| 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 |
Question: How can the inbound and outbound utilization have total
percentage 3.76%????
INBOUND UTILIZATION= 3.32
OUTBOUND UTILIZATION= 0.44
TOTAL UTILIZATION= 3.76
sho snmp PARKER_ST_W inter 2 all statis, for start 6:0 every 1:0 until
7:0 dur
1:0, in domain .rpt
!
!SNMP PARKER_ST_W Interface 2
!AT 6-JAN-1993 07:00:00 Statistics
!
!Examination of attributes shows:
! Duration = 3598 Secs
! Rcvd Octets = 5.731256E+06 Octets
! Inbound Packets = 12205 Packets
sho snmp PARKER_ST_W inter 2 all statis, for start 6:0 every 1:0 until
7:0 dur
1:0, in domain .rpt
!
!SNMP PARKER_ST_W Interface 2
!AT 6-JAN-1993 07:00:00 Statistics
!
!Examination of attributes shows:
! Duration = 3598 Secs
! Rcvd Octets = 5.731256E+06 Octets
! Inbound Packets = 12205 Packets
! Delivered Packets = 12205 Packets
! Average Size Of In Packet = 469.58 Octets/Packet
! Rcvd Octet Rate = 1592.9 Octets/Sec
! Inbound Packet Rate = 3.39 Packets/Sec
! Delivered Packet Percent = 100 Percent
! ********* Inbound Utilization = 3.32 Percent
! Delivered Non-unicast Packet Percent = 0 Percent
! Delivered Unicast Packet Percent = 100 Percent
! Delivered Non-unicast Packet Rate = 0 Packets/Sec
! Delivered Unicast Packet Rate = 3.39 Packets/Sec
! Xmitd Octets = 756577 Octets
! Outbound Packets = 6947 Packets
! Xmitd Packets = 6947 Packets
! Average Size Of Out Packet = 108.91 Octets/Packet
! Xmitd Octet Rate = 210.28 Octets/Sec
! Outbound Packet Rate = 1.93 Packets/Sec
! Xmitd Packet Percent = 100 Percent
! ******** Outbound Utilization = 0.44 Percent
! Outbound Non-unicast Packet Percent = 0 Percent
! Outbound Unicast Packet Percent = 100 Percent
! Outbound Non-unicast Packet Rate = 0 Packets/Sec
! Outbound Unicast Packet Rate = 1.93 Packets/Sec
! Total Octets = 6.487833E+06 Octets
! Total Packets = 19152 Packets
! Total Octet Rate = 1803.18 Octets/Sec
! ********* Total Utilization = 3.76 Percent
! Average Size Of Packet = 338.75 Octets/Packet
! Inbound Discarded Packets = 0 Packets
! Inbound Bad Proto Packets = 0 Packets
! Inbound Other Error Packets = 0 Packets
! Inbound Error Packets = 0 Packets
! Inbound Error Packet Percent = 0 Percent
! Inbound Bad Proto Packet Rate = 0 Packets/Sec
! Inbound Discarded Packet Rate = 0 Packets/Sec
! Inbound Other Error Packet Rate = 0 Packets/Sec
! Inbound Error Packet Rate = 0 Packets/Sec
! Inbound Bad Proto Packet Percent = 0 Percent
! Inbound Discarded Packet Percent = 0 Percent
! Inbound Other Error Packet Percent = 0 Percent
! Outbound Discarded Packets = 0 Packets
! Outbound Other Error Packets = 0 Packets
! Outbound Error Packets = 0 Packets
! Outbound Error Packet Percent = 0 Percent
! Outbound Discarded Packet Rate = 0 Packets/Sec
! Outbound Other Error Packet Rate = 0 Packets/Sec
! Outbound Error Packet Rate = 0 Packets/Sec
! Outbound Discarded Packet Percent = 0 Percent
! Outbound Other Error Packet Percent = 0 Percent
! Total Error Packets = 0 Packets
! Total Error Packet Percent = 0 Percent
! Delivered Non-unicast Packets = 0 Packets
! Delivered Unicast Packets = 12205 Packets
! Outbound Non-unicast Packets = 0 Packets
! Outbound Unicast Packets = 6947 Packets
! Total Error Packet Rate = 0 Percent
!
exit
[EOB]
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 4349.1 | wild guess. | TOOK::MCPHERSON | pre-retinal integration | Thu Jan 07 1993 12:48 | 11 |
What sort of interface is "Interface 2" ? The calculation Total Util = Inbound Util + Outbound Util would make sense for half-duplex links, but not full-duplex. Wouldn't an ethernet interface be a half-duplex link ? Just a guess. I don't know the internals of the PA FM... /doug | |||||
| 4349.2 | point to point full | CTHQ::HEALY | Thu Jan 07 1993 14:15 | 5 | |
Doug,
It's a point to point Full duplex...
