The Motley Fool Discussion Boards
Speaker's Corner / FrozenCanuck's FiberHead Friends
|Subject: Re: Metro Products||Date: 3/3/2002 7:37 AM|
|Author: DirtyDingus||Number: 2587 of 2606|
Here is a summary post I did 2 years ago on wireless and optical players.
Its still reasonably current - as is/are the posts I did on the definitions of a switch vs a router etc.
Another post about metro network topologies is at http://boards.fool.com/Message.asp?mid=14858051
Slightly more recently I wrote an article for the RTW newsletters produced by MAtt Tomkins (Y2Krash) - its in the August 2001 issue available at http://y42.briefcase.yahoo.com/rtwnewsletter/ - about the challenges of metro deployment.
With that as background here is some answers to your questions. Let me start by saying that until Sept 01 I worked for Nortel as a product manager for the Passport 8600 which has quite a lot of featuress that make it applicable to the metro space so I was up until that point depply involved in the metro market. Since then I've changed tack rather dramatically (see profile if U care) so I am no longer plugged into the currents of the switch market.
I'm trying to identify: 1.] what the primary "metro products are? 2.] What are the companies producing the products?? 3.] Does it seem sensible to begin studying all the identifiable "metro" companies, or are there 2 or 3 that stand far enough better than others that it would be reasonable to narrow study to these 3:
See links above. Metro products are generally cheap SONET/SDH ADM's, (gigabit) ethernet l2 and L3 switches and relatively coarse WDM systems and combinations of the above.
Sorrento make switches I thought...now I find in their advertising they call it a "switch router." Would that be on layer 3?? They list smart-looking group of customers, but they show low income [< 50M/yr ...which is about their mark cap].
Well actually no. If you look at the above mentioned definition of a switch link and the sorrento product page (http://www.sorrentonet.com/products_11.asp) you see that Sorrento have a dynamic optical cross-connect. This is a layer 1 device as it is unaware of the protocols and packets passing on each wavelength.
I understood that the ONIS product also is a "switch", and I am trying to learn more about all this because the CIEN ceo said the ONI metro product would make CIEN more competitive in that space.
ONI's products seem to be somewhat more smart than Sorrento's but that is because they appear to (optionally at least) do an optical-electrical-optical conversion. Of course during the electric phase ONI can do clever things like packet-baed routing, SONET/SDH stream aggregation (i.e. add/drop multiplexing) and so on. They do not yet apparently do packet based routing although there is mention of an MPLS product and it looks like that sometimes they can be a pure optical device BUT compared to sorrento they are less pure optical and more complex and thus further up the OSI stack but still mostly no further up than Layer 1.1
What is the issue about metro space that matters whether the switching process occurs at layer 3 or layer 3?
Well its a case of scalability and flexibility. If you build a L1 network (as ONI/Sorento etc want you) have a very scalable system (in that you can build many endpoints and many fiber rings/meshes) but its extremely inflexible and hard to change what traffic goes where once its been set up. This is a problem because in the metro about the only constant is change. Also (and related) this kind of network is hard to engineer to be robust towards network outages such as equipment failureor fiber cuts without reserving two paths for every connection.
On the other hand if you build a layer 2 network using a packet protocol such as ethernet you have a lot of flexibility because every device can potentially see every other one at the expense of scalability - managing the resulting broadcast storms is a nightmare - and the number of VLANs (vitual networks) is limited to some fairly small number (generally about 4000 but with proprietary extensions can be more than that). An Ethernet L2 network gives you lots of flexibility and reasonably network recovery arounbf aoutages etc. but at a cost of significant scaling problems.
An L2 network using SONET/SDH looks just like the all optical L1 network with the disadvantage that available bandwidth is a lot less and the advantage that the flexibility in terms of changing where traffic flows is much improved. However although the network is flexible in terms of provisioning it is not so flexible in terms of coping with network outages as the packet based network would be - as with the L1 network the usual solution is to waste 50% of available bandwidth by carrying the every signal over two different paths.
An L3 network is forced to be a packet based network. An L3 network will generally combine with L2 or L1 networks in that it will use the L1/L2 bits for the initial access layers and the interconnects between the L3 nodes, however the L3 devices solve much of the redundancy/flexability/scalability issues in the pure L1/L2 designs simply because they reduce the required size and complexity of each L1/L2 network. The disadvantage of L3 is that it is further removed from the wire and thus the advantage gained by saving network bandwidth has to be paid off against as slower response tim ein the event of an outage (typically 30+ seconds rather than the 50ms response time of optical or L2 SONET networks). On the other hand the L3 network significantly eases provisioning and with