Networks - Bioinformatics Computing - page 89

results of any analysis would have to be separately archived to a floppy, Zip ® disk, or CD-ROM. In
addition, sharing experimental data would require burning a CD-ROM or using other media
compatible with the other workstations in the laboratory. Simply attaching a data file to an e-mail
message or storing it in a shared or open folder on the server would be out of the question. Data
could also be shared through printouts, but because the computers aren't part of a network, each
workstation requires its own printer, plotter, modem, flatbed scanner, or other peripherals. For
example, unless the expression analysis workstation has its own connection to the Internet, results of
the experiment can't be easily communicated to collaborating laboratories or even the department in
an adjoining building. Furthermore, even though many of the public online bioinformatics databases
accept submissions on floppy or other media, the practice is usually frowned upon in favor of
electronic submission.

Without the wireless component of the LAN, researchers in the lab would not be able to instantly
explore the data generated by the scanning and analysis workstation, but would have to wait until
the other researchers operating a workstation have time to write the data to a disk or other media.
More importantly, every workstation operator would be responsible for backing up and archiving their
own data—a time-consuming, high-risk proposition. It's far more likely, for example, that a
researcher in the laboratory will fail to manually archive local data on a regular basis than it is for a
central, automated backup system to fail.

This brief tour of this prototypical microarray laboratory highlights several applications of networks in
bioinformatics. The underlying advantage of the network is the ability to move data from one
computer to another as quickly, transparently, and securely as possible. This entails accessing online
databases, publishing findings, communicating via e-mail, working with other researchers through
integrated networked applications known as groupware, and downloading applications and large data
sets from online sources via file transfer protocol (FTP) and other methods.

Although many of these features can be had by simply plugging in a few network cards and following
a handful of instruction manuals, chances are that several key functions won't be available without
considerably more knowledge of network technology. For example, selecting and configuring a
network requires that someone make educated decisions regarding bandwidth, reliability, security,
and cost. Furthermore, mixed operating system environments typical of bioinformatics laboratories,
which tend to have at least one workstation running Linux or UNIX, presents challenges not found in
generic office networks.

What's more, it may not be obvious from the simple network depicted in Figure 3-1 that
bioinformatics networks present unique networking challenges that typically can't be addressed by
generic network installations. The first is that there is a huge amount of data involved. The network
isn't handling short e-mail messages typical of the corporate environment, but massive sequence
strings, images, and other data. In addition, unlike networks that support traditional business
transaction processing, data are continually flowing from disk arrays, servers, and other sources to
computers for processing because the data can't fit into computer RAM. As a result, the network and
external data sources are in effect extensions of the computer bus, and the performance of the
network limits the overall performance of the system. It doesn't matter whether the computer
processor is capable of processing several hundred million operations per second if the network
feeding data from the disks to the computer has a throughput of only 4-5 Mbps.

This chapter continues the exploration of the Internet, intranets, wireless systems, and other network
technologies that apply directly to sharing, manipulating, and archiving sequence data and other
bioinformatics information. The following sections explore network architecture—how a network is
designed, how the components on the system are connected to the network, and how the
components interact with each other. As illustrated in Figure 3-2 , this includes examining networks
from the perspective of:

Geographical scope
l

l

Underlying model or models used to implement the network
l

Signal transmission technology
l Bandwidth or speed

l Physical layout or topology