04-03-2013 11:57 PM
for configuring an ISL between two switches, I understand, that B2B credits are important and deciding factor of the distance and speed between the two switches.
L0 is the normal mode and supports upto 5 KMs, and LE is the mode that supports from 5 to 10 KMs.
Also, each port has been provided with a default of 20 buffer credits.
I have a confusion here in understanding the difference between L0 and LE, or rather when to use it.
It says in the FOS guide that, in L0 mode, a port operating at 4 Gbps can be ISL-ed to a distance of 2 KMs, and upto 5 KMs at 2 Gbps.
Putting a port in LE mode will enable it to obtain more buffers.
However, the buffer calculation indicates that with 20 Buffers, an ISL can go upto 5 KMs at 4 Gbps.
So, why would we need to put that port in LE mode if we want it to work at 4 Gbps for a distance of 5 KMs? It already has sufficient buffers that it needs for operation.
Would it not work at L0 mode?
I am sorry if I have not been able to frame the question better, but hopefully I have put my query at least for the experts to understand.
04-04-2013 10:39 AM
Yes there is overlap in buffer credits allocation with these two port modes. The L0 mode is the default for all switch ports. Actual buffer credits are assigned depending on the type of device (F or E Port) that logs into the switch when in this mode. The LE mode is use to statically configure additional buffer ports for distance. I always validate BB credits required using the formula found in the FOS admin guide.
Use one of the following formulas to calculate the reserved buffers for distance:
• If QoS is enabled:
(Reserved Buffer for Distance Y) = (X * LinkSpeed / 2) + 6 + 14
• If QoS is not enabled:
(Reserved Buffer for Distance Y) = (X * LinkSpeed / 2) + 6
X = the distance determined in step 1 (in kilometers).
LinkSpeed = the speed of the link determined in step 2.
6 = the number of buffer credits reserved for Fabric Services, Multicast, and Broadcast
traffic. This is a static number.
14=the number of buffer credits reserved for QoS. This is a static number.
Alternatively you can calculate BB credits based on frame size (if it's a known quantity):
desired_distance = roundup
04-04-2013 10:04 PM
Thank you for your response jdaignea. It makes things a little clearer, but my confusion still remains.
I will try and put my query across with sepecific examples.
If I want the following ISLs to be set up, which mode should I be selecting? I have used the buffer calcuation without QoS and as per the formula you had pointed out earlier.
Buffer required: 11. Default buffers in port is 20 in L0 mode.
So L0 mode would be sufficient. But, the admin guide also says to select L1 for distances between 5 - 10 KM.
the default 20 buffers is good enough to run, AND .
So, which mode should we be selecting here?
04-05-2013 02:24 AM
In response to you questions:
Mode L0 provides 26 BB credits (20 actually for data frames). There are more Data Frames that the theoretically required due to the distance because these 20 BB credits are shared out among 4 Virtual Channels.
Mode LE, may provide fewer BB credits, (16 BB @ 2Gb, 10 BB for Data Frames though), but this 10 BB credits are assigned to an only data Virtual Channel.
I suppose that using mode L0 in long distance links, could cause a BB credit starvation situation in one of the VC
as distance increases (for example, a TrueCopy replication will always use the same VC inside an ISL). There could be a situation in which there is a starvation of BB credits in a VC whilst the rest of the VC are underused. I reckon that, for long distances, having more BB credits shared among different VC may not be the best configuration. Additionally, if Brocade recommends to use Mode LE, I would take his advice.
I hope this helps.