You want it when?

*You want it when? InteropNet team must install 120 miles of network connections in 48 hours. (Bo Pitsker, manager of Interop+NetWorld network)(PC Week Netweek)
PC Week May 2, 1994 v11 n17 pN1(3)
PC Week May 2, 1994 v11 n17 pN1(3)

You want it when? InteropNet team must install 120 miles of
network connections in 48 hours. (Bo Pitsker, manager of
Interop+NetWorld network)(PC Week Netweek)

by Blakeley, Michael

Abstract
Bo Pitsker is in charge of installing the 7,000 host computers that
comprise the computer network during the NetWorld+Interop trade show
and has plenty of advice for corporate managers of information systems
concerning network installations. Equipment redundancy is important
because it helps keep the network running and is more important to
Pitsker than performance characteristics. He also points out routers
using the Open Shortest Path First (OSPF) protocol adds another layer of
redundancy to the system. Pitsker's crew installed matrix switches in
the system to remotely link nodes to hubs because of their flexibility
and believes the cost of manually patching nodes is higher than the
initial cost of the matrix switches. He advises network managers to
hot-stage a network to solve any problems before the system becomes
an integral part of a company's operation.

Full Text
IS managers and network designers looking for clues to set up a
corporate LAN, take a look at the heart of this week's NetWorld+Interop
trade show in Las Vegas. The InteropNet is a live network linking every
booth on the show floor, using Ethernet, FDDI, ATM, and WAN
technologies.

The man behind the network -- Bo Pitsker, manager of worldwide
networking at ZD Expos, in Foster City, Calif. -- is no wirehead theorist.
His experience putting together the InteropNet makes him a valuable
source of information for corporate network managers.

Pitsker doesn't just plug all the wires together on the InteropNet. His
job also includes designing the network, managing a team of up to 120
technicians, and dealing with logistical and political problems.

"We're doing it in a trade show environment, but every corporate IS
planner has the same problems," Pitsker said. Technology selection is
based on "our need to implement the network rapidly and keep it alive in
a chaotic, less-than-perfectly ordered world."

InteropNet connects approximately 7,000 host machines to one another
-- as well as to the Internet -- using 120 miles of wiring. It makes use
of Token-Ring, 10BaseT, FDDI, and ATM (asynchronous transfer mode),
running in several buildings at NetWorld+Interop. The buildings are
linked using dedicated lines, fiber stands, and wireless bridges.

Logistics are a big part of solving the "rapidly" portion. Like any
corporate IS network, changes to the InteropNet can't happen after the
users arrive in the morning. In Pitsker's case, there's a 48-hour period in
which his team must install and troubleshoot the InteropNet -- starting
just after NetWorld+Interop arrives to take over the Las Vegas
Convention Center, and ending when the exhibitors' freight arrives.

In the course of making certain the disparate technologies work
together, Pitsker and his team work weekends and night shifts.

"Sound familiar? Our deadline can't slip. If we don't get it installed by
the time exhibitor freight comes in, we can't get it installed at all,"
Pitsker said.

It's a situation resembling that of a corporate planner who must cut
over to a new operating system, protocol, or topology, according to
Pitsker. "Speed of implementation influences the technologies we use,"
he said.

Keep it redundant

Redundancy is another factor used to differentiate the technologies. Any
extra redundancy helps Pitsker and his team manage less and keep the
network running longer.

FDDI and Token-Ring, for instance, have built-in redundancy. "Frequently
they're purchased for their performance, but the redundancy is more
important to us," Pitsker said.

Routers also lend themselves to redundancy in the network. "Say you've
lost your link. OK, go find another path and use it until the primary path
comes back," he said. Routing based on the OSPF (Open Shortest Path
First) protocol makes this possible, but such on-the-fly switching isn't
possible in a bridged environment.

Another important design goal for the InteropNet is the ability to make
changes in a controlled manner -- and this is where some information
systems go astray.

"Frequently, their choice of technology is completely decoupled from its
implementation. That's a mistake -- you're betting the farm without
understanding the implications of implementing your changes," Pitsker
said.

Structured wiring is an example of Pitsker's commitment to flexible,
controlled systems. He uses matrix switches to remotely connect nodes
to hubs, including the ability to switch a single Network General Co.
Sniffer between any one of dozens of Ethernet segments.

"Suppose a vendor has just changed booths, which is exactly the same as
a person moving offices," he said. "Do you physically have to go and move
every wire, even punching down in a closet?"

Matrix switches cost more initially than a punch block or patch panel,
but Pitsker recommends corporate planners sit down and figure out the
life-cycle cost of changes, in terms of lost time and lost opportunity.
"You're going to be shocked at how much money you spend patching nodes
manually," he said.

Return on investment is the key to justifying this kind of automation,
both for InteropNet and in the real world. "At some point, there has to be
a connection between business need and the investment in networking,"
he added.

Sometimes, Pitsker acknowledged, the best way to get a network on
track is to bring in outside help.

"Because we build our show network using industry experts, we are, in
effect, outsourcing the InteropNet," he said.

Outsourcing can save companies money, time, and headaches. When
implementing a new technology, "you could send your staff to a
three-day course, but they won't be able to do it successfully the first
time or over the weekend," Pitsker said.

Whether companies outsource or not, Pitsker recommended getting the
highest possible level of expertise. "[Then] you get to go on with your
first-time-right implementation."

To hot-stage or not to hot-stage ...

Pitsker also recommended hot-staging a network, as he does with the
InteropNet. Hot-staging allows IS managers to find problems before the
product hits their production networks.

Pitsker tells stories about "customers buying something like 150
bridges, then incrementally installing them on the production net.
They've never hot-staged anything -- this is not a formula for success."

Another lesson that the InteropNet team has learned is to power-up the
different technologies in stages. It's a controlled test, according to
Pitsker. If something fails, they turn it off and troubleshoot the problem.

This method also allows a network-management team to prioritize its
feature set and implement the most important features first. If a team
can't make it to the bottom of the list by deadline, the last remaining
features get cut.

"Then you can measure the post-deployment success," said Pitsker. "That
process shouldn't take years. Ideally, it shouldn't even take months."

Finally, Pitsker recommended conducting trials of new technologies. If
LAN managers plan to take quantum leaps over existing technologies,
"try before you buy," he warned. "But you have to buy that trial."

Justifying this sort of expense can cause political problems, he said.
Imagine telling a chief financial officer that "you want to buy an ATM
switch on this year's budget, but I'm not going to actually use it" for
another year or two, he added.

To avoid this hard sell, Pitsker recommended taking another approach.
"You don't go from driving a car to riding a motorcycle without taking
some lessons," he said, "and if you do, your fatality rate goes up."

This year's InteropNet is a good example of technology trials. Pitsker
has switched the backbone of the show entirely over to FDDI, after
several years of using a backup Ethernet circuit.

Now that FDDI is reliable, Pitsker claimed, he has introduced ATM into
the InteropNet. Either the ATM or FDDI network can carry the whole
backbone, and the OSPF interior gateway protocol switches between
them automatically. Pitsker said he tried to "break" InteropNet's ATM
before he had to rely on it for the show.

Type: Interview

Topic: NetWorld
InterOp
Trade Show
Network Architecture
Interview
Hardware Selection

Record# 15 350 518