10-30-2014 06:09 AM - edited 03-07-2019 09:18 PM
Folks:
I am looking for a work around - Currently I have a four member 2960x switch stack - I have Ten 1/0/1 and Ten 3/0/1 in a port-channel; however, when I try to set the 'bandwidth' on PO1 interface to 20000000 I receive an error, which corresponds being out of range, due to the links being 10Gbps.
Is there a work around so my PO1 interface reflects the correct bandwidth?
Switch I am using and version of software
Switch Ports Model SW Version SW Image
------ ----- ----- ---------- ----------
* 1 52 WS-C2960X-48TD-L 15.0(2)EX5 C2960X-UNIVERSALK9-M
Thank you
JJ
Solved! Go to Solution.
10-30-2014 08:57 AM
Hi,
The speed of each link is 10Gig but that does not mean you have a total of 20Gig of bandwidth to use.
Etherchannel load balances traffic between the 2 10Gig links using mostly source MAC address.
So, when host-1 sends a packet it uses one of the 10Gig links and when host-2 sends a packet it uses the other 10gig link, but you can't add the 2 10gig to get 20gig. It is not additional bandwidth, it is load balancing.
HTH
10-30-2014 08:57 AM
Hi,
The speed of each link is 10Gig but that does not mean you have a total of 20Gig of bandwidth to use.
Etherchannel load balances traffic between the 2 10Gig links using mostly source MAC address.
So, when host-1 sends a packet it uses one of the 10Gig links and when host-2 sends a packet it uses the other 10gig link, but you can't add the 2 10gig to get 20gig. It is not additional bandwidth, it is load balancing.
HTH
10-31-2014 09:20 AM
HI Reza,
Maybe I am looking at this the wrong. I understand the etherchannel group/interface is load balanced; however, wouldn't I still get 20Gbps in the aggregated PO1 interface since two 10Gbps links are bundled together? I am looking at this from the reliability 255/255, txload 1/255, rxload 1/255 perspective.
Since the 'bandwidth' is not configured correctly, the rxload and txload will reflect wrong values.
10-31-2014 09:53 AM
Hi Jason,
wouldn't I still get 20Gbps in the aggregated PO1 interface since two 10Gbps links are bundled together?
No, these are 2 10Gig interfaces and each flow uses one link at a time. Another word, the maximum bandwidth you can get is 10Gig. If you were to transfer a 20Gig file using one flow it would have to fit in one of the 10Gig ports. Portchannel does not increase the bandwidth, it just provides more lanes for you to use.
HTH
11-03-2014 05:33 AM
Hi Reza,
I was using poor choice of words - you following post "Portchannel does not increase the bandwidth, it just provides more lanes for you to use" - is how I understand it, so thank you for your professionalism and patience on explaining this.
Thank you for your help -
11-03-2014 05:46 AM
Hi Reza,
one more question about "etherchannels" - can I think of 2x 10Gbps links bundled together as the following?
20Gbps throughput, but only 10Gbps of bandwidth?
I am looking at it more from the STP issue that 'pille1234' - and STP will not see the aggregated links on the 20Gbps etherchannel.
11-01-2014 12:54 PM
The tx-/rxload is probably the least of your problems. Assuming we're talking about a L2 port-channel the interface BW is base for spanning tree cost calculation. With a BW 10000000 kbit/s STP would consider the port-channel equal to a single 10GE link and could possibly decide to block your port-channel for a less favorable link.
That being said, the port-channel is supposed to have the correct BW corresponding to the number of bundled links without any need for manual user configuration.
I suppose it is either a software bug or there is an interface not bundled correctly.
Post the result of 'show etherc sum' here to check.
Regards
11-03-2014 05:37 AM
Hi Pille1234,
I think I am hitting a bug. The port channels are configured correctly, which I can verify via 'sh etherchannel summary' and 'cdp neighbors detail' ; however, my Po1 interface is only reading 10Gbps.
Now, the odd part about it - my 6kvss shows the correct bandwidth 20Gbps on the other side of the portchannel.
I am going to dig through and see if there is an IOS bug... and will let you folks know.
10-31-2014 11:40 AM
If the port-channel is defined and the ports are up and aggregated correctly the bandwidth under the port-channel SVI should already read 20000000 bps . That being said everything Reza indicated about a given flow is true , it will have a total bandwidth of 10 gig to use as each flow is hashed over either one port or the other but not both at the same time.
11-04-2014 03:43 AM
This is a sh int po1 from one of our 20GB port channels.
Port-channel1 is up, line protocol is up (connected)
Hardware is EtherChannel, address is a80c.0dff.9f1d (bia a80c.0dff.9f1d)
Description: DOWNLINK TO *****
MTU 1500 bytes, BW 20000000 Kbit/sec, DLY 10 usec,
reliability 255/255, txload 1/255, rxload 1/255
Encapsulation ARPA, loopback not set
Keepalive set (10 sec)
Full-duplex, 10Gb/s, link type is auto, media type is unknown
It shows 20GB of available bandwidth but as each port is connected at 10Gb, this is shown also.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide