Boston Fiber Cuts Us Out

Readers take notice: With one swing of a backhoe in Boston, Light Reading's Websites went dark for more than two hours on Friday afternoon.

We apologize for any inconvenience. The event also gives us all good reason to question whether Internet data services are yet carrier-class.

A large fiber cut is believed to have caused server connectivity trouble to our service provider New Agora Corp., which leases collocation space from Allegiance Telecom Inc.'s (Nasdaq: ALGX) Boston Data Center.

An Allegiance technical support representative acknowledged that there was an outage in Boston that is believed to have been caused by the cutting of two OC3 connections. "Apparently, it affected [WorldCom Inc. operating unit] MCI and several other carriers as well," he says.

The outage could have been a lot worse. Given that the links were only down for between one and two hours, the cuts were probably clean, with no shredding, which takes longer to repair, the support rep says. Fiber cuts can take several hours to fix, depending on how clean the cut is, where it's located, and whether there is enough slack in the line to bring the two cut ends of the fiber together.

— Phil Harvey, Senior Editor, Light Reading

Page 1 / 2   >   >>
Carrier_Insider 12/5/2012 | 12:21:14 AM
re: Boston Fiber Cuts Us Out All Sprint's Fibers are redundant. So if a Sprint fiber is cut, the traffic is re-routed. You will not notice it.

ThouShaltNotJudge 12/5/2012 | 12:21:13 AM
re: Boston Fiber Cuts Us Out "You will not notice it."

... until the bill comes.
philipurso 12/5/2012 | 12:21:11 AM
re: Boston Fiber Cuts Us Out Consider TowerStream. Wireless T1 - T3, SLA, QOS guarantee. All POPs with at least 2 separate paths home to the internet. About 1/2 price. Not beholden to any phone company. One wire and one wireless connection give you true seperate egress and real redundancy. (Almost all our customers use us as primary.)
Consultant 12/5/2012 | 12:21:10 AM
re: Boston Fiber Cuts Us Out Sprint's fiber redundancy is meaningless if both sides of the SONET ring are in the same trench.

I bet Sprint has more collapsed SONET rings than buildings in Iraq.

You going to have to do a better sales job than that.
Carrier_Insider 12/5/2012 | 12:21:07 AM
re: Boston Fiber Cuts Us Out I'm an Engineer not a Sales person
arch_1 12/5/2012 | 12:21:04 AM
re: Boston Fiber Cuts Us Out Light reading is a customer of a data center and not directly the customer of a carrier. Light reading therefore has two choices if it wishes to improve its access resiliency: They can either find a data center with true carrier diversity, or they can place servers in separate data centers. The second choice is certainly more resilient, since you can put the servers in different cities or even on different continents, thus defending against natural disasters, DDoS attacks against co-located sites in the data center, and server and (with care) software failures.

It may even be cheaper, because they can use cheaper, less reliable data centers. This is particularly true if they can partner with another web site so that each serves as backup to the other rather than running duplicate sites.

The problem with this approach is this it requires the internet to "swing" to the backup server when the primary fails. This requires that the DNS be set up properly.

A second problem is that you need additional software to keep the secondary synchronized with the prmary. This is not rocket science.

This is a classic example of a much more general phenomenon: It is often possible to substitute computing power for bandwidth, or "good enough" for perfect. Also, most data applications, simply do not need SONET-like protection and will not buy it unless the cost is very low.

Just to calibrate the cost of this: Light reading can almost certainly run on a server that costs about $99/month. (This is a guess. what say you, LR?) Doubling this is not a huge expense.
joestudz 12/5/2012 | 12:21:02 AM
re: Boston Fiber Cuts Us Out Use FSO and a Mesh Network rather than a point to point interconnection
philipurso 12/5/2012 | 12:21:01 AM
re: Boston Fiber Cuts Us Out We think wireless is the solution for true seperate egress. Here's what TowerStream has built over Boston so far:

An OC3 wireless ring around the city with 7 POPs, each with at least 2 seperate egresses. The ring is connected to the internet at multi-carrier locations in 1) Boston and 2) Waltham.

The ring is self healing a la BGP. In addition, we are about to deploy a fiber connection from this ring to our Providence market, which also has a wireless ring with 2 seperated carriers linked wirelessly. It's not perfect, but pretty good. It goes home to 4 internet connections in 3 cities with no single point of failure.

For the last mile we use 2nd gen P2MP to deploy highly reliable wireless t1 to 6 Mbps and use a host of proven P2P radios for larger connections.

Most customers wanting absolute redundancy use us and a wireline connection with BGP for true seperate egeress, though it is possible in some cases to use 2 wireless connections to seperate POPs and have a degree of redundancy with TS alone.

We think of it as a seperate network above the city and not vulnerable to the same outages.

arch_1 12/5/2012 | 12:21:01 AM
re: Boston Fiber Cuts Us Out This is another example of diversified routing. It only works if the FSO (or other non-carrier local interconnect) connects to the internet backbone via truly separate paths. To me, FSO, Wireless, private fiber, etc., configured in a mesh, is a great way to bypass the local loop monopoly, but the local mesh still needs to connect to the core internet via truly separate paths to create the needed resiliency. Otherwise, a fire in a CO or in a tunnel in Baltimore can still isolate the mesh.
arch_1 12/5/2012 | 12:20:59 AM
re: Boston Fiber Cuts Us Out If my company were in the Boston area I would use your service to connect my building to the internet, bypassing the telco local loop and getting the robust resiliency you describe to allow high-speed access to the internet for my employees.

However, if I had a high-volume web site, I would not have it on a server in my building, sharing your OC3 with the rest of your customers. Instead, I would have my web site on a server at a data center connected via multiple OC-12s to the internet core.

(Disclaimer: I have no practical experience with colocation or with local access, so please educate me if this makes no sense.)
Page 1 / 2   >   >>
Sign In