Why another Ping utility?
Many Ping utilities are already available, one is even released with Windows itself, called Ping. But hrPing has some advanced features other Pings have not.
In short:
- Graphical display of ping results
- Uses high resolution timers, so ping times are accurate to the usec
- Can ping as well with UDP packets or ICMP timestamp messages
- Times and handles ICMP error replies as well
- Can have multiple pings “in-flight”, no need to wait for a reply before sending the next ping
- Improved statistics
- Size sweep: Send increasing packet sizes
- Can show only a summary of results
- Is a Traceroute and a Pathping as well
You can do much more with hrPing than with Windows Ping.
More text, please.
Like every Ping, hrPing sends “ICMP Echo Request” packets to the remote computer and listens to the matching “Echo response” packets. What’s more, hrPing can send UDP packets and ICMP timestamp packets as well. Not all packet types pass all firewalls and networks equally easy. With hrPing you have the possibility to vary. (-M and -u switches)
What’s more, hrPing times the round trip delay in microseconds (1/1000 msec). This is usually done by using the Windows’ “Performance Counter” which has a resolution of some MHz. You can even ask hrPing to use the CPU’s “Time Stamp Counter” which is incremented with the CPU’s clock cycle. You can’t get any more accurate with standard PCs today!
The next thing Windows Ping can not do is send more than one ping packet at a time. Windows Ping always sends one packet, waits for the reply, then prints its output line, repeat.
hrPing sends out one ping packet every x milliseconds (you can adjust this time with the -s parameter) while listening for incoming replies and printing the output if there is any.
The reason why you should like this is easy: with broadband you often have a delay of some 40 msec, while the upstream bandwidth of the whole connection is some 500 kbytes/sec. So, with a “standard” ping packet of 60 bytes (IP header + ICMP header + ping payload) you can send thousands of packets before you get the first reply. If you want to test line conditions, throughput, etc. this “overlapped” way of sending is really helpful.
Plus, hrPing has much better statistics than Windows Ping. You get the round trip times for ICMP error message replies as well! This way you can e.g. monitor the delay of a TTL exceed. hrPing counts the replies and error messages separately, so the global statistics don’t mess up one another. Plus, for the statistically inclined, hrPing calculates the standard deviation as well, to show you how much the values “jiggle”. hrPing shows you the standard deviation of the timings as well as the average times.
hrPing displays the IP identification field of the replies and thus makes it possible to do “silent load measurements”; see german c’t magazine 23/2003, p. 212 for details. (-I switch)
When sending a lot of packets hrPing’s “Summary mode” comes in handy: it will suppress the printing of each reply on its own line, but instead print a summary for all replies so far and a summary for the last 10 seconds (time can be adjusted). This lets you keep a nice overall view. (-y switch)
Is there more? Yes: hrPing can send out pings with increasing sizes: the “Size Sweep” function, where after each send, the size is increased until it reaches a maximum, then reset. Plus, we do some math when processing the replies and estimate the line speed, if the data is conclusive enough. (-l and -L switches)
Ah, one more thing: hrPing is a traceroute (-r switch) and a pathping (-p switch) as well 🙂
There’s a lot more goodies hidden in hrPing, just use it and you will find out about small but useful features.
And this is how to use it:
hrPING [<options>] <host>
<host> may be the IP address or the hostname. In the latter case the name will be resolved to its address at the beginning of the ping loop.
There are a couple of options that let you specify the data you’re sending:
| -f | Set Don’t Fragment flag in packet. |
Set the “Don’t fragment” bit in the IP header of the PING packet. Default is not set.
Set the “Time To Live” value in the IP header of the PING packet. Default is 255.
