[Click here for Novell Intranetware] [Cascade @ the heart of the Internet]
------------------------------------------------------------------
ADSL trial surprise: 6M bit/sec just a pipe dream
Internet servers and access gear fail to keep pace with broadband
line speeds.
[FYI:] By Tim Greene
Network World, 10/7/96
Contact Senior
Writer Tim Waltham, Mass. - Through no fault of its own, it
Greene looks like Asymmetric Digital Subscriber Line
(ADSL) technology by itself is not the answer to
ADSL FAQ - From high-speed Internet access.
the ADSL Forum.
Recent tests by GTE Laboratories, Inc. found
More technical that even if Web browsers were to connect to Web
info about ADSL servers via 6M bit/sec ADSL lines, they probably
- From the ADSL could not take advantage of the bandwidth.
Forum.
''Web browsing performance is substantially
HDSL moves to poorer than ADSL bandwidths allow for,''
data forefront concluded a GTE Laboratories report submitted to
- Network the ADSL Forum in London.
World, 10/7/96.
The news is discouraging for carriers such as
Network World GTE Corp., which is looking to offer ADSL
articles about carriers such as GTE Corp., which is looking to
ADSL offer ADSL service to support Web browsing and
other applications.
The actual test involved client machines
equipped with Microsoft Corp. and Netscape
Communications Corp. browsers downloading files
from their own cache and from a Web server
attached via an uncongested 10M bit/sec Ethernet
link.
Downloads of text, image and mixed text/image
files topped out at less than 1M bit/sec. But
downloads of mixed text/image files barely took
advantage of the bandwidth offered by a standard
28.8K bit/sec modem (see graphic, page 1).
''If it's only as fast as your modem goes, who
needs ADSL?'' said Ira Machefsky, an analyst
with Giga Information Group.
The problem with the test, though, is that it
does not finger the culprit, Machefsky said.
Bob Olshansky, manager of advanced services
platforms for GTE, acknowledged the problem
could be with the server, the browser, HTTP
itself or the TCP/IP stack used.
Machefsky said the problem might also rest with
the PC disks, transfer rates, memory or other
possibilities.
However, one analyst pinned the rap on the
browsers, which open several TCP/IP sessions to
download each Web page.
''If those sessions were opened simultaneously
instead of sequentially as they are now,
throughput would increase dramatically,'' said
Kieran Taylor, broadband consultant for
TeleChoice, Inc., a consultancy in Verona, N.J.
GTE is doing more tests to try to find the choke
point.
Testing, testing
The GTE tests were designed to determine how
fast browsers could download different types of
files.
The browsers ran on five client hardware
platforms: 486, Pentium 90 and Pentium 166 PCs,
Mac Quadra and Mac Power PC.
They accessed a Macintosh PowerMac 6150 running
Quarterdeck Corp.'s WebStar server software over
a 10M bit/sec Ethernet connection. The server
fed them 13 types of files, including text,
JPEG, GIF and HTML text with in-line imaging.
GTE also performed FTP transfers across the
platforms.
Download times were measured from when the end
user clicked the mouse button until when the
final display appeared on the screen. So the
times recorded were affected by server file
fetches, data transfer, client decompression and
formatting, and caching to the client disk.
GTE allowed that the speed problem might lie in
the browsers themselves, where data is
decompressed and formatted before being
displayed.
''Our goal was to stimulate the industry to
address the problem,'' Olshansky said.
------------------------------------------------------------------
[NW Fusion home ] [News+ ] [Netref ]
[ Feedback ] [ Search ] [ Product Locator] [Stocks ]
Copyright 1996 Network World, Inc.