The Ethernet Revolution -- Scott Clavenna

February 13, 2001

8 Min Read
Comes the Ethernet Revolution

Scott Clavenna - Director of Research at Light ReadingAll advocates of optical networking are revolutionaries at heart. After all,we want to build a better tomorrow, one that consigns the bottlenecks andbackups of today’s Internet to the Dustbin of History and delivers cheap,plentiful bandwidth to the masses.

Ironically enough, this is one revolution that’s got the backing of thenetwork “owners” — the carriers, CLECs, and service providers who see bravenew profit in this brave new world.

As I detailed last time out, Ethernet is the key to transforming the publicnetwork (see Endless Ethernet?). All the carrierswill eventually use this technology as a universal interface to theirnetworks, leveraging its low cost and scalability to transport all manner ofIP services end to end.

I also pointed out that optical Ethernet need not be “pure packet.” With theright hardware and software it can easily be provisioned over a next-genSonet structure.

But if the past 20 years of networking has taught us anything, it’s thattop-down revolutions rarely succeed. So why do I believe so fervently thatoptical Ethernet will not wind up being ATM all over again?

Because this is a populist movement, driven by legions of network users.They may not sing defiant songs while ATM heads roll into a basket. But theywill line up behind optical Ethernet because its message is compellinglystraightforward and pragmatic.

Thus far, discussions about optical Ethernet have centered on how it speedsprovisioning, reduces operations and services costs, and manages IP servicesas deftly as Sonet add/drop muxes manage phone calls. David Yates of Atrica Systems puts itsuccinctly: “Transmission bandwidth is the foundation for all value-addedservices in the network.”

That’s good news for network operators looking for low-cost solutions thatsupport IP, something that can turn all their bandwidth into real servicerevenues. In fact, optical Ethernet delivers bandwidth at such low coststhat carriers have an opportunity to completely rethink how they design and build their networks.
This begs a simple question: Do enterprise networks need to be transformed?Well... Yes. They’re expensive to scale and maintain. They require constantreconfiguration as new users or remote sites are added. And they lock ITmanagers into an endless cycle of training to keep up with new technologiesand services.

I defy you to find an enterprise network manager who wouldn’t like to addusers easily; increase bandwidth between locations without deploying a meshof private lines; reduce WAN services costs; trim specially trained ITstaff; and outsource some apps or backup storage servers when necessary.

And then try to find one who doesn’t want more bandwidth — way morebandwidth.Can optical Ethernet deliver?

Yes — given the right combination of equipment and services support.

Let’s take that net managers’ wish list item by item:

  • Add users to the network
    With an optical Ethernet service, users andlocations can be added via simple LAN-switch port attachment. Ideally, theyshould have the equivalent LAN-like access to all network resources,collapsing the network to a virtual LAN. If you’re not on the network, youneed your service provider to get you on. This could take time and involvedeploying new fiber or alternate last-mile solutions. Time will tell howlong the last-mile bottleneck remains.

  • Increase bandwidth easily
    Most optical Ethernet services offer some kindof customer network management that allows service providers or end-users toprovision bandwidth via software commands. And this bandwidth is cheap,compared with adding T1 lines between sites. Typically, no new facilitiesneed to be added to the network, just increased switch capacity on the ring.Optical Ethernet solutions add the verifiable and assured SLAs (service-level agreements) missing intoday’s gigabit Ethernet switches.

  • Reduce WAN service costs
    This is a real home run and will be realizedearly and gratefully by most enterprise users. There are two ways toquantify this advantage. First, WAN bandwidth should be cheaper withEthernet because the associated transport equipment is so much cheaper thanATM/Sonet gear or multiple routers.

    Second, the CPE (customer premises equipment) is much cheaper,requiring only a port on a low-cost gigabit Ethernet switch. You can getthose from Cisco Systems Inc. (Nasdaq: CSCO), Extreme Networks Inc. (Nasdaq: EXTR), or Foundry Networks Inc. (Nasdaq: FDRY) for less than ten grand. Compared with a framerelay switch, router, CSU/DSU, or whatever a WAN manager has been usingbefore, this can result in savings of up to 90 percent.

