Are there any best common practices for the CFM levels use
Since my pure Ethernet aggregation layers are small I believe I only need
two CFM levels
I plan on using Level 5 between CPEs managed by us and Level 4 between
Aggregation devices -that's where MPLS PWs kicks in
So leaving Level 7 and Level 6 for customers and carrier-customers
respectfully -would this be enough please?
I'm also interested on what's the rule of thumb for CCMs Frequency, Number
of Packets, Interpacket Interval, Packet Size and Lifetime for the
Thanks a lot for any inputs
The ITU recommend the following levels:
5,6,7 = Customer
3,4 = Provider
1,2 = Operator
0 = Local segment
I don't know if there are any rules of thumb for the CCM interval - faster is more sensitive & unstable, slow is sluggish but stable. The spec allows between 3.33 ms and 10 minutes in 7 steps, with 1s being the midpoint. So we use 1s intervals.
I'm not sure if the other parameters you mention are configurable for CCM. I think the packet has a constant size. Are you wanting to also do Y.1731 performance management?
Thank you so much Jonathon.
This is exactly what I what I was searching for.
Oh and yes I should have mentioned I'd like to do the Y.1731 and measure the
delay and delay variance
Just yesterday evening I found a great article about how ATT did theirs
active measurements though for IP - wondering if I could do the same for my
They used dedicated servers and I'll be running this form the routers
ME3600X and CX and ASR9K so I'm a bit worried about the scaling of the whole
ATT basically used two probes and each 24-hour day is divided into 96 test
cycles of 15 minutes
A Poisson probe sequence of duration equal to the test cycle
- Poisson distribution with average interarrival time of 3.3 s
- Packet size of 278 bytes, including headers
- UDP protocol
Two periodic probe sequences in every test
- Interval of 20 ms between successive packets (or 50 packets/s)
- 1 min duration
- Random start time within the 15 min cycle
- Packet size of 60 bytes (including headers)
- UDP protocol
I don't know if any Y.1731 gear will do anything other than constant interval probes. I'm also not sure what the values of randomly spaced probes would be.
As far as I know the Y.1731 performance measurement probes are intended to be used to obtain performance data on circuits with as little impact on the available bandwidth as possible. Using a random departure time sounds like the sort of load test where you want to generate a synthetic data stream and consume bandwidth.
I have used BRIX active measurement for IP for many years, but here is a
link that describes BRIX in conjunction with ADVA for Ethernet probes.
There is an article in IEEE Communications Magazine circa 2004-2005 by AT&T
researchers describing their roll your own active measurement system,
theoretical assumptions, and theory of probe data collection.