Each of the three types of network adapters that we tested had to supply UTP (unshielded twisted-pair) connectors (some also included BNC and AUI connectors) and directly support Novell NetWare 386 version 3.11.
Our NIC suite consisted of 18 data transfer tests that simulated a variety of network-usage patterns and load conditions. We also evaluated the design of each product, its drivers, and its documentation to determine our rankings.
PERFORMANCE
Our test-bed server was a Compaq Deskpro 66/M (see the illustration). The server's and NetFlex controller's memory creates a large buffer to minimize the server's effects on client/server performance. We designed all tests so that there was a minimum of disk usage on the server, since frequent disk access could lower throughput and mask the performance of the adapter being tested. Additio
nal test equipment included a diskless 286-based workstation with 1 MB of memory, to generate network traffic, and a DynaStar MiniHub.
We used NetWare 386 version 3.11 drivers supplied by each network-hardware vendor. We also used monolithic drivers for the 16-bit NICs if they were provided. We tested PCMCIA cards and portable adapters with ODI (Open Data-Link Interface) drivers, which reflects common usage. We tested the portable adapters in unidirectional mode and in the fastest mode that each one supported (i.e., bidirectional or EPP).
APPLICATION TESTS
We created the spreadsheet test by replicating the file-system access of Microsoft Excel 4.0. Each call, including the file, location in the file, and size of the operation, was re-created in detail for the reading and writing of small (20 KB) and large (100 KB) spreadsheets. We created the word processing test in a similar fashion by replicating the file-system operations of WordPerfect for Windows 5.1.
We designed two tests using
cc:Mail for Windows. Each is an actual cc:Mail application, which sends a message with an attached file to a specified user. One test attaches a small representative text file, while the other attaches a larger spreadsheet. Two similar tests run on DOS. The last Windows test was an application that reads and writes an indexed database.
Once we compiled our transaction-per-minute data, we computed a speed index relative to the performance of the 16-bit ISA-based National Semiconductor InfoMover NE2000plus.
The performance of the 16-bit ISA adapters cannot be directly compared to that of the PCMCIA and portable adapters because they are tested on different client workstations. Also, the NIC scores presented here aren't directly comparable to those from the August 1993 Lab Report because we used a different client workstation in our test-bed. This Lab Report also presents speed in a simplified index format rather than in transactions per minute.
USABILITY
We awarded points to products
according to the number of features that they provided from our list of essential features (see the text box "Hassle-Free NICs" on page 248). The portable adapters that received the highest features scores offered both bidirectional and EPP support, the ability to draw power from an external mouse or keyboard port, and a parallel port pass-through that enables you to simultaneously attach a printer and a NIC.
We assessed three areas for our ease-of-use ratings: configuration of the adapter, installation of the network drivers, and quality of the technical manual.
Contributors
Alan Joch, Senior Editor/BYTE, coordinates the combined testing between the BYTE Lab and NSTL.
Siva Kumar, Technical Analyst/NSTL, specializes in hardware and network-operating-systems testing.
Anthony J. Lennon, Project Manager/NSTL, evaluates servers and network hardware, as well as portable systems and peripherals.
The Lab Report is an ongoing collaborative project between BYTE magazine and Nationa
Performance Test
Windows 3.1
-- Spreadsheet
-- Wordprocessing
-- Database
-- E-Mail (small file)
-- E-Mail (large file)
MS-DOS 6.2
-- Low-level throughput
-- E-Mail (small file)
-- E-Mail (large file)
Illustration: Test-Bed Hardware and Software
How We Tested