  • Trim specially trained IT staff
    Ethernet services can be managed by a LANmanager. In fact, some Ethernet services don’t even require the customer to have a router on the premises. A managed service that includes firewallingcan easily deliver the features offered by frame relay, at a much lowerprice with a significantly lower cost of ownership. The only gear at thecustomer premises is a LAN switch, which was already there.

    Managing network services now becomes as easy as managing the LAN. The ITstaff can go back to overseeing apps, setting up a high-bandwidth intranet,or other more valuable activities. This also means that small andmedium-sized businesses can migrate to high-bandwidth services withoutstaffing an IT group.

This gets to the heart of why Ethernet will not go away. It so easilysatisfies the requirements of both the service provider and the user.

It also gets to a possible big payoff for the enterprise: network-centriccomputing.

No sooner are those words out than the specter of Larry Ellison, Oracle, appears. Afterall, he has long argued that the network is where the action is, not the PC,which can be dumbed down to take advantage of advanced networkingtechnologies and available bandwidth. In Ellison’s view, the PC provides aneasy-to-use interface to the resources of “The Network,” which can be theInternet or a departmental LAN or anything in between.

Bill Gates, Ellison’s archnemesis, has become the world’s richest humanbeing by betting that the PC is where it’s at. For Gates, the network doesn’t do much more than deliver basic connectivity between processing- andstorage-rich desktops.

Optical Ethernet proves they’re both right. The network can nowcarry sufficient bandwidth to support a mix of centralized resources andoutsourced storage, for example, while empowering PCs to be as user-definedand customizable as possible. You can focus your energies on either your PCor your databases — either way, you have the bandwidth to make it happen.

To see how, let’s take a look at typical enterprise. Distributing networkresources like file servers, databases, and storage means distributing themanagement of those resources. That means cost, complexity, andcomplications trying to scale the network uniformly.

Consider an enterprise with one large data center and eight remote sites ina metro area. Corporate networkers typically link those sites with privatelines. Some may opt for frame relay, creating a VPN (virtual private network) among sites — particularly if they’re distributed over a relatively largegeographic area.

This level of interconnection is fairly low speed, one or a few T1s, whichdoesn’t add up to much if end-users are accessing remote storage ormultimedia databases and content. It certainly doesn’t qualify as LAN-likeaccess. To get that, enterprises typically distribute servers to each ofthose sites, giving users fast, reliable access to the content theyregularly work with.

But here’s the catch. Eight 10-user application licenses cost more than one80-user license. What’s more, if a new software revision comes out, usersmay not all be upgraded at the same time, limiting the sharing of fileswithin the enterprise. In all likelihood, folks at corporate HQ will havemuch better IT support than those at remote sites.

Here’s where optical Ethernet can lend a big hand, courtesy of its cheap,flexible bandwidth.

In the network-centric model, an enterprise subscribes to an opticalEthernet service or builds its own private network, connecting remote sitesto HQ over low-cost gigabit Ethernet links. Remote users appear as nodes onan Ethernet LAN. They have the same access to network resources as anyoneelse. According to Bob Schiff of Lantern Communications, “The distinction between LAN and WAN becomes more that of administrative andoperations ownership than that of bandwidth and geography.”

The enterprise can now move all databases, file servers, storage servers,and other resources back to headquarters (or outsource some of them if theygrow too large to handle). This immediately improves operations,reliability, and manageability. Remote sites require only a single LANmanager. The specialists remain at HQ, but in this scenario there’s no unequaldistribution of services or support. If a new version of some corporatesoftware appears, all users immediately have access to it.And when the MIS director decides to outsource storage, optical Ethernet is equally capable of delivering the goods.

Clearly, it takes time to transform computing paradigms. But the success ofoptical Ethernet doesn’t depend on enterprises adopting an entirely newcomputing model. Optical Ethernet allows the enterprise to put computingwhere it makes the most sense — centralized, outsourced, or wherever it belongs. Unlikemost things in life, optical Ethernet has two upsides: one for serviceproviders, the other for the enterprise. Both will benefit to such a largeextent that there is one inescapable conclusion: Vive la révolution!

You can use our message board to comment on the revolution in the making orregister your counter-revolutionary complaints. Just use the link below.

— Scott Clavenna is president of PointEast Research LLC and director of researchat Light Reading

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like