WAN Ethernet

It depends on what you mean by WAN.

All the fiber ethernet types will run half duplex, but that still requires two fibers. With the right couplers it might be possible to get both directions into one fiber.

The traditional WAN, or, more correctly MAN (metropolitan area network) was 10broad36, pretty much gone by now. The current cable modems, I believe, should be part of the broadband ethernet family, though they don't have 10broadXXX labels.

Otherwise it is usual to go to T1 or T3, which are not ethernet but are more suitable for longer distances.

T1 can be used either point to point or through telephone company switching systems, point to point for metropolitian distances, through the phone system for longer distances.

-- glen

Yes. The primary constraint is distance.

Yes, but you would probably find full duplex 100Base-FX to be easier, less expensive, and less latency.

10Base-FB does not sound very promising to me, not unless you already have the equipment in place and plenty of spares.

I would not bother putting in 10Base-FL into any new locations unless that was all you could get your network service provider to support. The same multimode fibre can be used for 100Base-SX which is probably easier and less expensive to get parts for.

Around here, when they install a fibre link, even if all the customer purchases is 3 megabits/s, they put in 100Base, in case the customer wants to upgrade later (the fibre installation is the expensive part.) The ISPs then either put rate limiting on the line to prevent more than the contracted use, or else they just let the customer use more than the base megabits/s... and then bill the customer according to the

90th percentile actual usage.

Bonjour Glen,

Thanks for these words that make us thinking.

Yes, two fibers at the physical level but FD at the MAC level. As for any fiber ethernet type, you have either FD mode or HD mode. You can see that, for example, in section 30.5.1.1.2 (802.3-2005), where it is noted both 100Base-TXHD and 100Base-TXFD.

It's exactly the same thing for 100Base-T2, which can be HD or FD. On the transceiver interface, you have one dedicated pair to transmit data and one dedicated pair to receive data.

It's an interresting confusion.

Have you a typical reference, please ?

Today, many providers are using Ethernet in their long range networks, especially for 1G and 10G. My question was: are all the fiber Ethernet types could be used in these WAN networks?

And Walter said the distance was very important. It is right, we cannot imagine links of 400 m maximum length (the case for 100Base-FXHD). It's better using 1G and 10G with links of 40 km length.

Physical T1 and T3 are not fitted for long distance. And fiber interfaces are not standardized for these signals.

T1 and T3 frames can be carried on SDH, or OTH, or Ethernet.

Best regards, Michelot

Bonjour Michelot,

I read that 40 Gb/s Ethernet and 40 Gb/s SONET/SDH use the same optical transceivers. Maybe this is also true for 10GBASE?

So why should there be any physical layer differences in how they would be used in a WAN? I think the main differences now would be OAM and isochronous features (or lack of).

IEEE 802.3ae-2002 gives many PMD layers for 10 Gb/s Ethernets, among which are ones that seem well suited for WAN use, much like SONET OC-192. For example, 10GBASE-L, operating at about 1300 nm over single-mode fiber, is good to 10 Km, which is adequate for a MAN.

10GBASE-E over single-mode goes to 30 or 40 Km.

There are also multimode fiber PMD layers that are only good for short distances, as shown in Table 52-24. some of these, like the 62.5/125 um multimode fiber used in FDDI, is only good for 26 or 33m at 10 Gb/s! Obviously, not adequate for WAN or MAN use. But really, none of the short wavelength and multimode options are good for any WAN or MAN application.

So I'd say that the while Ethernet has fiber optic PMD layers that are not adequate for WAN or MAN use, it also has options that are the same as SONET/SDH, i.e. single-mode at long wavelengths, and they should be competitive with SONET/SDH in terms of distance.

Salut, Albert

Recognize that 100BASE-T2 exists only on paper; there have never been any commercial products or transceiver silicon produced.

I do have some 100baseT4 transceivers, though...

-- glen

10GE has several different physical layers, one of which is compatible with OC-192/STM-64. That particular PHY tends to cost several times what the more common Ethernet-based PHYs do, even at the same "distance" rating on the lasers.

I have no idea whether this severe price difference is due to engineering difficulty, lower demand, or market segmentation.

10GE WAN PHY can be handled by a carrier as if it were an OC-192/STM-64 signal; that's the entire point. The other 10GE PHYs cannot.

Vendors typically label these SX, LX, ZX, etc.

S

The big questions are whether you can lease dark fiber between locations and the distance (as the fiber runs, not the roads or birds) is within the transmit power capabilities of your hardware. If both of those are true, you can use any PHY you want, from 10BaseFL up to 10GBase-xx.

If they are not, you will need to find a carrier that offers an Ethernet transport service, and what speeds and interface types they offer will vary widely based on what hardware they use themselves. I wouldn't expect to find many that support 10Mb/s Ethernet, but here in the US there are dozens that will transport any of several 100Mb/s and 1Gb/s variants for you.

The only exception today is that you can use a 10GE WAN PHY with a carrier that thinks they're selling you an OC-192/STM-64 circuit. However, very few people can afford circuits that big or the equipment to terminate them.

S

Bonjour Stephen,

The word compatible is not really adapted, it's truly an OC-192/STM-64 interface. So the SDH compatibility is like between SDH interfaces.

[802.3] "The WIS provides a 10GBASE-W device with the capability to transmit and receive IEEE 802.3 MAC frames within the payload envelope of a SONET STS-192c/SDH VC-4-64c frame".

