BladeSystem Virtual Connect
Showing results for 
Search instead for 
Did you mean: 

Can Flex10 NICs use more bandwidth than that has been assigned to them?

Trusted Contributor

Can Flex10 NICs use more bandwidth than that has been assigned to them?

There has been a lot of discussion regarding how Virtual COnnect only does rate-limiting in the Transmit side of Virtual Connect. Here is some of the conversation and what Alex in our engineering group helps to clarify things:






On the recent Flexible adapter / infrastructure concall (last week).  The point was made that bandwidth is only enforced on outbound connections, from the blade to the Virtual Connect module.  There is no capping/shapping/limiting on the inbound.


So theoretically you could get 10G inbound on one flex port, even if it was set to 100Mb.


Can remember what they said about the time line and over subscribing flex nic's - ie being able to allocate a min and a max throughput.




Gary commented: 

Is this still true?  So if there is a DOS attack against one of the FlexNICs, will it be limited to that NIC or will the entire LOM be crippled resulting in loss of comms on the other three flex NICs on the LOM?




And this was Alex response to this issue:




Individual Ethernet, iSCSI or FCoE functions Rx traffic flows are not limited and could consume up to full line rate of 10Gb. With FCoE, however, ETS flow control management will guarantee minimum bandwidth set by VCM. What this means is that when there is no congestion the FCoE or LAN bandwidth can exceed the specified data rates for the traffic flowing from VC to FlexFabric adapter. But, under congested conditions, the VC module will enforce a fair allocation of bandwidth as determined by the FCoE function rate limit defined in the server profile and the remainder will be set as the aggregate rate limit for the FlexNICs.


On the Tx side, FlexNIC will be limited by whatever is defined in the server profile and set as max in the network definition.




Other questions or comments?