Set the “Type Of Service” bits in the IP header of the PING packet. Default is 0. It is possible that Windows erases or overwrites this field when sending the packet. Furthermore, TOS is deprecated nowadays. hrPing will use IP_HDRINCL option to set TOS.
| -l size | Send buffer size (ICMP payload size). |
How may bytes payload should be send? Remember that each packet is of the form: IP header (20 bytes) + ICMP header (8 bytes) + payload. You may only specify the payload size. Minimum is 0, maximum is 64k-1-20-8, i.e., 65507 bytes. Default is 32 bytes.This works for UDP mode as well, but not for ICMP timestamp mode. There the ICMP length is always 10 bytes.
| -l s1[:s2[:i]] | Size sweep: send buffer size from s1 to s2 step i |
With this syntax hrPing will start sending a payload of s1 bytes, increase the payload by i bytes for each send (if i was set, otherwise increment by 1) until s2 is reached or exceeded and then restart with s1 bytes. This switches the correlation calculation on that tries to see if there is a correlation between the size of the packet you send and the time it takes for a reply.
| -L size | Total IP datagram size (ICMP payload size + 28). |
Same as the above, only that this size here is the size for the total IP datagram.
| -L s1[:s2[:i]] | IP datagram size (payload size + 28, default 60) [with sweep] |
Analogue as above for -l s1[:s2[:i]]
| -M | Send ICMP timestamp requests |
This will send out ICMP timestamp requests and print ICMP timestamp replies. In the reply there is the send time off the other side in milliseconds, so it’s possible to distinguish between delay that was caused sending and delay that was caused when receiving. These numbers have only millisecond resolution. To synchronise clocks hrPing will calculate the clock offset from the first ICMP timestamp reply, assuming that half of the delay came from each direction.
| -u [port] | Send UDP packets (port 7 by default) |
You can send UDP packets as well. This causes UDP packets to be send to and from port port. Hopefully, the other side has no UDP port under that port number and will reply with a “port unreachable” message, which will be counted as a proper reply.
Options that allow you to specify how hrPing operates:
| -t | Ping the specified host until stopped. |
Loop forever. You can abort hrPING any time with CTRL-C or CTRL-Break. Unlike Windows PING, hrPING will still print the statistics gathered so far when you abort. CTRL-C waits for some time for replies still to come in before it aborts. If you are fed up with waiting, press Ctrl-C 5 times. Ctrl-Break just prints the statistics, but doesn’t abort. That’s nice in quiet mode or with many replies.
| -n count | Number of echo requests to send. |
Specify the number of PING packets to send. Default number is 4. For traceroute or pathping modes (see below) this specifies the number of pings per hop. For tracerroute mode, the default number is 3.
| -w timeout | Timeout in milliseconds to wait for each reply. |
Maximum timeout to wait for a reply. This time applies when hrPing is waiting for the last replies to come in. Furthermore, hrPing will even count a reply as timeouted if it took too long. Default is 2000 milliseconds.
| -s time | Interval in milliseconds between packets. |
This is the number of milliseconds between sending of two PING packets. hrPINGwill try to stick to this number quiet accurately. If sending took a little longer for one packet it will send out the next packet a little earlier. Default is 500 milliseconds. (You can use decimals for a very fine grained interval: -s5.4 will send a packet every 5400 microseconds on average.)
| -c [num] | Concurrent sending of up to num pings at a time (default 1) |
Try to keep a maximum of num pings in transmission without a reply. When either a reply comes in or a timeout occurs, send more pings, so num pings are again in flight.-c (without a number) effectively turns off overlapped send/receive, since a second ping is only sent after a reply to the first or an timeout. This, together with the -w option make hrPing behave like Windows’ ping.
| -r [count] | Be a traceroute (do count pings each hop, default 3) |
hrPing contains a traceroute utility! It works almost the same as Windows TRACERT, except that you can specify how many packets are sent per hop, default is three. By default, IP addresses are not resolved to names. Use -a to do that.
| -a [hop] | Resolve addresses to names for traceroute (start at hop) |
This tries to resolve the IP addresses into DNS names. If you specify hop hrPing will not try to resolve the first hop-1 hops, since they are often not resolvable and it only costs time to try (and who has time nowadays?)
| -p | Trace path to destination, then ping all hops on path |
Do a path ping: work like traceroute to get the adresses on the path to a given destination, then ping each of them individually (-u and -M honoured as well as other options that make sense (-l, even with size sweep, -c, -s, -w, etc). At the end, print some statistics for all hops on the path.
