| Title: | DECmcc user notes file. Does not replace IPMT. |
| Notice: | Use IPMT for problems. Newsletter location in note 6187 |
| Moderator: | TAEC::BEROUD |
| Created: | Mon Aug 21 1989 |
| Last Modified: | Wed Jun 04 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 6497 |
| Total number of notes: | 27359 |
Hello,
I had done a simple test earlier comparing PA stats to manually
computing via counters and all seemed well for a single duration
of 15 minutes. The two methods produced numbers quite close at least
for packet rates.
I'm now using MCC to determine packet rates on our current network to
better determine our future needs and how routers of the future will
hold up.
There are two logs below, one which captures the packet rate each minute
for an hour and one which captures the statistics as an average for the full
hour (each run for the same hour). The total hourly average indicates a
packet rate of 209.76 pps. But looking at the 1 minute averages it appears
the total average of each of the 1 minute intervals is somewhere under
100pps. They aren't even close when I would think they should be almost
identical. Why??? Can someone please tell me where I'm going wrong here.
Am I polling too quickly and MCC isn't using actual time for computations
but the user specified duration?
thanks,
brad...
.............................................................................
$mcc
show node4 zklan1 packet rate,at every 0:1:0 until +1:0:0, -
for start (-0:00) duration 0:1:0,to file=[decmcc.router_test]zklan1_pr_2.txt
***Output from zklan1_pr_2.txt***
Node4 zklan1
AT 20-MAY-1991 14:02:21 Statistics
Packet Rate = 4.2 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:04:20 Statistics
Packet Rate = 119.72 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:05:28 Statistics
Packet Rate = 99.43 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:07:20 Statistics
Packet Rate = 67.28 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:09:20 Statistics
Packet Rate = 35.95 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:10:30 Statistics
Packet Rate = 125.85 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:12:20 Statistics
Packet Rate = 126.13 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:13:28 Statistics
Packet Rate = 116.98 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:15:20 Statistics
Packet Rate = 128.87 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:16:29 Statistics
Packet Rate = 124.6 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:18:20 Statistics
Packet Rate = 135.75 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:19:27 Statistics
Packet Rate = 114.67 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:21:20 Statistics
Packet Rate = 37.2 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:22:28 Statistics
Packet Rate = 110.17 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:24:20 Statistics
Packet Rate = 131.57 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:25:26 Statistics
Packet Rate = 84.27 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:27:20 Statistics
Packet Rate = 54.87 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:29:20 Statistics
Packet Rate = 117.78 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:31:20 Statistics
Packet Rate = 131.88 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:32:27 Statistics
Packet Rate = 73.65 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:34:20 Statistics
Packet Rate = 112.55 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:36:20 Statistics
Packet Rate = 75.2 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:38:20 Statistics
Packet Rate = 89.85 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:39:28 Statistics
Packet Rate = 24.57 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:41:20 Statistics
Packet Rate = 118.67 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:43:20 Statistics
Packet Rate = 87.87 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:44:28 Statistics
Packet Rate = 95.5 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:46:20 Statistics
Packet Rate = 94.6 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:48:20 Statistics
Packet Rate = 81.18 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:49:27 Statistics
Packet Rate = 30.1 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:51:20 Statistics
Packet Rate = 102.42 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:52:27 Statistics
Packet Rate = 114.03 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:54:20 Statistics
Packet Rate = 29.23 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:56:20 Statistics
Packet Rate = 112.97 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:58:20 Statistics
Packet Rate = 109.55 Packets/Sec
Node4 zklan1
AT 20-MAY-1991 14:59:31 Statistics
Packet Rate = 89.88 Packets/Sec
............................................................................
