Forum Discussion
Tal_BenHaim_112
Nimbostratus
NimbostratusSNMP - Collecting data on throughput rates
Hi,
I wrote an application that makes a report of traffic throughput rate on each
virtual address, and the interface using i-control.
The results i get are much low than i expected.
I would like to get these statistics using SNMP.
where can i find instructions on how to install the net-snmp for windows?
I read the article on how to configure the big-ip to work with SNMP.
(https://tech.f5.com/home/bigip-next/manuals/bigip9_2_3/bigip9_2_3mgmt/BIG_IP9_2_3_nsm_sysguide-16-1.htmlwp1024696)
As I understand I should use the sysStatClientBytesIn, sysStatClientBytesOut ... objects.
What is the difference between those object?
what is the difference between those objects and STATISTIC_SERVER_SIDE_BYTES_IN/OUT (i-control)?
Thanks,
Tal
- Tal, I can't really help you on the SNMP side of things (mainly because I just don't know that part of the product). As for your issues with statistics in iControl, are you sure you are correctly building the 64 bit numbers from the UInt64 structures returned in the stats calls? If you are just using the low 32 bit number in that structure, then the results will be unpredictible when the counters flip the 32 bit boundary.
-Joe - Hi Tal -
I might be able to shed a little light on the SNMP stats definitions:
Client bytes in/out refer to data on the client side of LTM, where "in" means data sent from the client to the LTM, and "out" means data sent from LTM to the client.
Server bytes in/out refer to data on the server side of LTM, where "in" means data sent from the server to the LTM, and "out" means data sent from LTM to the server.
I'm sure all 4 are exposed in iControl, although I am not sure how exactly they are named or categorized for iControl.
(Ever hear that story about 4 blind guys trying to explain what an elephant looks like...?)
HTH
/deb - (edited for clarity)
Tal_BenHaim_112Hi,
Im using this function:
UInt64 build64(CommonULong64 value)
{
return ((ulong)value.high << 32) | (uint)value.low;
}
If I would multiple the results I get from i-Control by 5 than the results would make sense....
(excepy of that they seem to be ok , the proportion of throughput between the Virtual Servers and Pools seems to make sense , and the interface stats is quite the same as the sum of the virtual servers.
The only thing is that the results are 5 times smaller...
Do you have any idea?- Your conversion code looks good. The data that iControl returns is the exact data (ie, comes from the same place) that is presented in the GUI. Are you saying that the counters presented in the GUI are 5 times less than the counters returned via iControl? Or you are referring to rates that you are calculating on the client code. If it's a question of rates, where are you getting the 5x rates that you are comparing it to?
-Joe - Tal_BenHaim_112Hi,
I was referring to rates that Im calculating in my client code (using i-Control).
These values are 5 times less than the values I get from the throughput rates I receive from ISP provider.
I had a look on the graphs in the GUI (Overview ---> Performance ---> Throughput)
Could you provide us with more information of what we actually see in the graphs. also, how come we see a rate 60MBPs on the graph while the maximum capacity of our telecom lines is 30MBPs ? 
John_Arnold_106Link to Manual:
https://tech.f5.com/home/bigip-next/manuals/bigip9_4/bigip9_4mgmt/BIG_IP_9_4_nsm_guide-16-1.htmlwp1053731
The two throughput lines in the GUI graph represent Client-side bits and Server-side bits. The lines are an aggregate of TX and RX.
If your telecom limit is 30Mb/s Full Duplex (that's a weird rate...) your servers could theoretically send 30Mb/s and receive 30Mb/s at the same time, and BIGIP would report the (RX+TX) aggregate of 60Mb/s.
