[Cutting Edge II]
>>>So amati is not on the cutting edge. Where is their product?<<<
There are thousands in trials around the Globe.
45 DSLAMS were sent to a French Telco!
Dave - Dont stand to close the The Edge, you'll get Cut!
Proof-of-concept tests show that ADSL and
HDSL modems have what it takes to deliver
high-speed services.
Front-Runners
Several varieties of DSL technology are now in various stages
of development (see "Hurry Up and Wait"). But the two top
contenders for the local loop are ADSL (asymmetric digital
subscriber line) and HDSL (high-data-rate digital subscriber
line). On paper, ADSL delivers data rates as high as 9 Mbit/s
downstream (from the central office to the customer site) and
up to 640 kbit/s upstream (from the customer site to the CO).
By contrast, HDSL is symmetric--it delivers data at rates up
to 1.544 Mbit/s in each direction. The symmetric design has
enabled telcos to deploy HDSL as a T1 access technology in
the local loop (see "The Mother of Invention"). SDSL
(single-line digital subscriber line) is similar to HDSL in that it
offers two-way 1.544-Mbit/s channels. The key difference is
that, as its name implies, SDSL works over a single copper
wiring pair; HDSL requires two copper pairs. The single-pair
design makes SDSL better suited for residential use, but it also
limits maximum operating range to about 10,000 feet.
For this proof-of-concept lab test, tele.com teamed up with
National Software Laboratories Inc. (NSTL, Conshohocken,
Pa.) and TeleChoice Inc. (Verona, N.J.). NSTL, an
independent test lab owned by the McGraw-Hill Companies
Inc., handled all product testing. TeleChoice, a
telecommunications consultancy, helped develop the test
methodology and evaluated the final test results (see "Test
Methodology").
tele.com invited 10 different DSL modem vendors to
participate. Three accepted the invitation: Amati, Orckit, and
PairGain (see "The DSL Modem Test Lineup"). Amati and
Orckit submitted their ADSL modems for test; PairGain
submitted its SDSL product. Although the test bed was
optimized to evaluate ADSL modems, the line simulator used
in the test also was able to accommodate SDSL tests.
Because these units represent DSL's first generation, it's
important to remember that they don't provide an indication of
where the technology could be but where it is right now. In
fact, Orckit refers to its submission as a prototype; at test
time, it said a second-generation product would be shipping
this month. Next-generation devices likely will offer some key
improvements not only in operating speed but also in power
consumption, size, and ease of installation.
Noticeably missing from the lineup of ADSL modems in this
evaluation are any using CAP (carrierless amplitude and
phase) modulation. This de facto standard is present in the
majority of the ADSL trials worldwide, but at press time it still
had not attained standards status. Although vendors of
CAP-based modems elected to not participate in this round of
evaluations, many have said they will participate in future
tele.com lab tests (see "Where's CAP?").
Given that this is a first crack at ADSL--hence the term "proof
of concept"--the goal for this lab test isn't to compare
performance numbers but to get an idea of how well the
modems as a class perform under different line conditions. As
DSL technology matures and more suppliers offer established
working products, tele.com will revisit the technology to take a
close look at how products match up in terms of performance.
To The Test
For this proof-of-concept evaluation, the test bed was set up
to answer one overarching question about DSL modems: How
well do they deliver the bandwidth needed to accommodate
high-speed data services? The nine different industry-standard
tests used in the evaluation measure performance over lines
characterized by different transmission lengths and with
different sources of interference. One test--the CSA 0
test--served as a baseline. It measures throughput at a very
short distance (10 feet) with no signal interference. The other
eight industry-standard tests measure throughput at various
transmission lengths (from 6,000 feet to 15,000 feet) and with
various line impairments, such as white noise and near-end
crosstalk generated by the presence of other ADSL, HDSL,
ISDN, or T1 signals.
One battery of tests measured the raw throughput of DSL
modems under these different line conditions. The underlying
protocol used for the raw throughput tests was UDP (user
datagram protocol). UDP is a good benchmark for raw speed
because it doesn't require acknowledgments to be passed
between the sending and receiving modems. A second set of
tests measured modem performance under more real-world
conditions. In that second battery, traffic was sent between
modems using FTP (file transfer protocol), a protocol
commonly used in Internet applications. FTP's underlying
protocol is TCP/IP, which unlike UDP does require
acknowledgments to be passed between modems. In Internet
applications, each FTP session opens a separate TCP/IP
session. Downloading a Web page can require the opening of
several simultaneous sessions. For this reason, the test bed
included two separate FTP tests: one for a single session, and
one for connections involving as many as eight simultaneous
sessions. As part of the FTP evaluation, testers also measured
latency--the delay introduced by the DSL modem in passing
FTP data. Although latency is not a big problem for most
conventional Internet applications, such as the downloading of
World Wide Web pages, it can be troublesome for real-time
applications like Internet telephony or videoconferencing.
Now that many such "streaming" applications are showing up
on the Internet, latency could be a key factor in overall modem
performance.
As was expected, all four modems tested showed their best
results in the CSA 0 tests. In the CSA 0 raw throughput test,
Orckit's ORvision modem delivered 512-byte packets at
6.949 Mbit/s downstream and 661 kbit/s upstream (see the
performance tables for test results logged by each modem). In
its product literature, Orckit claims top rates of 8 Mbit/s
downstream and 640 kbit/s upstream. In the CSA 0 FTP
tests, ORvision clocked in at 6.168 Mbit/s downstream for the
simultaneous-session test.
The two Overture 8 ADSL modems supplied for the test by
Amati were configured with different software loads. One
modem was configured to deliver 1.5 Mbit/s downstream and
160 kbit/s upstream; the other was set to operate at 4 Mbit/s
downstream and 384 kbit/s upstream. Both Amati modems
performed as advertised in the CSA 0 tests for raw throughput
and simultaneous FTP sessions. The sole SDSL modem in the
test, PairGain's Megabit Modem, came very close to matching
the 768-kbit/s two-way performance claimed by its vendor. It
registered 744 kbit/s downstream and 751 kbit/s upstream in
the CSA 0 raw throughput test and 740 kbit/s in the
simultaneous FTP test.
Model Of Consistency
Both Amati modems and the PairGain product came close to
matching the CSA 0 benchmark speeds for every distance and
line-impairment condition tested. The version of the Amati
Overture 8 configured for 1.5-Mbit/s performance was an
absolute model of consistency: In all nine raw throughput tests,
it delivered exactly 158 kbit/s of upstream throughput, while
downstream speeds stayed at a consistent 1.515 Mbit/s for all
but one test--in which measured performance dropped by all
of 1 kbit/s, to 1.514 Mbit/s. Although this Overture 8 model
did show some drop-off in performance for simultaneous FTP
sessions in some tests, throughput at distances greater than
12,000 feet actually improved slightly over the CSA 0
throughput benchmark.
The Amati Overture 8 modem configured for 4-Mbit/s
performance and the PairGain Megabit Modem also
demonstrated a high level of consistency across the distance
tests for both raw speed and FTP traffic. With both these
products, however, tests covering distances over 12,000 feet
were not applied because in both cases the vendors said their
products weren't intended to work over lines greater than
12,000 feet.
Any one have the Tel.com URL, so ole Dave can finish the stories...
JW@KSC |
