- Integrated Systems
- About Us
- Integrated Systems
- About Us
09-06-2013 06:34 AM
What is the difference between the IfInDiscards and IpInDiscards counters in VC?
Lawrence was looking to help a customer:
I am trying to get my head around the differences between IfInDiscards and IpInDiscards. From page 241 of the VC 4.01 user manual (http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c03791917/c03791917.pdf ) the definitions of IfInDiscards and IpInDiscards sound almost the same to a “non-network”
IfInDiscards - The number of inbound packets that were chosen to be discarded, even though no errors had been detected, to prevent their being delivered to a higher-layer protocol. One possible reason for discarding such a packet is to free up buffer space.
IpInDiscards - The number of input IP datagrams for which no problems were encountered to prevent their continued processing, but which were discarded (possibly for lack of buffer space). This counter does not include any datagrams discarded while awaiting re-assembly.
What is the difference between a packet and an IP datagram?
My customer finds that the IfInDiscards number and the IpInDiscards number are fairly static, but when they do vMotions of guests from one host to another, the IfInDiscards number starts to grow rapidly. They are asking why.
What scenarios cause IfInDiscards and IpInDiscards counters to start increasing?
My customer is asking whether monitoring IfInDiscards and IpInDiscards counters is a good way of monitoring the health of the VC modules and the traffic bandwidth. How are others monitoring whether their VC modules have sufficient bandwidth?
Reply from Armand:
Packet is usually an Ethernet frame, which include Ethernet header (src and dest address field, type field, VLAN field if tagged), **AND** IP datagram if IP traffic is carried **OR** FC frame is FCoE traffic us carried, **OR** any other type of traffic (designated in type field before), it’s mostly management traffic (LLDP, BPDU….). At this point, we are at L2 level.
An IP datagram is the IP payload in the Packet, it contains src and dest ip address, sequence number, …, and data : for instance, block of SCSI data if iSCSI is used. At this point, we are at L3 level.
Concerning Vmotion, a specific subnet is used for carrying that traffic, and at L2 level, would stay within VC domain (not forwarded to upstream switch).
Therefore, VC would consider this Vmotion traffic discarded at the uplink level, from where you’re probably monitoring the VC traffic.
Follow-up question from Lawrence:
Thanks for the explanation, Armand.
As a follow on question – what statistics should a customer be looking at, in order to see whether their VC bandwidth is coping with the traffic from their servers to their uplinks? My customer wants to know whether their uplinks (and stacking links) are being saturated or not by the traffic generated from their servers.
Input from Vincent:
Show statistics-throughput from the CLI or its GUI equivalent with a graph have been added for that purpose.
And info from Mark:
We also published this CN to describe behaviors you may see in regards to discards.
Other comments or questions?