Why is SNC switching Time <50ms
Arun Kumar Pillai
Need to understand why the SNC switching time is less than 50ms.
As i know in TDM if signal fail is more than 50 ms ,then phone
connections will be disconnected.
Is it right or any other reason is there?
pavan
According to the ITU-T standard the any protection
switching time should be less than 50ms.
pavan
The 50ms switch time is kind of the SDH/SONET standard for switching.
This includes fault detection and the actual switching and resynching
of the circuit. While it is 50ms for the transport gear, I can tell
you that it is usally longer for the circuit to recover as a whole.
It also depends on the type of equipment on the circuit. Ideally,
your using equipment that has at least a 50ms worth of data buffering,
then you won't notice a thing. The equipment will keep transmitting
what's in the buffer and by then, the circuit has switched and new
data will arrive in the buffer.
In worse case scenario's, it takes time just for the voice carriers to
reconnect and resume calls. If it's IP, then its going to take time
for the interfaces to re-establish connection etc.. So 50ms for
equipment doesn't mean 50ms for the circuit. A circuit can take
anywhere from 5 seconds to 20 seconds to recover. But again, this is
all dependant on the equipment.
UPSR (Unidirection Path Switched Rings) DO take switch hits, but not
as bad as 50ms. In this type of ring, the traffic is being
transmitted in both directions on the ring. So there is no electronic
switch from one card to another, like on a linear system. All that
happens is the APS bytes (K bytes that reside in SONET overhead) from
one node tell the distant node to start receiving traffic from the
other side of the ring. With equipment buffering, you'll barely
notice a 10ms hit.
Huub van Helvoort
You wrote:
> Need to understand why the SNC switching time is less than 50ms.
> As i know in TDM if signal fail is more than 50 ms ,then phone
> connections will be disconnected.
You are almost right.
The requirement for protection switching less than 50 ms is already
applicable to PDH (E1 and T1) before SDH inherited it.
When a call is started, signalling (dialling information) is used
for setting up the connection. These signals have a duration of
about 100 ms, an interuption of 50 ms will not affect this signalling.
Calls that have been established can be interrupted for more
than 50 ms (several seconds) without being disconneted.
Best regards, Huub.
--
reply to hhelvooort with 2 'o's
================================================================
http://www.van-helvoort.eu/
================================================================
Always remember that you are unique...just like everyone else...
Huub van Helvoort
You replied:
>>> Need to understand why the SNC switching time is less than 50ms.
>>> As i know in TDM if signal fail is more than 50 ms ,then phone
>>> connections will be disconnected.
>>
>>> Is it right or any other reason is there?
>>
>> According to the ITU-T standard the any protection
>> switching time should be less than 50ms.
>
> The 50ms switch time is kind of the SDH/SONET standard for switching.
And OTN, and Carrier Class Ethernet, and MPLS-TP as well.
> This includes fault detection and the actual switching and resynching
> of the circuit.
No! According to the recommendations it is the time *after* the
defect has been detected. It is the time for the switch to take place
and re-syncing after the protection switch is *not* included.
> While it is 50ms for the transport gear, I can tell
> you that it is usally longer for the circuit to recover as a whole.
See above.
> It also depends on the type of equipment on the circuit. Ideally,
> your using equipment that has at least a 50ms worth of data buffering,
> then you won't notice a thing.
No! Ideally the equipment should have 0 (zero) ms buffering to keep
propagation delays as low as possible.
In SDH the objective is transfer delay per node of 50 ns (nanoseconds).
> The equipment will keep transmitting
> what's in the buffer and by then, the circuit has switched and new
> data will arrive in the buffer.
This may work for voice, but if data is transported this is not useful.
> In worse case scenario's, it takes time just for the voice carriers to
> reconnect and resume calls.
Calls are never resumed, if the interruption it too long, calls will
be disconnected.
> If it's IP, then its going to take time for the interfaces to
> re-establish connection etc..
Maybe you are talking now about restoration which may take up to
several seconds.
> So 50ms for equipment doesn't mean 50ms for the circuit.
The requirement for <50 ms is for the connection, which can include
multiple nodes, so it is not for a single node/equipment.
> A circuit can take anywhere from 5 seconds to 20 seconds to recover.
This is again recovery, not protection.
> But again, this is all dependant on the equipment.
It depends on the technology.
> UPSR (Unidirection Path Switched Rings) DO take switch hits, but not
> as bad as 50ms. In this type of ring, the traffic is being
> transmitted in both directions on the ring.
Ring protection is link protection and linear protection is
connection (circuit) protection.
> So there is no electronic switch from one card to another, like on
> a linear system.
This is equipment protection, which is a special case of linear
protection.
> All that happens is the APS bytes (K bytes that reside in SONET
> overhead) from one node tell the distant node to start receiving
> traffic from the other side of the ring.
APS is used in both ring and linear protection to support the
in-band protection protocol.
The APS protocol resides in the connection function of a node
where the switching takes place. In linear protection at the
end nodes, in ring protection in evry node in the ring.
> With equipment buffering, you'll barely notice a 10ms hit.
Even without buffering, as explained above, you will not *hear*
a hit, but your computer will notice a hit.
Best regards, huub.
pavan
Thanks a lot, Hubb. You clarified so many things.
Thanks&Regards
Pavan