Bandwidth utilization Calculation

Unanswered Question
Dec 12th, 2007

Hello,

I want small help for the calculation of Bandwidth utilization calculation.I have read the cisco document but still I am confused with that as I'm not understanding the poll cycles interval.Help me with the formula too..

Help me with these..

Thanks,

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 2.7 (3 ratings)
Loading.
David Stanford Wed, 12/12/2007 - 06:21

I assume you are reviewing this document:

http://www.cisco.com/en/US/tech/tk648/tk362/technologies_tech_note09186a008009496e.shtml

Here's an example:

You poll ifInOctets and ifOutOctets every 10 seconds. 10 Seconds is your poll cycle.

ifSpeed should remain the same so you should not need to poll again.

So the first time you poll ifInOctets the result is 100 an ifOutOctets is 50.

After 10 seconds you poll again and ifInOctets is 110 and ifOutOctets is 70.

This is where you calculate the delta for the formula.

Take the second result of each and subtract the first result:

ifInOctets --> 110 - 100 = 10

ifOutOctets --> 70 - 50 = 20

Then your formula would be:

(10 + 20) x 8 x 100 / (10 seconds) x ifSpeed

freebird123 Wed, 12/12/2007 - 18:50

Hello ,

Thx for the reply.I have tried the same way as you mentioned.But if at some time say after 10 seconds we are polling and if the no of octets previously polled are greater than the recently polled then that would be negative result..I am asking because i got this result too..so what to do for this situation??any sujjestions??

David Stanford Wed, 12/12/2007 - 21:32

Not sure what you are getting at here. The second poll should have an increase in the counter value...and you calculate the value

So 100 first time, 110 second time would be a delta value of 10

Now if the first poll was 100 and the second was 80 we have two possible causes:

1) high speed interface with 32-bit counters wrapping

2) IOS bug showing negative values

freebird123 Wed, 12/12/2007 - 21:42

Hello ,

If I have understood the reason better,if its for 32 bit counters wrapping,how would i resolve the problem.As its not giving me negative values all the time.

David Stanford Fri, 12/14/2007 - 08:36

What speed interfaces are these? If these are high speed you should be using HC counters.

freebird123 Sun, 12/16/2007 - 18:54

These are Gigabit interfaces.

But the thing is if its a 32 bit counter then according to formula negative values at some instant of time are for sure for del(ifInOctets) & also for OutOctets .

Kindly suggest me other way which will give me finer values.

Thanks

freebird123 Mon, 02/18/2008 - 20:38

Hi,

I have one more doubt on this bandwidth calculation.Can u let me know if there is any other oid which tells the total ifIn & IfOut traffic.

Thanks,

Regards

GERARD PUOPLO Sun, 02/24/2008 - 19:14

Most tools (e.g. MRTG, VitalNet etc) that use SNMP to calculate bandwidth utilization suffer a certain degree of inaccuracy that gets large and larger as the speed of the interface traffic increases and the poll cycle decreases.

The results are usually ok for trending but here is the issue most folks forget. Most tools do not usually poll for sysuptime when they poll for HC ifIn and ifOut octets. They are calculating delta time based on server time delta. SNMP timeout and retry occurrences (especially on high speed interfaces) of a couple of seconds of error gets larger and larger as the poll cycle is set smaller and smaller.

The most accurate way to get SNMP bandutil statistics via SNMP is for the application to do a poll for sysuptime and hcifin and hcifout all together in the snmp get and then used the delta of the sysuptimes to calculate delta time.

I know only one vendor that time stamp delta time using polled sysuptime and not server receive time but most don't.

jp

Actions

This Discussion