 |
|
 |
|
Engine Options
The "Engine" settings control what and how PingPlotter sends data. The "Packet Type" settings allows you to pick what kind of data you want to send to tailor PingPlotter to your network needs. PingPlotter supports 4 packet types:· ICMP using Windows DLL (default). This method is the traditional method and matches the data that the Windows TRACERT command uses. It works on all Windows operating systems, and is a good balance of reliability and capability.· ICMP using Raw Sockets. This method sends the same data that the Windows DLL does, but does so by manually creating the packets in PingPlotter. This sometimes works in cases where ICMP.DLL doesn’t, but requires administrative rights on the PC.· UDP Packets (Unix-Style). Instead of sending ICMP packets, this sends UDP packets, which might be able to get through some firewalls that are blocking ICMP. This uses ports 33434 – 33500 and closely mirrors Unix’s traceroute command. This requires administrative user rights.· TCP Packets. If a firewall is blocking ICMP packets, it’s sometimes possible to get a response using TCP packets instead. TCP is the protocol used for all web browsers in addition to FTP, Telnet and others. This requires Administrative user rights, and on some operating systems (Windows XP SP2) also requires a helper library. When PingPlotter sends out a packet, it waits a certain amount of time for a response. The longer it waits, the more resources it needs to use (to keep sockets open), but the more likely that it will get a response. 9999 is a pretty high number – very few packets take longer than this to return. Some people use 1500 here. Time interval between hop traces PingPlotter sends out packets without waiting for previous packets to return. This doesn’t mean, however, that it sends out all packets at once (you don’t want to do this, or it might overwhelm your bandwidth). This is in ms (or 1/1000s of a second). The Packet Size can make a considerable difference in latency performance. Normally, you want to use a relatively small number here. The default is 56 bytes, but in some cases you might need to lower this (especially on TCP port 80 packets, which sometimes get dropped unless they are 40 bytes). 1500 is lot of data, and should be used with great care. A 1500 byte packet means PingPlotter will be sending out 30-50 K per second worth of data, which can cause its own problems (and makes measuring latency more challenging). Some networks use this byte to prioritize data. If that’s the case, you can manipulate this value to see how the network responds. This is relatively advanced course material – ie: if you don’t know what this means, just leave it at 0. Sometimes hop 1 or hop 2 might never respond. Rather than continuing to pound away at these hops and never getting a response, it sometimes makes sense to just ignore the first hop or 2. This is totally normal on lots of cable modems, and can happen on any connection – where the first hop is always "silent". Ignoring the first non-responding hops will save some resources. TCP Specific Settings When using TCP packets, you can specify which target port to use. Usually, you’ll want to use port 80 here, but you’re welcome to use any reasonable port. Windows XP SP2’s firewall blocks creation of TCP packets, so you’ll need to use WinPCap to create packets under that OS (and possibly others). See our online page with more details at http://www.pingplotter.com/winpcap.html.
This makes PingPlotter do all DNS lookups in the main application thread. If you’re having DNS lookup problems, it might be worth trying. Setting this too high can cause PingPlotter to crash, especially on old operating systems. 45 is proven safe. If you have a really fast trace interval, you may need to increase this number to support the trace interval. Usually, you want to lower the "Timeout Speed" setting before raising this, as that might be adequate. Every packet has a minimum size (28 bytes for ICMP/UDP, 40 bytes for TCP). Anything in addition to that core size is padded with a value. Modify these settings to pad with the value you’re interested in. Different packet "cargo" values might help expose network problems. |
