|
Hi all-
It looks like you Ascendites could use a little cheering up. This
is off the Pairgain thread:
Ascend vs. competition modem tests. See nwfusion.com
Review of DSL modems from four
vendors
By Jim Brown
Network World, 11/3/97
Luck has more influence than solid planning on whether you'll hit pay dirt with Digital Subscriber Line (DSL) services. For openers, you've got to be in an area with a service provider that is knowledgeable and daring enough to deal with the offering. Then you have to cross your
fingers and hope you'll get a pair of copper wires clean enough to carry the signal the short mile or two it's designed to travel.
At least that's what we found in testing a DSL circuit supplied by HarvardNet, an Internet service provider and competitive local exchange carrier (CLEC) in Harvard, Mass. While we had little trouble getting the circuit installed within the promised two-week window, we met with varying levels of success in getting different vendors' DSL equipment to operate over the circuit. In fact, only two of the five products we examined worked over the line and both operated at less than 1M bit/sec - a far cry from the 7M bit/sec many DSL equipment vendors tout.
The products that successfully shipped data over the line, Paradyne Corp.'s HotWire and Ascend Communications, Inc.'s DSLpipe, support rate adaptive DSL (RADSL), which enables the equipment to adjust its speed to account for line conditions. We never got a symmetric DSL (SDSL) unit for Ascend, asymetric DSL (ADSL) equipment from Amati Communications Corp. or a RADSL unit from PairGain Technologies, Inc. to work over the line.
It was a different story in a lab test done at Virtual Interoperability Testing and Applications Lab (VITALNet), a partner test lab in Research Triangle Park, N.C. Under controlled conditions, all the equipment operated at or near top-rated speed when linked over a short length of twisted pair cable. It's more common than you'd think for some units to work on a particular line while others do not. Chip Ach, HarvardNet's director of engineering, says it's typical for the service provider to have to install several different pairs of wires to customer sites before finding one that works well enough to carry the DSL signal.
HarvardNet also realizes that line conditions may prohibit equipment from hitting top speed, so it guarantees a top rate of only 768K bit/sec on its DSl lines. This is because HarvardNet and many other service providers rely on so-called burglar alarm or local- area data service (LADS) circuits they acquire from local exchange carriers (LEC). Also known as dry copper or an unbundled local loop, these wire pairs may be old, in need of repair susceptible to crosstalk, or fall victim to some other malady.
Unfortunately, there is no easy way to tell exactly what's causing a problem on the line when you encounter one. Just follow along with what happened to us.
We ordered the line on Sept. 8 and it was installed on Sept. 22. The line connected our lab to a Bell Atlantic central office (CO) 1.8 miles away. That office was linked to a HarvardNet point of presence a few hundred feet from the CO. In other words, we were within DSL distance limits. With Ach's help, we installed the CO equipment component of each vendor's DSL product at the HarvardNet POP and tried to get it to communicate with the customer premises equipment (CPE) in our Framingham, Mass. lab. In addition to a DSL interface, the CO and CPE pieces of each vendor's product has a 10Base-T Ethernet port for connection to a hub on either side of the DSL link. This configuration sets up the DSL equipment to act as an Ethernet bridge or router, depending on the type of software each vendor provides.
We tried PairGain's Megabit Modem CRA first and quickly ran into problems when it failed to connect using a default configuration. Using the management interface on PairGain's self-enclosed CO unit, which looks like a modem, we dropped the top speed the pair would try to connect at and adjusted the amount of noise the units would tolerate to the highest level permitted. Still, nothing. To ensure the DSL line would support a connection, we plugged an analog phone into each end and from our lab had a conversation with someone at the POP.
HarvardNet asked Bell Atlantic to trace the circuit and make sure it was clean and all connections were tight. Despite checking several times over several weeks, we never heard back. Oddly, Ach says the PairGain units linked up at near top speed on another DSL line to a building a few hundred feet from ours. This further shows that dealing with DSL takes a little luck.
After fiddling with some parameters on Paradyne's chassis-based HotWire CO unit, such as stepping down the maximum link speed it would establish, we finally got it to connect with the modem-like unit in our office lab. However, we had to settle for a much lower speed than the 1M bit/sec uplink and 2.5M bit/sec downlink speeds the HotWire units achieved in the VITALNet lab. Likewise, Ach said the HotWire products connected at higher speed over the DSL line to the building a few hundred feet from ours.
HotWire only operates in static routing mode, which made configuration a little more difficult. However, that should not concern you because the HotWire CO unit clearly is designed to be managed by ISPs or carriers, who can downline load settings to the CPE. About the only thing you will need to do is make sure your internal TCP/IP addressing scheme is compatible with what was fed into the HotWire. Paradyne says it will add bridging to ease configuration woes.
Ascend's RADSL equipment deserves a tip of the hat for being one of the two products that operated over our DSL line. But, it only established a link after we dropped the maximum speed it would operate at from 7M bit/sec to 2.5M bit/sec. Even then we wound up with a downlink rate of only 640K and an uplink rate of 272K bit/sec.
Once it started working on the DSL line, we had no problems with Ascend's RADSL unit and it was a cinch to configure in bridging mode. However, configuration can become more complicated once you try to take advantage of the product's rich feature set, which was the richest of all the products we examined. Those features include support for IP and IPX routing, as well a various forms of packet filtering and security protocols such as Challenge Handshake Authentications Protocol (CHAP) and Password Authentications Protocol (PAP).
In looking at Ascend's SDSL and Amati's product, we wonder if the reason they didn't work over the DSL line was because they were trying to connect at a higher speed than the line would bear. Ascend's SDSL unit tried to establish a bidirectional 768K bit/sec link. Amati's chassis-based Allegro ADSL product seems to be in the same boat. The unit tries to link up with the Overture 810 CPE at a speed that is faster than the top speed we got off the line. We could not find a way to adjust that speed downward.
Because of the line problems, our testing was inconclusive on whether Carrierless Amplitude and Phase Modulation (CAP) line encoding is any better than the competing Discrete Multi-Tone (DMT). CAP was developed by AT&T Paradyne and is now controlled by Globe Span Technologies, the sole producer of CAP chip sets. DMT is an emerging international standard that has won support from a number of chip makers. Of the five products we examined, only Amati's supported DMT.
Overall, our experience with DSL shows you'll be taking shots in the dark in trying to get a good operational DSL link until the carriers - both the traditional local exchange and CLECs - reengineer their local loops to provide top-notch copper pairs or equipment providers design affordable products that can overcome the problems in existing lines.
|