This is not the only case in IEEE. The RPR protocol specified in 802.17 uses GFP over SDH. But this protocol has not had the wanted achievement.

I don't understand that. I'm ready to pay 10G ten times more than 1G.

10GBase-W physical layer is STS-192c/VC-4-64c, added to a reduced SOH because the application is for UNI interface, transmitted by the client side.

In fact, you can transport any other 10G Ethernet type over a SDH network, classically, through GFP by using VCAT, e.g. VC-4-66v.

10GBase-W doesn't use the GFP adaptation. [from Albert]

It became common 802.3 in december 2005.

All Ethernet type can be carried over SDH, except 10Gbase-W which is really SDH at the physical level.

In the 10GBase-W family, you can have a MMF access link of 400 m with

10GBase-LW.

Best regards, Michelot

One can debate whether 10GBase-W is really a SONET/SDH interface or not, but the point stands that it's the only 10GE PHY that can _directly_ be connected to native SONET/SDH equipment. I used the word "compatible" because a direct connection can be made without any sort of encapsulation equipment, as would be required with any other PHY.

I mean that, for the same speed, fiber type, and distance rating on the lasers, a WAN PHY interface will cost significantly more than a LAN PHY interface; i.e. 10GBase-W products will cost more than 10GBase-S products even though both are MMF for up to 400m.

And that's why I used the word "compatible" above: there is no need for an additional adaptation or encapsulation layer.

Sure, you can transport another 10GE variant over a SONET/SDH network, but you need additional equipment to convert or encapsulate the payload. This is similar to, say, transporting Ethernet over a T1/E1; it's possible to encapsulate the former in the latter with special equipment, but nobody would say they are "compatible" in common speech.

One would be technically correct to say that 10GBase-W "is" SONET/SDH, but in my experience that confuses people, particularly those with an IP/Ethernet background. They're almost completely different technologies from a network engineer's perspective.

Actual vendors provide many more options than that; you can get WAN PHYs for both MMF and SMF with lasers ranging from 400m to 100km, just like the LAN PHYs.

S

This makes a good case for the importance of some understanding of the

7-layer ISO/OSI model. Without it, confusion is sure to occur. That was a topic of discussion on here some time ago.

By the way, a similar problem occurs when folk cavalierly compare 802.11 with 802.16 (Wi-Fi vs WiMax), which are quite different beasts. Here again, the OSI model helps.

Michelot a ecrit:

Yes, but I didn't consider 400 meters to be compatible with anything related to WANs or LANs, which I thought was your original question.

I don't believe it's essential to layer Ethernet on top of SONET/SDH for WAN or MAN service any longer. Of course, this can be done. And may make good sense in cases where the SONET layer is used as a common infrastructure, to carry packet switched as well as circuit-based, synchronous channels. (And then expect long tirades about whether or not such circuit-switched channels are ever needed at all.)

Bon soir Albert

Sorry. I meant WANs or *MANs*.

Bert

drive distance and fibre compatibility

"classic" telco provide single mode fibre, so you arent going to get very far trying to use anything else directly.

however the telco can give you a CPE device that adapts between ethernet versions so that the choice used over the WAN is invisible to you - see

a few examples....

yes - but you are limited to the "collision distance" - ie. where both ends reliably detect a collision. unfortunately - this is 400m.

there are no real advantages to using 10Base anything at present (except maybe on copper).

if you have dark fibre then drive at faster rates - more bandwidth for the same money (or even less given the cost of GBICs). If you dont then the telco service will dictate what you do and the standard to use.

Bonjour Rich,

Yes, for 100Mbit/s, the 100BASE-TX and 100BASE-FX are used today.

It seems that 802.3 standard doesn't mention the wavelength for MMF

100BASE-FX. Perhaps I don't see it. Is it only 1310 nm and never 850 nm ?

Best regards, Michelot

Bonsoir Stephen,

All client source signals can be connected as tributaries to a SONET/SDH node equipment, using the right interface and the right transport configuration.

Encapsulation through GFP, and not directly. This special equipment is a multiplexer.

So those people have to know IEEE 802.3 (12/2005 or previous amendments) where it is full specified, and not G.707 ITU-T that just gives some quick added things, like clocking.

You're right. Just 40 km without optical amplifier.

Best regards, Michelot

Bonsoir Albert,

You're right, SDH doesn't need Ethernet. We have IP over GFP over SDH, or Voice over MPLS over GFP. In WAN, the main advantages of SDH is OAM, reliability, protection, latency and very extended infrastructure.

In MAN we can discuss between Ethernet and SDH. And, in LAN, Ethernet is the king.

Best regards, Michelot

Bonsoir,

Thanks to Walter, Stephen and stephen for these word, your talking about dark fiber.

Your links are interresting: about 10 Mbit/s history, and the curious BT information where I need to look it again.

Best regards, Michelot

There is no rationale for using any half-duplex technology for WANs or MANs today, and it's getting very difficult to justify it even in LANs. "Collision distance" is a thing of the past.

S

Interestingly enough, the company I work for *prefers* SDH circuits without protection - we would rather create the necessary redundancy ourselves using several unprotected point to point links (using either SDH or WDM from a capacity provider). And we don't find SDH equipment to be sigificantly more reliable than Ethernet equipment.

So for us, the main advantage that SDH offers over Ethernet is OAM.

I'm afraid I don't understand the latency argument. Speed of light in fiber is the dominant part of delay here, and is of course the same for SDH and Ethernet equipment.

Steinar Haug, Nethelp consulting, snipped-for-privacy@nethelp.no