And some options control the output:
| -lic | Show public license and warranty. |
We need you to accept the software license. This is done the first time you start hrPING. If you want to re-read it, use this option.
| -fwhelp | Print firewall help text |
Remarks on how to set up the Windows firewall to let hrPing’s packets pass.
| -F file | Log output into file as well, even if -q is set |
All output is logged to file file as well as to the screen. If -q options are set, all output goes to log file, even if it’s not printed to the screen.
| -T | Print timestamp in front of each line |
Preceed each line of output with a timestamp of the form
“2012-05-22 18:19:53.508: “
| -q[r|e|t] | Be quiet (-qr=no replies, -qe=no errors, -qt=no timeouts) |
Be quiet. Use -qr to not print replies, -qe to not print IMCP error messages or -qt not to print timeouts.
| -y [sec] | Print summary of the last sec secs (default 10) |
Be quiet, but print a summary of all packets (with counts and statistics) and one of packets of the last sec seconds. Useful with small -s send delays or long sends (-t or high -n).
Show graph of the ping times. To do that, grping.exe is started. More on that later on. Use -gg to close graph on hrping exit. Use -G to utilize a already running grping.exe.
Prints a help screen, -?? or -hh prints an even longer one.
And then, we have options for the connoisseur:
| -I id | Set ICMP id field to id |
Set the “Identification” IP header field to the value specified. It is possible that Windows erases or overwrites this field when sending the packet.
| -W | “warm up” with one uncounted echo request at beginning |
If specified, hrPING will send one uncounted ping before all others. This “warm up” is useful with some firewalls that somehow cause the first block to be much slower than the following ones. -s, -w apply.
| -A | Abort after the first echo reply (-AA => or error) |
Loop as long as there are no proper replies (or even error messages if -AA).
| -H | Use IP_HDRINCL socket option (default for UDP or with TOS) |
With this option set, hrPing sets up his IP headers for send itself, otherwise Windows does it. -H is selected automatically with -u or -v, because it doesn’t work otherwise.
| -E file | Stop pinging if file exists |
This is nice for batch files or for coordinating with a background job. hrPing will loop as long as usual (i.e. depending on -t or -n options), but will furthermore check for the existence of file. If file comes into existence, hrPing will exit the loop.
| -diag | Diagnose your connectivity: send different packet types |
Send all kinds of different packets (ICMP echo, ICMP timestamp, UDP) with high and low TTL to the destination and see if we get an answer. That might be useful to check what is coming through your firewall or ISP.
| -K | Wait for any key press when done |
Useful if hrPing is started from Windows Explorer.
| -qr time | Suppress all replies under time msec |
Suppress all reply outputs under time msec, but write them still to the logfile, if provided.
| -O ofs | Set time offset in msec for timestamp mode |
Specify the time offset between your time and the receiver’s time by hand, otherwise hrPing will do that on the first reply.
Well, somewhere debug info is printed. I don’t want to be too specific. Actually, it’s a secret. 🙂
You can mess around with the timer, if you want:
| -perfcnt | Use performance counter (default) |
Use the performance counter of Windows. Usually, it’s pretty accurate (some MHz). That’s the default.
hrPING automatically decides if it uses the CPU’s timestamp counter (TSC) or the operating system’s performance counter for timings. On some CPU’s the TSC is not reliable, since it doesn’t tick at the same speed all the time. On multiprocessor systems, not all TSC have to tick exactly in sync. In almost all cases, hrPING will use the performace counter. If you want to force TSC usage, use -tsc, but hrPing will only use the TSC if it thinks it’s accurate.
| -mmtime | Use multimedia timer; not very accurate |
Use Windows’ multimedia timers. They should be at least accurate to the msec.
| -tick | Use Windows GetTickCount; very inaccurate! |
Use the standard Windows timer. That usually has a resoltion of 15 msec or so. Berk!
| -prec | Measure timer precision |
Not sure how accurate the timer is? Measure it!
