Content area
Full Text
While Microsoft has the edge over Linux as a server operating system for the enterprise, don't count the little guy out
Linux has made a name for itself as a formidable Internet application platform. But does Linux have what it takes to play against the big boys as an enterprise server operating system?
In a head-to-head comparison of Microsoft's Windows NT Server 4.0 and Caldera's OpenLinux 2.2, which is based on the most recent release of the Linux kernel, we sized up what each had to offer in the areas of per formance, manageability, connectivity, scalability and security.
We concluded that for small and mid-size networks, it's a close race because performance and administration features for the most part are equivalent. Both operating systems offer good extensibility and high availability, but need improvements regarding their ease of installation. Both can run a host of native applications. NT's applications tend to be better integrated, but Linux's are free.
However, NT's ability to host Microsoft's electronic commerce wares and transaction integrity products, such as Transaction Server and Message Queue Server, (which to date have no counterparts in the Linux world), make NT the clear choice for enterprise use.
What about Novell's NetWare? We're big NetWare fans, but because NT and Linux are better positioned as application servers, we looked at them head-to-head.
The basics: File and print performance
We tested each platform as a file server and as a File Transfer Protocol (FTP) server by running test scripts on a dozen clients that imposed heavy loads on the servers (see "How we did it," page 54).We ran the tests twice for each operating system - once using a Mylex RAID storage array and once without it.
The matchup between NT and Linux for serving up files resulted in a split decision. OpenLinux had faster disk support overall, but faltered when we tested it with a Mylex ExtremeRaid SCSI controller.The NT drivers for the controller seemed to have an edge over those for Linux.
The Ethernet connection, through a Compaq Netelligent Ethernet 10/100M bit/sec switch set to run at 100M bit/sec with no other traffic, was consistently saturated during the batch/script test.The connection was less saturated during the FTP put test. Ethernet collisions and retries seemed to...