XML File Size Bloat fingered for perf probs [ComputerWorld]

Date view Thread view Subject view Author view

From: Rohit Khare (rohit@uci.edu)
Date: Wed May 10 2000 - 12:11:13 PDT

Bloated file size an issue for XML

IT must plan servers, networks accordingly

By Carol Sliwa
05/08/2000 United Parcel Service of America Inc. plans soon to launch
a beta project using XML to exchange shipment, tracking and rate
information with select business customers.

But XML file size is a looming issue. The carrier manifest, one of
the largest files the Atlanta-based company receives via electronic
data interchange and Internet-based data streams, is now no more than
100MB in size.

However, UPS discovered that comparable XML files could be nearly
three times as large -- which will influence the infrastructure
decisions the carrier makes in the near future, said Augie Picado,
director of e-commerce deployment.

"The size of files is a key consideration," Picado said. "It's
something that we have to plan for."

Picado said he doesn't think bloated files will derail adoption of
the content-tagging language, nor do many other users, industry
analysts and consultants. But they cautioned that file size is one of
the many issues companies must take into account as they set up
networks, storage and servers to handle large volumes of XML
documents. For example, companies that use dial-up modems for EDI
might have to upgrade their connections to the data center.

"Just because it's easy to create an XML document, they shouldn't go
hog wild in sending a lot of extraneous, unnecessary data," warned
Rachel Foerster, a project team leader for the electronic-business
XML standards effort and a principal at Rachel Foerster & Associates,
an e-commerce consultancy in Beach Park, Ill.

Indeed, several early adopters said they use XML only to exchange
small amounts of data and have encountered no problems with large XML

John Deaton, vice president of planning at Office Depot Inc. in
Delray Beach, Fla., said his company's purchase-order transaction
sets are typically less than 1KB. "It's a nit on the network," he

Competitor Staples Inc. has seen no XML-related bottlenecks either.
Garn Evans, a manager of information systems at the Framingham,
Mass.-based office supply company, said a WebMethods Inc. B2B server
strips out the incoming XML tags and translates them into Staples'
proprietary document format before the information traverses the
internal network.

Some companies have more ambitious XML plans. David Westmoreland, CIO
at Arrow Electronics Inc. in Melville, N.Y., estimated that his
company handles $3 billion to $4 billion worth of EDI transactions
per year, but he said he expects all transactions, large and small,
to shift to XML during the next three to five years.

"You have to make sure, as you're scaling your network and computers,
that you're getting the business benefit out of the technology," said
Westmoreland. "And if you're not, you shouldn't do it."

Westmoreland said he thinks his company will win more business
because it can respond to customers in real time, using XML, rather
than waiting hours or even days for EDI transactions to turn around
in batch mode. While cognizant of XML file size, Westmoreland said
he's not worried, since network and storage costs are continually
coming down.

But some analysts and users caution that XML might not be suited for
every type of transaction. They expect that many companies will have
a blended XML/EDI strategy.

Ken Vollmer, an analyst at Cambridge, Mass.-based Giga Information
Group Inc., has predicted that EDI files will grow 30% to 40% as they
shift to XML but says he doesn't think file size will create a
crisis, because XML adoption will be gradual.

Forrester Research Inc. analyst Josh Walker said companies will learn
to architect around the XML file size issue. Products that can
quickly translate XML, from vendors such as Active Software Inc.,
Neon Software and Vitria Technology Inc., will also help, he said.

Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed May 10 2000 - 12:12:44 PDT