Best of the ISP-Lists

Fixed Wireless Equipment

Loooong-Haul Wireless

WISPs search for ways of overcoming the lack of available wired backbone/backhaul connections—or simply to beat their high cost. When customers are a long, long way from NOCs, challenges multiply and choices shrink.

[July 1, 2003]

On ISP-Wireless in April 2003, DW polled the membership for ideas on how to span the wide open spaces wirelessly:

"Hi, how are you guys handling backbone for real rural areas? From what I saw, the longest distance you can shoot a backhaul is about 50 miles (with the Redline gear). What if from 50 miles from where you are, you're still in the middle of nowhere? I know I could probably get a T-1 from ALLTEL or heaven forbid an ATM from the local Sprint telco, but I'd really like to backhaul this back to a DS3.

I need about 80-100 miles. Anything out there that can do that? 10 Mbps should do fine."

[PG (a satellite consultant) provided some basics on satellite solutions] "You can get satellite backhaul just about anywhere at just about any speed you want. This is the most common method of backhaul in many countries outside of the U.S. There are both shared and dedicated bandwidth solutions, both symmetric and asymmetric. A satellite provider will land your traffic at their teleport and connect you to the Internet backbone at high speed with very few hops.

You can also get point to point [PtP] circuits and backhaul the traffic to wherever you want it to go. It comes down to a financial decision. The cost of dish equipment for the speeds you discuss are not trivial, but building towers on mountains in the middle of nowhere isn't trivial either.

Note that latency will be a factor for applications like gaming."

[FM advised] "Make sure the sat provider is going to be around though. I picked up a rumor a while back on ISP-CEO that Starband isn't doing too well. Since I don't know this as fact, please don't take it as such, but it might be a good idea to get a prospectus, FWIW."

[KB seconded this] "I have picked up three new accounts this month alone for my WaveRider system that were with Starband . Everyone I talk to has nothing but complaints about them, so it wouldn't surprise me."

[PG waded back in with an extended primer on satellite backhaul] "I should have said more in my first message: Starband and DirecWay are consumer class products and are not designed to provide bandwidth that can be resold. These products would not come remotely close to providing the type of service you need for Internet backhaul. The technology used to deliver these solutions is DVB/TDMA. DVB is what DirecTV and Dish network deliver TV satellite service with; IP is merely piggybacked on top and inserted into MPEG frames—albeit with a significant loss of efficiency. TDMA is a way for many remote sites to contend for and share inbound bandwidth using a mechanism that works much like shared Ethernet with all the collisions, backoffs, retransmits, etc. It also is very inefficient in most implementations, but it's cheap—relatively speaking.

Backhaul-class connectivity is generally provided using dedicated bandwidth with technology called SCPC or Single Channel Per Carrier. Because the bandwidth is not shared, it is more expensive. Latency will be about 300 ms in each direction. The customer has a dedicated pipe, just like a point-to-point T-1. Voice quality can be excellent. Satellites have been carrying analog voice with SCPC for decades. There are a number of satellite space fleet providers who provide SCPC bandwidth including SES-Americom (formerly GE), PanamSat, Telesat, Loral Skynet, New Skies, etc. These are large stable companies in a completely different class from Starband (a service provider that does have financial issues).

There are also enterprise class shared broadband satellite solutions that will support backhaul, VoIP, etc. The first class of solutions is generally DVB/SCPC—that is, shared bandwidth outbound and dedicated bandwidth inbound. In addition to dedicated SCPC solutions, a shared enterprise class solution that I work with has an advanced TDM/TDMA solution that provides a unique mix of dedicated and shared bandwidth, with the ability to apply CIR and QoS to ensure voice quality. It is the most bandwidth-efficient shared solution I am aware of. This solution provides cost savings by using shared bandwidth, but also provides data reliability features with CIR and application-QoS capabilities. It is a very popular solution for backhaul connectivity in places like Africa, Mid-East, South and Central America, etc. Here in the states it is very popular as an alternative to T-1 or Frame Relay for remote offices and businesses that have difficulty obtaining these services—or who must pay very high costs for wireline services."

[MV (a representative of Redline Communications) headed off in a new direction] "Without knowing much about your terrain, I'd say you should be able to build a highly reliable link going 80 to100 miles using a couple of hops using Redline gear. We officially support 50 miles, but have one customer in Idaho doing 72 miles in one hop. (He had a couple of mountaintops to help.)"

[SS (a representative of DefactoWireless.com) added his two cents] "We can typically provide service within 100 miles of our fiber line. We have gigabit capacity and it's all Ethernet based so it's very simple for you to connect. We drop a port at your NOC or tower. Also, if you can tap into the fiber in two places it will take over 20 major acts of god to take out your connection."

[MS offered his own take] "I run T-1 for that kind of thing. If you need 10 megs, though, you're probably better off building 2 to 5 point-to-point (PtP) links. Monthly payments on that will probably be much less than multiple T-1 circuits."

[SS agreed] "Right. Until you are in a colo facility and start buying megabits you'll find the T-1 approach with radio backhaul cheaper. Also remember that your T-1 providers in rural areas are not Tier1 for service provision and failures are increased greatly.

[MS continued] "I'd try to keep the PtP shots at less than 30 miles if possible. I've known people that have done 40-ish miles but they take more errors and seem to need more work. Some of the newer gear may be better, but there's not enough of it out there yet to be sure. And the more I learn, plus the fact that there are going to be a lot more wireless systems out there in the future, I'm leaning more and more in the direction of short links where possible."

[SS enthused] "That's the idea!"

—End