Test Specs

We based our evaluation mostly on performance because fast network access is a big concern when choosing a remote-access server. We also, however, placed a strong weighting on usability and features, and then we looked at technology and value. Our best-overall rating formula breaks down this way: technology (10 percent), implementation (40 percent, which is further divided into 40 percent for usability, 40 percent for features, and 20 percent for value), and performance (50 percent).

The performance test suite simulates from one to 16 remote users attempting to dial into a centralized LAN through the servers. To simulate multiple remote-access sessions, we had eight Windows 95-based Dell Dimension XPS P75 PCs dialing into the servers using Hayes Microcomputer Products' ((770) 840-9200) 28.8-Kbps Optima V.34/V.CF +fax +voice modems. The serve rs would then route the c alls to a NetWare-based file server. The client-side systems were configured with 16 MB of RAM. We used telephone-line simulators from Northeast Innovations ((603) 229-0500) to create dial tones.

NSTL's InterMark file-access tests force the servers to seek files in the following three file directories:

--
 Applications -- This directory has read-only access, and we use it the most in the tests. For this reason, it tends to remain in a remote-access server's cache memory.

--
 Data -- This directory is much longer than the applications directory, so its individual files aren't used as frequently. Typically, a server doesn't cache these files.

--
 Personal -- This directory simulates where you must go to access a personal file; the primary traffic to this directory is to write files. The tests in this directory also accommodate print traffic, which goes into a server's print queue.

To round out the evaluation, we rated usability and features. We checked out how easy the routers are to set up and use, and we weighted important features like customer support and warranty length.

---