of the
- third#party products discussed in the AppNotes. Novell reserves the right to
- revise these AppNotes and to make changes in their content at any time,
- without obligation to notify any person or entity of such revisions or
- changes. These AppNotes do not constitute an endorsement of the third#party
- product or products that were tested. Configuration(s) tested or described
- may or may not be the only available solution. Any test is not a
- determination of product quality or correctness, nor does it ensure
- compliance with any federal, state or local requirements. Novell does not
- warranty products except as stated in applicable Novell product warranties or
- license agreements.
-
- Copyright { 1991 by Novell, Inc., Provo, Utah. All rights reserved.
-
- As a means of promoting NetWare AppNotes, Novell grants you without charge
- the right to reproduce, distribute and use copies of the AppNotes, provided
- you do not receive any payment, commercial benefit or other consideration for
- the reproduction or distribution, or change any copyright notices appearing
- on or in the document.
-
- Contents
-
- Introduction 29
-
- The San Jose - Provo Internet 29
-
- LAN/WAN Analysis Method 30
-
- Determining the Route 30
-
- Determining the Delay for the First Hop 32
-
- Determining the Delay for the Final Hop 33
-
- Introduction
-
- Resolving performance problems on large internets can be challenging because
- of the number of components and LAN or WAN segments that need to be analyzed.
- There are techniques, however, that can make the analysis and problem
- resolution easier. This AppNote presents a method for using the LANalyzer to
- diagnose internet bottlenecks. It explains how this method was applied to a
- real problem at Novell.
-
- The San Jose - Provo InternetA number of PC users at Novell in San Jose,
- California, use LAN Access Servers in Provo, Utah, to access data on a server
- in Provo. To access the remote database, users' connections must traverse
- three IPX routed segments. One of those segments is a 1/3 T1 WAN link
- operating at approximately 52K bits per second. shows a map of the internet
- between San Jose and Provo.
-
- : Map of the Internet Between San Jose and Provo
-
-
- The users in San Jose complain that this remote data access is intolerably
- slow. They often wait several seconds for data to appear on their screens.
-
- LAN/WAN Analysis Method
-
- To isolate the performance bottleneck, you can measure the delays between the
- user's segment and each of the intermediate segments and the end segment.
- Novell's LANalyzer version 3.0 has an application called BRIGVIEW which was
- developed specifically for this purpose. BRIGVIEW allows the user to send
- IPX diagnostic packets through bridges and routers. To help in measuring the
- transmit delay, the LANalyzer has high resolution time stamping on both the
- transmitted packet and the response packet.
-
- Although the method here uses IPX diagnostic packets, the technique itself is
- general. The LANalyzer can perform this same experiment for any protocol.
- There are LANalyzer applications similar to BRIGVIEW for TCP/IP, AppleTalk,
- DECnet, and XNS.
-
- Basically, we're going to put a LANalyzer on network C9 99 00 02 and measure
- round trip delay times to components on segment 00 00 00 BD and also to
- components on segment 01 08 00 05. BRIGVIEW can transmit an IPX diagnostic
- packet from C9 99 00 02 to any specified segment. We will repeat this
- experiment ten times to calculate an average round trip delay time. After
- doing this for each segment, we will compare the Access Server 110 round
- trip delay to the delay for other devices on that segment.
-
- Determining the Route
-
- The first step is to determine the route used to go from San Jose to Provo.
- We used the LANalyzer BROADCST application to capture IPX RIP (Routing
- Information Protocol) packets and look for the route to 01 08 00 05. Figure 2
- shows the RIP packet which advertises the three#hop route from San Jose to
- Provo. The routing is done by a NetWare server called OSD#ROUTER.
-
- : IPX RIP Packet
-
- Next, we modified the BRIGVIEW application to transmit packets to OSD#ROUTER
- (using the Ethernet station address contained in the RIP packet). We also
- entered the source address of C9 99 00 02 and the destination address of 00
- 00 00 BD and 01 08 00 05. Since the LANalyzer has six transmit channels, we
- could test all connections in one pass. Figure 3 shows an IPX packet that
- traveled from network C9 99 00 02 to 01 08 00 05.
-
- : IPX Diagnostic Packet
-
- Determining the Delay for the First Hop
-
- To determine the average delay from the first to second hop of the three#hop
- path, we transmit ten packets from C9 99 00 02 to 00 00 00 BD. Then we take
- the average of the round trip delay times through OSD#ROUTER to the attached
- segment. Figure 4 shows the LANalyzer trace summary screen.
-
- : Round Trip Delay to Segment 00 00 00 BD
-
- Note that the round trip delays average about 18 milliseconds. This indicates
- the delay going through the first part of the three#hop route is about 18 ms,
- which is a reasonable delay. Thus, this is not a bottleneck.
-
- Determining the Delay for the Final Hop
-
- Next we transmit a single packet to 01 08 00 05 (using the default broadcast
- node ID convention) to get responses from all devices on the Provo segment.
- This allows us to identify other nodes that we can use as a control group in
- our round trip delay measurements. If all the devices respond in about the
- same amount of time, then the problem is not with any particular device. If,
- however, some device accessed by our complaining users responds more slowly
- than the other devices, we will have identified the problem. Figure 5 shows a
- response from a server on that segment.
-
- : Response from Segment 01 00 08 05
-
- For our control experiment, we transmit ten packets directly to the server
- identified in Figure 5. That will tell us about how long the delay is going
- from San Jose to that segment in Provo without going through the access
- server.
-
- Figure 6 shows the LANalyzer trace summary from our control experiment. From
- the interpacket arrival times, we can see that it takes 30 to 80 milliseconds
- to go from San Jose to Provo and back. Subtracting the 18 ms delay from our
- first hop tests yields a 12 to 62 ms delay across the WAN link. This is also
- a reasonable delay, ruling out the entire route from San Jose to Provo as the
- bottleneck.
-
- : San Jose to Provo Control Experiment Results
-
-
- If the delay time through Access Server 110 is significantly slower than the
- 30 to 80 ms average from our control experiment, then we have found our
- bottleneck.
-
- To check this, we transmit ten packets to Access Server 110 (whose Ethernet
- address we know from USERLIST /A). Figure 7 shows the LANalyzer trace summary
- of the packets going from San Jose to Access Server 110 in Provo and back.
-
- : Access Server 110 Round Trip Delay We can see that the round trip delay is
- 300 to 800 milliseconds. That is ten times longer than the delay in our
- control experiment. We can conclude (happily) that our LAN and WAN routing
- links are not the bottleneck. Rather, Access Server 110 is either slow or it
- is overutilized.
-
- We can repeat this experiment during off hours and see if Access Server 110
- responds better when no one is using it. If so, it's an indication that the
- Access Server is being overutilized. If it doesn't respond faster during off
- hours, then we have a basic performance problem with this machine.
-
- Editor's Note: The author accepts written feedback at FAX (801) 429#5511.
-
-