Return codes:
For ping mode:
- 0
- All sent pings were answered with their proper replies.
- 1
- All sent pings were answered with their proper replies or ICMP error messages.
- 2
- At least some pings were answered with either their proper replies or ICMP error messages.
- 3
- No replies at all.
- 9
- Error.
For traceroute mode:
- 0
- Reached destination
- 2
- At least some pings were answered with either their proper replies or ICMP error messages.
- 3
- No replies at all.
- 9
- Error.
Now a test run:
Ping us:
C:\> hrPING www.example.com
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=cc09
Pinging www.example.com [123.45.67.89]
with 32 bytes data (60 bytes IP):
From 123.45.67.89: bytes=60 seq=0001 TTL=55 ID=becc time=42.803ms
From 123.45.67.89: bytes=60 seq=0002 TTL=55 ID=becd time=43.768ms
From 123.45.67.89: bytes=60 seq=0003 TTL=55 ID=bece time=44.842ms
From 123.45.67.89: bytes=60 seq=0004 TTL=55 ID=becf time=43.449ms
Packets: sent=4, rcvd=4, error=0, lost=0 (0.0% loss) in 1.546366 se
RTTs in ms: min/avg/max/dev: 42.803 / 43.715 / 44.842 / 0.737
Bandwidth in kbytes/sec: sent=0.155, rcvd=0.155
You see that hrPing numbers the packets in ascending order. The sequence numbers of the replies are listed (if there are out-of-sequence packets, hrPing will write “SEQ=” instead of “seq=”, so you notice).
Notice the times in microseconds (milliseconds with 3 decimals)
Furthermore, we see the TTL, which is the TTL of the sender, minus the number of hops the packet took to come here. I guess TTL was initially set to 64, so it took the packet 10 hops to come here (64-55+1=10, TTL is only decremented on forward, not on the first send).
Plus, hrPing prints the number of bytes in the received packet and the IP identification field.
Please also notice the statistics with average and standard deviation (which is a measure of how wide the data points are spread out. So 4, 6, 4, 6 has a smaller deviation than 1, 9, 1, 9, even though the average is the same).
Now let’s look at a second test run:
Try to ping us:
C:\> hrPING -i1 www.example.com
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=140d
Pinging www.example.com [123.45.67.89]
with 32 bytes data (60 bytes IP), TTL 1:
From 192.168.2.1: TTL count exceeded; bytes=56 seq=0001 TTL=64 ID=33b7 time=0.990ms
From 192.168.2.1: TTL count exceeded; bytes=56 seq=0002 TTL=64 ID=33b8 time=1.237ms
From 192.168.2.1: TTL count exceeded; bytes=56 seq=0003 TTL=64 ID=33b9 time=0.578ms
From 192.168.2.1: TTL count exceeded; bytes=56 seq=0004 TTL=64 ID=33ba time=1.041ms
Packets: sent=4, rcvd=0, error=4, lost=0 (0.0% loss) in 1.514205 sec
RTTs in ms: min/avg/max/dev: 0.578 / 0.961 / 1.237 / 0.239
Bandwidth in kbytes/sec: sent=0.158, rcvd=0.147
Notice that this is a list of ICMP error messages, yet there are still the sequence numbers and round-trip-times listed.
