Talk:Optical Carrier

From Wikipedia, the free encyclopedia

Contents 1 Unsupported claims 2 Cost? 3 CRS-1 for enterprise? 4 Consolidation of Optical Carrier pages

[edit] Unsupported claims

The article previously contained a number of claims which I feel confident are inaccurate, specifically that OC768 was already replacing OC192 in large scale networks, and that 10GE is more popular in 'carrier grade' networks than OC192. Direct experience tells me that both of these claims are inaccurate enough to be laughable, and I can find no citations to back them up... So I've removed them.

The article also seemed to imply that 10GE is used over OC192 for internet traffic links because 10GE provides more bandwidth. Even under best case conditions for internet traffic on 10GE (1500 byte packets), 10GE only provides 2.36% more bandwidth. SONET's troubleshooting tools (alarms, and link trace, BER measurment) and OAM (which is not yet standard for ethernet) would justify the selection of SONET over ethernet against a 2.36% gain in any carrier enviroment. Worse, common internet traffic has a mean packet size around 300 bytes (and it's on its way down due to VoIP and other real time services), and at 350bytes/packet OC192 carries about 5% more capacity than 10GE. --Gmaxwell 19:53, 18 October 2005 (UTC)

I agree with you Gmaxwell, and i agree with your edits. However, i feel the sentence, as it currently is, is totally confusing to the reader. The sentence i mean is, "Although OC-192 operates at a somewhat slower line-rate than 10GE, it provides greater capacity for internet sized packets because Ethernet has higher per-packet overhead." Ommitted are: line rate of 10GE, and the definition of an "internet sized packet'. I think it would be great if you converted what you wrote above into information in the article. Remember, right now, we don't have a "too much information" problem, we have lack of information problem, so feel free to add as much as you can. — Fudoreaper 06:41, 28 October 2005 (UTC)

Gmaxwell asked me to comment on my changes with regard to comparing 10 GbE to OC-192. Specificly, I deleted "Although OC-192 operates at a somewhat slower line-rate than 10GE, it provides greater capacity for internet sized packets because Ethernet has higher per-packet overhead." because I felt it wasn't accurate since OC-192 does not always provide greater capacity, even with the qualifer about "internet sized", a term which isn't likely to be well understood by most audiences. I work in the industry, and I can't even be sure I understand what "internet sized" means. I think I have a good guess, but a layman would have no chance. Back to the issue at hand: per-packet overhead depends upon packet size (as discussed above), just like data inflation with HDLC depends on data content. Either the discussion needs to go fairly indepth to explain the comparison, or, in my opinion, it shouldn't be made since it is not important to explaining what OC-192 is. Mentioning it in passing simply invites misunderstanding by a large part of the audience. But this is in the past, working with what the article has now -- I see Gmaxwell updated it, and I like the improved wording about inter-operating. I question, however, if we really need the diversion about 10GE-LAN. Would it not be enough to just say OC-192 is compatible with 10GE-WAN and leave the LAN vs. WAN discussion in the 10GE article? I've thought quite a lot about what else could be added to flesh out this article - and honestly, I don't think there is much. Most of the info should be handled by the SONET article - almost to the point where I find myself wondering if the OC-n articles shouldn't just redirect to SONET (or be combined into the Optical_Carrier page). Mrand 05:33, 17 March 2006 (UTC)

We should probably have an entire article on link encap efficiency, you're right that the text is somewhat misleading if we over simplify. I just don't want to go do the converse route of misleading people to believe that 10GE always has higher capacity. :)

As far as LAN vs WAN... It *pains* me to mention 10GE Wan PHY without some clarification.. I don't have hard data in front of me, but I'd still be willing to bet that most 10GE wide area links are currently LANphy over wave.. The name is somewhat misleading.

As far as merging and redirecting, I'm all for it.. The proliferation of articles on speculative SONET speeds has been making me sad for a while. --Gmaxwell 05:42, 17 March 2006 (UTC)

[edit] Cost?

How much do OC-192 lines cost?

How much does the a fancy house cost?  :) Your question is a bit too broad to answer, if you're talking about a OC-192 link from point A to B we need to know where A and B are. If you want to know what an OC-192 worth of internet connectivity costs, that too is hard to answer because, like many other high end things, such purchases are negotiated on a case by case basis. --Gmaxwell 03:24, 26 February 2006 (UTC)

OK, but what ballpark are we talking about?

[edit] CRS-1 for enterprise?

that's a new one for me. Last time I looked, the CRS-1 was solidly aimed at the ISP market, not enterprise. --Alvestrand 18:38, 23 February 2006 (UTC)

You're quite right. Sound like someone is taking his CCNA a bit too seriously. --72.165.204.2 07:05, 25 February 2006 (UTC)

although I think the poster might have been correct if he'd replaced "enterprise" with "ISP" - OC-768 (40G) is indeed being rolled out in major backbones, and support of 40G interfaces is a design criterion for any serious new high-end router today. --Alvestrand 08:43, 25 February 2006 (UTC)

Whos backbone? ... The pre-existing optical transport gear can't support it without special (and insanely expensive) coding transponders like those made by StrataLight and even then it's difficult to get an existing 10g wave carrying OC-768. Worse, the signal produced by such specialized transponders is non-standardized and non-interoperable. At the current time it's much cheaper to run four DWDMed OC-192s than a OC-768. There has been a lot of testing and some trials, but I don't believe that anyone is actually rolling it out.

Now, as far as 40G interfaces as a design criteria for a router.. If you're talking about a router positioned at core provider networks, then I'd agree. But there are only two vendors who really have market share in that space: Juniper Networks and Cisco Systems.... and both of them offer OC-768 interfaces on their highest end boxes. So the point is somewhat moot. --72.165.204.2 16:53, 25 February 2006 (UTC)

[edit] Consolidation of Optical Carrier pages

Hi guys, I've consolidated all the OC-<whatever> pages to Optical Carrier, and the Talk pages of OC-192 and OC-768 --NigelJ ^talk 23:12, 28 April 2006 (UTC)