For more details, please see ourCookie Policy.

Anonymous
Posts: 0

# Bottleneck measure

Hi All

I need some help in understanding bottlenck better.

------------------------------------------------------------------------------

Latency threshold for alert - 0.100
Congestion threshold for alert - 0.800
Averaging time for alert - 300 seconds
Quiet time for alert - 300 seconds

--------------------------------------------------------------------------

Can some one explain me the above output.

Valued Contributor
Posts: 761
Registered: ‎06-11-2010

## Re: Bottleneck measure

Hi,

That's easy. The definition of each parameter is the following:

Congestion threshold for alert: The severity threshold for triggering a congestion alert. This threshold indicates the percentage of one-second intervals affected by congestion conditions within a specified time window. The congestion threshold is expressed as a fraction between 0 and 1.

Latency threshold for alert: The severity threshold for triggering a latency alert. This threshold indicates the percentage of one-second intervals affected by latency conditions within a specified time window. The latency threshold is expressed as a fraction between 0 and 1.

Averaging time for alert: The time window in seconds over which the percentage of seconds affected by bottleneck conditions is computed and compared with the threshold. Quiet time for alert The minimum number of seconds between consecutive alerts. The value assigned to this parameter applies to both latency and congestion monitoring.

Quiet time for alert: The minimum number of seconds between consecutive alerts. The value assigned to this parameter applies to both latency and congestion monitoring.

Applying this definition to you bottleneck configuration: the bottlenecks are measured in 300-seconds windows, if in that period of time, the switch detects a Congestion situation the 80% of the time, it will trigger a Congestion bottleneck alert.

If the switch detects a latency during the 10% (or more) of those 300 seconds, a latency bottleneck alert is triggered.

Also, if an alert is triggered, it will not report any new alerts in the following 300 seconds as specified with "quiet time for alert".

Rgds

Felipon

Anonymous
Posts: 0

## Re: Bottleneck measure

Okay that was a pretty good explanation.

Now Let us take a scenarion where we have 8Gbps SFP,

Can you pls tell me, Theoretically and practically how much data can be transferred through it.

Valued Contributor
Posts: 761
Registered: ‎06-11-2010

## Re: Bottleneck measure

Hi there,

Let's try to figure out what the best we can get is:

• When working a 8G FC, the line rate is 8500 Mbps (8.5 Gbps) and the encoding is 8b/10b
• A full Frame payload can be up to 2112 Bytes, but in practice a payload of 2048 Bytes is used.
• The overhead (SoF + header + CRC + EoF) introduced by FC is 36 Bytes.

So, the size of a FC frame would be: Payload + Overhead = 2048 + 36 = 2084 Bytes

• Additionally, accoring to the FC standar, At the N_Port transmitter there shall be a minimum of six Primitive Signals (Idles and R_RDY) between frames. Each Primitive is 4 Bytes.

So every time we send a FC frame we also have to transmit 6 4-Byte primitives with a total size of 24 Bytes.

This makes that for every FC frame we send a total of 2108 Bytes are transmitted: Payload + overhead + Primitive signals = 2048 + 36 + 24 = 2108 Bytes

A payload of 2048 Bytes corresponds to 97,15% aprox. of the total number of Transmitted Bytes (2038*100/2108)

Once said this, we can try to calculate the real available throughput:

1. We lose 20% of the BW due to encoding, so 8500 Mbps * [8/10] = 6800 Mbps of FC traffic.
2. We also lose, in a best-case scenario with a payload of 2048 Bytes, a 2.85% of the Bandwith with overhead (plus primitives)

Total throughput: 8500 Mbps * [8/10] * [97,15/100] = 6606 Mbps / 8 = 825,78 MB/s

Hope its clear enough.

Felipon

### Join the Broadcom Support Community

Get quick and easy access to valuable resources across the Broadcom Community Network.