[j-nsp] questions regarding submarine cables

[Mark Tinka]

On Monday, December 12, 2011 06:54:14 AM Martin T wrote:

> This isn't directly related with Juniper, but hopefully
> not totally off-topic as well :) At least here might be
> some industry insiders, who have some experience with
> submarine cables.

It's operational in a way that affects the Internet, so I'm 
happy :-).

> 1) According to engineer in this video:
> http://vimeo.com/29975179 ..modern submarine cables are
> often built as self-healing rings. For example "FLAG
> Atlantic (FA-1)" between USA and Europe seen here:
> http://www.cablemap.info/ is a good example. How does
> such self-healing ring works?

In essence, it's not that different from SDH or SONET rings 
built on land.

See here:

	http://en.wikipedia.org/wiki/EAC-C2C


However, it is also feasible that some cable operators 
consider options for building 2x or more linear paths, as 
opposed to ring topologies (ring topologies do have their 
own sets of challenges, e.g., choosing the best route, 
guarantees on latency, management, e.t.c.; which is not to 
say that linear topologies do not bring with them their own 
unique set of constraints - it's a balancing act).

> Am I correct, that in case
> both cables are healthy, only one of two is utilized
> with traffic?

Cable operators are unlike IP operators. They don't normally 
load balance traffic across all available physical media.

They could use the other part of the cable as a "Protect" 
circuit, i.e., if you buy a service from the cable operator 
and request for protection, the protection will be enabled 
on the other part of the ring.

Some operators run as many as 3x rings and designate that as 
"primary", "secondary" and "tertiary".

> Who provides equipment for such setups(I
> guess Alcate-Lucent is one of the vendors)?

Key submarine equipment (SLTE) vendors are:

	o ALU.
	o Fujistu.
	o NEC
	o Tyco.


Some of the terrestrial players are now looking to enter 
this space, e.g., Huawei, Infinera, Nortel, e.t.c.

> In addition,
> how long it will take such switch-over to occur? Are we
> talking about couple of milliseconds or few seconds?

Cable operators offer 50ms failover for pre-provisioned 
protect circuits. Typical electrical hand-offs to customers 
are SDH or SONET, although Ethernet is now commonplace. I've 
dealt with SDH and Ethernet only.

However, I know some cable operators are considering 
GMPLS/ASON which does provide protection, but may not be as 
quick as 50ms. GMPLS, for example, uses an IP control 
comprising:

	o OSPF-TE or ISIS-TE for routing.
	o RSVP-TE or CR-LDP for signaling.
	o LMP for link management.

Of course, what we normally do is rather than buy a service 
+ protect circuit from the same cable operator, we buy 2x 
service circuits from different operators so we can:

	a) Have both circuits active at the same time.
	b) Reduce risk by using different providers.

> 2) In case cable has multipoint landings(for example
> "GLO1", which has 15 landings in west coast of Africa
> and Europe) then submarine branching units should be in
> use. How this branching works? Are some waves(CWDM)
> inside the cables switched towards the landing and other
> continue via submarine cable? Because as I understand,
> on a wave level, we are still talking about point to
> point connections(?). In addition, are those branching
> units under the water or are they on small islands, oil
> platforms etc if possible?

That is what is called an OADM-BU. You may take a look here:

http://www.nec.co.jp/techrep/en/journal/g10/n01/100104-71.html

http://blog.pipeinternational.com/index.php?option=com_myblog&show=The-
OADM-BU.html&Itemid=53

Hope this helps.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20111212/b5214155/attachment.sig>