Measure the delay of your connection:
We use the smallest ping available and we use the shortest route available. We could use traceroute mode to find the first hop and ping that one. But experience shows that the hops often don’t answer to pings. So we use a trick: we send out packets with TTL 2: they will be bounced on the first external hop, since hop 1 is my internal router (use TTL 1 instead if you are directly connected to the Internet). (The trick is not new, this is how traceroute works.)
C:\> hrping -i2 -l0 www.example.com
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=a40e
Pinging www.example.com [123.45.67.89]
with 0 bytes data (28 bytes IP), TTL 2:
From 217.0.119.53: TTL count exceeded; bytes=56 seq=0001 TTL=254 ID=0000 time=17.168ms
From 217.0.119.53: TTL count exceeded; bytes=56 seq=0002 TTL=254 ID=0000 time=16.634ms
From 217.0.119.53: TTL count exceeded; bytes=56 seq=0003 TTL=254 ID=0000 time=16.740ms
From 217.0.119.53: TTL count exceeded; bytes=56 seq=0004 TTL=254 ID=0000 time=16.342ms
Packets: sent=4, rcvd=0, error=4, lost=0 (0.0% loss) in 1.529537 sec
RTTs in ms: min/avg/max/dev: 16.342 / 16.721 / 17.168 / 0.296
Bandwidth in kbytes/sec: sent=0.073, rcvd=0.146
So, the minimum delay seems to be some 16.3 ms. If you increase the number of tries you might find an even smaller delay, but it’s not very likely it will shrink a lot more:
C:\> hrping -i2 -l0 -n1000 -q -s20 www.example.com
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=040b
Pinging www.example.com [123.45.67.89]
with 0 bytes data (28 bytes IP), TTL 2:
Packets: sent=1000, rcvd=0, error=1000, lost=0 (0.0% loss) in 20.001239 sec
RTTs in ms: min/avg/max/dev: 15.501 / 16.181 / 19.551 / 0.547
Bandwidth in kbytes/sec: sent=1.399, rcvd=2.799
So, the real minimum seems to be around 15.5 ms. We might want to run hrPing all the time to see how well our line is. We can do that and instruct hrPing to be silent and only display a summary:
C:\> hrping -i2 -l0 -t -y www.example.com
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=1807
Pinging www.example.com [123.45.67.89]
with 0 bytes data (28 bytes IP), TTL 2:
Total:
Packets: sent=85, rcvd=0, error=85, lost=0 (0.0% loss) in 42.042626 sec
RTTs in ms: min/avg/max/dev: 15.815 / 18.558 / 34.122 / 5.022
Bandwidth in kbytes/sec: sent=0.056, rcvd=0.113
Last 10 seconds:
Packets: sent=20, rcvd=0, error=20, lost=0 (0.0% loss) in 9.531548 sec
RTTs in ms: min/avg/max/dev: 16.040 / 23.758 / 34.122 / 7.203
Bandwidth in kbytes/sec: sent=0.058, rcvd=0.117
[Aborting...]
Packets: sent=85, rcvd=0, error=85, lost=0 (0.0% loss) in 42.042626 sec
RTTs in ms: min/avg/max/dev: 15.815 / 18.558 / 34.122 / 5.022
Bandwidth in kbytes/sec: sent=0.056, rcvd=0.113
This will continue to show a summary (to end it I pressed Ctrl-C). The summary shows that we had an increased ping time in the last 10 seconds. This is no wonder, since I ran an upload at the same time.