$mcc
show node4 zklan1 all statistics,for duration 1:0:0, -
to file=[decmcc.router_test]zklan1_stats_2.txt
***Output from zklan1_stats_2.txt***
Node4 zklan1
AT 20-MAY-1991 15:00:22 Statistics
Examination of Attributes shows:
Duration = 3600 Seconds
Originating Packets Sent = 3700
Forwarding Packets Received = 747688
Terminating Packets Received = 3740
Packet Rate = 209.76 Packets/Sec
Originating Packets Sent Percent = 0.49
Forwarding Packets Received Percent = 99.01
Terminating Packets Received Percent = 0.5
Count of Aged Packets Lost = 225
Count of Node Unreachable Packets Lost = 6
Count of Node Out Of Range Packets Lost = 0
Count of Terminating Congestion Packets Lost = 0
Count of Transit Congestion Packets Lost = 0
Count of Packet Format Error = 0
Count of Partial Routing Update Lost = 0
Count of Verification Rejects = 0
Lost Packet Percent = 3.000000E-02
Count of Oversized Packets Lost = 0
Count of Response Timeouts = 0
Count of Receive Connect Resource Errors = 0
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 1034.1 | duration difference accounts for some | TOOK::CALLANDER | Tue May 21 1991 14:09 | 25 | |
The difference in the durations used (1minute in one, and an hour in the other) will account for some of the discrepency, since the hour will give a much more accurate view of the traffic. Also the fact that the duration time and the scheduling time are the name (every minute and duration 1 minute), the duration (first show of the stats) is taking longer than the every clause allows for so you are actually only going out every other minute. You can see this my looking at the time stamps on each of the shows. This will in some instances hide some of the peaks. I am a bit surprised at the size of the descrepancy though. Maybe some one with more knowledge of what real time packet traffic looks like could help fill us in on if this large of a difference in rates, based upon the formulas in the manuals, make sence. To get a bit closer look at what you seemed to be trying to do, attempt the same basic thing with the following commands: ......................................... $mcc show node4 zklan1 packet rate,at every 0:1:10 until +1:0:0, - for start (-0:00) duration 0:1:0,to file=[decmcc.router_test]zklan1_pr_2.txt show node4 zklan1 packet rate,- for start (-0:00) duration 1:0,to file=[decmcc.router_test]zklan1_pr_1.txt | |||||
| 1034.2 | an update | JETSAM::WOODCOCK | Tue May 28 1991 19:57 | 56 | |
re:-.1 The difference in the durations used (1minute in one, and an hour in the other) will account for some of the discrepency, since the hour will give a much more accurate view of the traffic. > Why would it be much more accurate? I could see it being a little off > but this is too much. Could someone please explain the polling and > computational process involved here. Does MCC poll once a minute and > subtract last_counters from present_counters and do a computation? Or > does it execute this command as seperate 1 minute commands and use two > polls for each minute?? Also the fact that the duration time and the scheduling time are the name (every minute and duration 1 minute), the duration (first show of the stats) is taking longer than the every clause allows for so you are actually only going out every other minute. You can see this my looking at the time stamps on each of the shows. This will in some instances hide some of the peaks. > This is not consistent with what I'm seeing. In other words if the > command reflects every 2 minutes and a duration of 2 minutes what I get > back is mostly around (3-4 seconds over) 2 minutes and an occasional > skipped interval (but not every other one). I am a bit surprised at the size of the descrepancy though. Maybe some one with more knowledge of what real time packet traffic looks like could help fill us in on if this large of a difference in rates, based upon the formulas in the manuals, make sence. To get a bit closer look at what you seemed to be trying to do, attempt the same basic thing with the following commands: ......................................... $mcc show node4 zklan1 packet rate,at every 0:1:10 until +1:0:0, - for start (-0:00) duration 0:1:0,to file=[decmcc.router_test]zklan1_pr_2.txt show node4 zklan1 packet rate,- for start (-0:00) duration 1:0,to file=[decmcc.router_test]zklan1_pr_1.txt > I tried a number of different schemes. The biggest impact was to simply > increase the poll interval even with leaving the duration=interval. By > just raising both to 2 minutes they came in very close. I also tried > increasing the interval to 10 seconds higher than the duration, this was > again close but didn't seem to produce any more accuracy. Finally it seems > that the longer commands of this nature run the further off the accuracy > becomes. As an example if I do 2 minute intervals/durations for 10 minutes > the numbers were acceptably close (290.06 compared to 292.44 pps). But when > this same commands were run for an hour accuracy appears unacceptable > (349.24 compared to 375.31 pps). The averaged short intervals always > produces the smaller number. Can anyone shed a bit more light on this > problem?? thanks, brad... | |||||
| 1034.3 | zko config | JETSAM::WOODCOCK | Thu May 30 1991 20:45 | 11 | |
I was asked to provide an idea of the configuration. There is a map in ANCHOR::NET$MAPS:AREA02.LN03 This is a pictorial view of the backbone WAN connects and the related connects to the site LAN. This is not up to date but will give you the general idea of connectivity. My polls are coming in from one of the WAN connects (I didn't check which one but it shouldn't matter). regards, brad... | |||||
| 1034.4 | Investigating | TOOK::ANWARUDDIN | Anwar | Fri Jun 28 1991 19:30 | 3 |
We are investigating the reason for the discrepancy. We did experiment with different durations and we did find some discrepancy which, depending on the traffic, varied from 5 - 20%. | |||||