[Customer router] 1--->
[FTTC modem] 2--->
[FTTC MSAN] 3--->
[BT Layer 2 Switch] 4--->
[GEA Cablelink(s) into the...] 5--->
[21CN cloud via MSIL?] 6--->
[Host link(s) to CP Infrastructure] 7--->
[CP transit to Internet] 8--->1
CP Return Paths to FTTC modem (
presumably a reverse of the above but may jump onto CP owned backhaul at certain points to save moola?)
So where might congestion present?
1. The infamous 'customer domain' [Responsibility=Customer] a. Customer router/PC e.g. high processor usage = reduced throughput and erratic latency
b. Intermittent/pervasive copper faults can look a lot like congestion and should be investigated before anything else. (
A recent example is where a HG612 was running over 90% CPU, presumably correcting FECs)
c. Misconfigured PPPoE clients, possible blackholing of frames with MTU between 1492 and 1500? (
Mentioned in a BT SIN somewhere)
d. Loads of other stuff we can add later...
2. The TP/copper line/metallic path [Responsibility=Openreach] a. A copper line fault, disturber or interferer could fire up within the classic peak utilisation period; disruption akin to congestion in some cases, perhaps.
3. Uplinks from street cabinet MSAN to Exchange/Metro node [Responsibility=Openreach] a. MSAN uplinks to Exchange/Metro node oversubscribed? (
Found documentation to suggest the Huawei 5600 supports 10GE uplinks but no proof these are actually used; do they take up a shelf that could otherwise be used for subscriber ports? Also no indication 10GE is supported on *all* FTTC MSANs... ECI? Information for this aspect of the network appears to be extremely thin, hopefully I've just overlooked it somewhere?)
b. MSAN high processor usage contributing to erratic pings/reduced throughput. Software vectoring (
which I just made up) or any other intensive calculation could contribute to it; this is ultimately congestion.
c. MSAN not informing BRAS of correct synch speed, wrong policer applied... somewhere/everywhere?
4. Uplinks from L2S to CP handover/aggregator [Responsibility=CP & Openreach] a. GEA Cablelink is oversubcribed, CP needs to order more [Responsibility=CP]
b. FTTC traffic unevenly distributed over Cablelinks? Worst case... no port aggregation and EUs are staticly assigned per port [Responsibility=Openreach]
c. Some other scenario?
5. MSIL entry point at exchange/Metro node[Responsibility=CP] a. MSIL bursting above contracted bandwidth excess traffic policed either by BT on upper limit or CP on lower limit to avoid charges.
b. Congestion within the bowels of 21CN (surely not?!)
c. Others...?
6. MSIL exit point at core node / Host links to CP (gleaned from a comment on the PN forum) [Responsibility=CP/Wholesale] a. Poor traffic distribution across Host links back to CP [Responsibility=Wholesale]
b. Poor traffic distribution across Host links back into 21CN [Responsibility=CP]
c. Host link(s) generally congested/oversubcribed. Presumably 1 and 10Gb. The latter representing significant capex if not opex too? [Responsibility=CP]
7. CP transit to Internet [Responsibility=CP] a. The transit link
b. The xconnect to get to the transit link
8. Return paths to FTTC modemAs you can tell, I am clutching at straws for some of this but it seems like we may all benefit from understanding this stuff a bit better?
Please offer a correction if you spot an error on this and I will update.
I think the CP handover has to happen at a core node, for example? One of 20~ in the UK? Plusnet have around 20 gateways... seems to fit?
My assumptions:
---------------------
I have found it extremely difficult to understand the BT nomenclature. I have almost zero practical experience of service provider networks so perhaps it's just me.
The MSIL/AP/EP riddle:
MSIL = "Multi Service Interconnect Link" = EVC Trunk? (Metro ethernet forum:
An EVC is a conceptual service pipe within the service provider network.)
EP = "Extension Path" = Still not entirely sure. Possibly EFP/Ethernet Flow Point/Service Instance or some accounting instance
AP = "Aggregation Point" = Port+VRF on the BEA (
An Aggregation Point (AP). The AP is a logical CP-specific virtual router, in which the CP’s
traffic is managed.)
I've been wondering if AP bandwidth is synonomous with traffic conform rate and the EP is the burst rate? I seem to have made up a few scenarios where they both fit can't be sure of any!
SVLAN = this word has spread like wildfire across various forums but noone seems to know what it specifically applies to. In my mind, this is simply a double tagging scenario. Inner (c) tag is customer, outer tag (s) is service provider/CP. With that in mind, it could apply to almost anywhere in the network, FMSAN to L2S, L2S to CP or elsewhere?
SVLAN upgrades? How one upgrades the outer tag on a frame I'm not sure. Surely the actual upgrade is better allocation to trunk ports? Perhaps this is implied in SVLAN but it's not obvious to me.
General thoughts which may help/hinder the above:
------------------------------------------------------------------
CPs can only order in 1Gb/s chunks from the FTTC (L2S) terminating switch:
From http://www.sinet.bt.com/sinet/SINs/pdf/498v6p0_C.pdf
1.3 GEA Cablelink
The GEA Cablelink product will be offered for the CP to order connectivity to the
L2S in the same Point of Handover building.
This will comprise:
• A 1 Gbit/s Ethernet port into the L2S. The Gigabit Ethernet interface will be
set to auto-negotiate, 1000Base-LX (SingleMode only); and
• Fibre connection from the port on the L2S to the location within the same
Point of Handover specified by the CP.
I've no idea which of the two below Plusnet use. Gut says they relieve as much responsibility from BT as possible and terminate their own PPP sessions?
WBC can be run in two customer handover modes from the AP:
• PPP Termination and Aggregation (PTA), whereby the End User PPP Session is terminated
on the WBC BRAS and Routed IP is used to handover the End User traffic to the CP;
• Layer 2 Tunnelling Protocol (L2TP) handover, whereby the End User PPP session is
tunnelled to a CP-owned L2TP Network Server (LNS) where the PPP session is terminated.