Q: In short, what is the future of data transmission?
I faith 0C 768 are rolled out, right? 41Gbps second?
I realize there are also systems in place, such as Internet 2 . any useful links?
Re:Thanks for the clarification guys. I thought it was switching and mux speed that limited the number of channels.
Generally as an Enterprise I don't care what a carrier does, just give me my interface fast, reliable and cheap. For example on the MAN side we just run POS on a telco sonet ring.
Re:There are some areas (Chicago being one of them) where the Telco offers a service called "GigaMAN" where companies can get point-to-point GigE with up to ~100Km (mile?) radius. Cost-wise, I believe it's at or below DS3.
FWIW
Scott
Re:guy, in the enterprise, expect to see an order of magnitude jump in Ethernets. 100Mb/s links -> gigabit, gigabit links -> ten gig. A fairly straightforward upgrade ratchet. What's also extremely interesting is the 10 gigE WAN PHY, which is still not widely available, but will allow 10gigE links to be built on top of SONET OC192c links. Why is that interesting? See guy's post about WDM, and see my next paragraph. SONET metro/carrier transmission gear is wicked expensive, and getting much of the same performance out of cheaper 10gigE switching equipment will help make high speed MANs more cost effective.
Metro area networks are a boom waiting to happen. Today's access network is terrible. We have 100Mb/s, gigabit, now 10gig LANs, but most of the business world lives on T1s, DS3s, and maybe if they're lucky OC3s. Want to interconnect a few offices? Unless you have a big budget, this is a frame net or an IPsec VPN over those same slowish uplink lines. There are realities of peering and cost structures today that make it still cost prohibitive to build very very high speed full net access, but a fast (100Mb/s, gigE) closed metro network that you can attach a few offices to and also get some traffic shaped full net access out of is still a major improvement from where we are now. Unfortunately, (a) the telcos are NOT helping cheap MAN deployment as it cuts directly into where they think they make money (if telcos had cheap MAN offerrings, they'd sell more of them, though, and have happier customers… *sigh*), and (b) the CLECs who have been in this business all were in the early part of their ramp-up when the bubble burst – most of them didn't survive and those that did had to retrench and worry about staying alive rather than growing. And unfortunately, in the access/meto market, footprint rules.
guy, to more narrowly address your question, Juniper has an OC-192c POS PIC. I don't know what Juniper and Cisco have as beta products and I don't know what random start-ups are trialing, but normally looking at Juniper's top end PIC will give you a pretty good idea of what the top end WAN line speeds in use are. Note that this is slightly deceptive because carriers often are using DWDM to mux many of those on a fiber, and optical switching or just plain patching in order to create cut-through circuits to different locations. So for example, a fiber from Washington, DC to Philadelphia, PA might well carry an OC-192c link between those two points, and also carry the first hop of a Washington, DC to New York, NY OC-192c link. The router thinks they're two separate fibers because that's what will come into the router, but the backhoe will demonstrate otherwise 
guy, in practice fiber and termination quality remains a limiting factor on much of this stuff.
Re:(nitpick mode)
The actual lamda separation is done optically with filters then the data for that wavelength is recovered, amplified, etc.
(end nitpick)
The limit is determined by the frequency of the medium (light, in ths case). The magnification in light microscopy is limited by the (relatively) low frequency (~1000-1300Xfor pure light microscopy). To get higher resolution, they moved to electron microscopy (several angstrom resolution is common), then X-rays. To break the upper limits for data transmission, they'll either have to find a medium that'll move modulated X-rays (per wavelength), or improve the resolution of thier optical filters (DWDM).
Actually these are two separate issues. DWDM just aggregates discrete signals onto a single fiber (pair if it's bi-directional). Better filters will improve the aggregation density. Medium that keeps the signals "purer" will also help.
To get mode data/bandwidth for a single signal, they need to improve the amount of data-per-clock (X data for Y clocks). … like "bits-per-baud" … they'll be needing some new compression or encoding schemes to get that done.
Watch for something that would equate to "Wireless DWDM" in the not-too-distant future. High-speed, high-density radio and/or media-free optical DWDM (point-to-point).
FWIW
Scott
Re:Originally posted by: guy
Currently, because of DWDM, single fibers have been able to transmit data at speeds up to 400Gb/s.
:Q…Is there a physical limit however?
No physical limit that I can think of, its just the precision of the electronics to separate all the wavelengths (lambdas) needed.
Re:Currently, because of DWDM, single fibers have been able to transmit data at speeds up to 400Gb/s.
:Q…Is there a physical limit however?
Re:Sweet. I guess I was reffering to carrier grade, but your info is very interesting.
I used to work at a place that did multiplexing. Neat stuff.
Re:Hmm. Anything at those kind of speeds isn't designed for the enterprise – That's definitely aimed at the carriers. I don't know anyone who needs that kind of bandwidth.
Keep in mind that large enterprises are often the slowest to change. They tend to stick with something that works, until they really need to change. (That's probably why we have 2,500+ OS/2 servers on our network!).
If I had to pick one technology that's been the biggest impact on the L1-3 network, it's definitely DWDM (http://www.webopedia.com/TERM/D/DWDM.html) – The ability to multiplex multiple connections across a single fiber. This has allowed companies to be far more flexible with their networks.
For example, you can lease a single pair of fiber between two data centers You can mux multiple Ethernet connections, your WAN circuits, the SAN replication, PBX, security system, etc. all together. No more need for ten pairs of fiber, just use one and some DWDM muxes. Need to add another SAN? No problem, spin up another lambda (aka virtual fiber) and you're good.
Lessee. What else. Voice over IP for a large-scale replacement of the enterprise voice system is starting to become a reality, MPLS is catching on more in the core, and more efficient delivery of HTTP – CDN's and HTTP Acceleration.
Lots of cool things out there that are just finally catching on in the enterprise.
- G
0 Comments.