What is my line speed?
hrPing has this nice size sweep feature. Let’s try it:
C:\> hrping www.example.com -l0:1400:96 -n20 -s20
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=e80d
Pinging www.example.com [123.45.67.89]
with 0-1400 bytes data (28-1428 bytes IP):
From 123.45.67.89: bytes=28 seq=0001 TTL=55 ID=bf45 time=43.591ms
From 123.45.67.89: bytes=124 seq=0002 TTL=55 ID=bf46 time=42.699ms
From 123.45.67.89: bytes=220 seq=0003 TTL=55 ID=bf47 time=44.006ms
From 123.45.67.89: bytes=316 seq=0004 TTL=55 ID=bf48 time=43.198ms
From 123.45.67.89: bytes=412 seq=0005 TTL=55 ID=bf49 time=44.212ms
From 123.45.67.89: bytes=508 seq=0006 TTL=55 ID=bf4a time=43.859ms
From 123.45.67.89: bytes=604 seq=0007 TTL=55 ID=bf4b time=44.744ms
From 123.45.67.89: bytes=700 seq=0008 TTL=55 ID=bf4c time=45.763ms
From 123.45.67.89: bytes=796 seq=0009 TTL=55 ID=bf4d time=44.257ms
From 123.45.67.89: bytes=892 seq=000a TTL=55 ID=bf4e time=46.415ms
From 123.45.67.89: bytes=988 seq=000b TTL=55 ID=bf4f time=46.215ms
From 123.45.67.89: bytes=1084 seq=000c TTL=55 ID=bf50 time=45.993ms
From 123.45.67.89: bytes=1180 seq=000d TTL=55 ID=bf51 time=46.744ms
From 123.45.67.89: bytes=1276 seq=000e TTL=55 ID=bf52 time=46.761ms
From 123.45.67.89: bytes=1372 seq=000f TTL=55 ID=bf53 time=47.214ms
From 123.45.67.89: bytes=28 seq=0010 TTL=55 ID=bf54 time=43.284ms
From 123.45.67.89: bytes=124 seq=0011 TTL=55 ID=bf55 time=43.345ms
From 123.45.67.89: bytes=220 seq=0012 TTL=55 ID=bf56 time=44.738ms
From 123.45.67.89: bytes=316 seq=0013 TTL=55 ID=bf57 time=44.561ms
From 123.45.67.89: bytes=412 seq=0014 TTL=55 ID=bf58 time=44.263ms
Packets: sent=20, rcvd=20, error=0, lost=0 (0.0% loss) in 0.434017 sec
RTTs in ms: min/avg/max/dev: 42.699 / 44.793 / 47.214 / 1.337
Bandwidth in kbytes/sec: sent=26.727, rcvd=26.727
Correlation: 91.6%, estimated speed: 281.11 kbytes/sec
As you can see, the “bytes=” are increasing, since the proper reply to a ping packet with N bytes is a pong packet with N bytes. Furthermore, you may notice that the reply time is increasing slightly with the packet size. We do some statistics on it (linear regression and correlation) and get a correlation coefficient: 91.6%. Generally, a correlation of 50% or more is called “correlated”, of 80% or more “highly correlated”. That means: the larger the packet, the larger the time and time is proportional to packet size. hrPing estimates the speed to be some 280 kbytes/s. As a matter of fact, that is too slow, but measurements show smaller throughputs the longer the route is. Let’s try the same with hop 2 (the first external hop):
C:\> hrping www.example.com -l0:1400:96 -n20 -s20 -i2
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=9c01
Pinging www.example.com [123.45.67.89]
with 0-1400 bytes data (28-1428 bytes IP), TTL 2:
From 217.0.119.53: TTL count exceeded; bytes=56 seq=0001 TTL=254 ID=0000 time=16.164ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0002 TTL=254 ID=0000 time=16.125ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0003 TTL=254 ID=0000 time=16.271ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0004 TTL=254 ID=0000 time=16.524ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0005 TTL=254 ID=0000 time=17.250ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0006 TTL=254 ID=0000 time=16.856ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0007 TTL=254 ID=0000 time=17.577ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0008 TTL=254 ID=0000 time=17.362ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0009 TTL=254 ID=0000 time=18.080ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=000a TTL=254 ID=0000 time=17.851ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=000b TTL=254 ID=0000 time=18.158ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=000c TTL=254 ID=0000 time=18.372ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=000d TTL=254 ID=0000 time=19.002ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=000e TTL=254 ID=0000 time=18.442ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=000f TTL=254 ID=0000 time=18.895ms