Andy
| |||||
| 4349.3 | So much for that hypothesis... | TOOK::MCPHERSON | pre-retinal integration | Thu Jan 07 1993 14:29 | 6 |
That about does it for me... Time for better-qualified personnel to
intervene. (... he says as they march "Bingo: the amazing typing
wallabee over to my terminal...)
/doug
| |||||
| 4349.4 | Works as documented (but maybe not as expected). | MOLAR::CHRISB::BRIENEN | Network Management Applications! | Thu Jan 07 1993 18:20 | 26 |
> Question: How can the inbound and outbound utilization have total
> percentage 3.76%????
>
> INBOUND UTILIZATION= 3.32
> OUTBOUND UTILIZATION= 0.44
> TOTAL UTILIZATION= 3.76
According to the PA FM Use book (I'm looking at the Bookreader version
for PA V1.2, Appendix E: TCP/IP Statistical Function Algorithms), the
algorithm for "Total Utilization" is:
(ifInOctets+ifOutOctets) * bits
Total utilization = ------------------------------- * 100
ifSpeed * time
Looks like it's being calculated as documented in the Use book.
Are you saying: "I think the equation is wrong, and should be
calculated as <insert-formula-here> instead" ? (I've been
surprized myself at some of the algorithm's used...).
Maybe we should add some text to the Release Notes?
Chris Brienen
| |||||
| 4349.5 | reply to node .4 | CTHQ::HEALY | Mon Jan 11 1993 10:01 | 7 | |
Yes we understand the equation, I does not make sense. Add the inbound
and outbound and divide by 2, seeing the given utilization it does not
work out right. It should be 0 to 100 right now is looks like 0 to 200.
P.S. This is a full duplex circuit..
Andy Healy
| |||||
| 4349.6 | total = MAX(in,out) | BLUMON::SYLOR | Architect = Buzzword Generator | Wed Jan 13 1993 17:18 | 14 |
Andy is right, the formula doesn't make sense. Full duplex lines have two independent utilizations, adding them (or averaging them) doesn't provide useful info. Taking the MAX of inbound and outbound utilization might make sense, it would give a single number to alarm on. Next question. What protocol is being run over that interface - what's the value of ifType (Inteface Type)? Many of the stats don't make sense for certain types, you could discard the ones that aren't useful for this type of interface. The value of average size of packet includes all the overhead, including any extra packets that the protocol might introduce (acks etc). Mark | |||||
| 4349.7 | yes and no | CTHQ::WOODCOCK | Thu Jan 14 1993 13:33 | 37 | |
>Andy is right, the formula doesn't make sense. Full duplex lines have >two independent utilizations, adding them (or averaging them) doesn't >provide useful info. Taking the MAX of inbound and outbound utilization >might make sense, it would give a single number to alarm on. It does and then sometimes it doesn't. It would depend on who is asking for this number :-). In most instances I agree that the highest of the two should be considered TOTAL because ultimately from a link standpoint this is how bandwidth is managed. When either In or OUT are saturated the link gets an upgrade. This is the preferred number to give upper mngmt and bean counters. Ultimately though the differentiation between IN and OUT means real $$$$. The potential savings is enormous in bandwidth costs if the topology actually can be designed around getting IN closer to OUT with better load balancing. This situation means that upgrades will potentially occur less frequently and in this scenerio TOTAL is the AVG of IN and OUT which will tell you how well you are achieving the goal. While in theory this is the goal, reality is that this is an immensely difficult task not even thought of by most. >Next question. What protocol is being run over that interface - >what's the value of ifType (Inteface Type)? Many of the stats don't make sense >for certain types, you could discard the ones that aren't useful for this type >of interface. ifType = propPointToPointSerial To be honest we don't use TOTAL UTILIZATION from MCC, Andy was just being a good guy and pointing out something that didn't look quite right. >The value of average size of packet includes all the overhead, including >any extra packets that the protocol might introduce (acks etc). Good to know for future reference. cheers, brad... | |||||