From 217.0.119.53: TTL count exceeded; bytes=56 seq=0010 TTL=254 ID=0000 time=16.676ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0011 TTL=254 ID=0000 time=16.259ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0012 TTL=254 ID=0000 time=16.400ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0013 TTL=254 ID=0000 time=16.941ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0014 TTL=254 ID=0000 time=17.573ms
Packets: sent=20, rcvd=0, error=20, lost=0 (0.0% loss) in 0.407530 sec
RTTs in ms: min/avg/max/dev: 16.125 / 17.338 / 19.002 / 0.911
Bandwidth in kbytes/sec: sent=28.464, rcvd=3.808
Correlation: 95.2%, estimated speed: 428.86 kbytes/sec
This is more like it (I have about 500 kbytes/sec upstream). The estimated speed is only shown if there is a correlation of at least 40%, otherwise the estimation is too vague. Linear regression is very sensitive to stray data. Often you ping 50 times and get 49 good replies (say in the 50 msecs) and one that is way too high (say 150 msec). Normally, this would badly offset your correlation and linear regression. But if you know a reasonable upper limit for your replies, you can filter out the stray replies with the -w option, which sets the maximum time to wait for a reply. The trick here is that hrPing counts replies as timeouts, even if they arrived back, but it took longer than the timeout time. In our above example, you may use a -w60 to filter out the stray 150 msec peak. If you have no idea what timeout time to use, try a -w with the average time plus 2-3 times the deviation.
Graphics!
To have hrPing open a window as well and graph the results type this:
C:\> hrping www.example.com -t -i2 -g
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
Source address is 192.168.2.106; using ICMP echo-request, ID=cc09
Pinging www.example.com [123.45.67.89]
with 32 bytes data (60 bytes IP), TTL 2:
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0001 TTL=254 ID=0000 time=16.494ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0002 TTL=254 ID=0000 time=16.673ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0003 TTL=254 ID=0000 time=16.017ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0004 TTL=254 ID=0000 time=16.532ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0005 TTL=254 ID=0000 time=16.969ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0006 TTL=254 ID=0000 time=16.910ms
From 217.0.119.53: TTL count exceeded; bytes=80 seq=0007 TTL=254 ID=0000 time=16.335ms
[Aborting...]
Packets: sent=7, rcvd=0, error=7, lost=0 (0.0% loss) in 3.027021 sec
RTTs in ms: min/avg/max/dev: 16.017 / 16.561 / 16.969 / 0.305
Bandwidth in kbytes/sec: sent=0.138, rcvd=0.185
This will open a window and show the ping times. After aborting, the window will stay open. In the window you can select the time scale that is shown and you can choose to show an average of the values, with different average times. If you like the window to be automatically closed when hrPing closes (even on Ctrl-C), use -gg instead of -g.
You can have multiple graphs in one window
Start the first hrPing regularly:
C:\> hrping www.example.com -t -i2 -g
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
[...]
After that, start the second hrPing with a -G instead:
C:\> hrping www.example.com -t -G
This is hrPING v5.00 by cFos Software GmbH -- http://www.cfos.de
[...]
This will produce two graphs in the same window. Nice! 🙂 You find the hrPing textual reply lines annoying? Add a -y to get a summary or a -q to leave it away altogether.
System requirements:
hrPING should work well on all systems running Windows XP and above (Vista, Windows 7, Windows 8 previews and the server OSes Server 2003 and 2008 as well).
Usually, you have to be a member of the Administrator group to run hrPing, since hrPing uses “raw sockets”.
Under XP, you can open access to raw sockets for every user by setting the following Registry key under HKEY_LOCAL_MACHINE to 1 (DWORD): System\CurrentControlSet\Services\Afd\Parameters\DisableRawSecurity This feature was removed in Vista. 🙁
