Free Unix: Do You Get What You Pay For?
May 04, 1998
Noncommercial Unix systems like Linux have been stirring increased market interest as well as confusion. Their suitability is discussed, and recommendations are provided.
Hardware & Operating Systems: Unix and Midrange Server Technologies
How will Unix and midrange servers evolve during the next five years?
Strategic Planning Assumptions
Linux and other Unix freeware will establish pockets of strength in technically competent market niches (academia, application development, Web servers) (0.8 probability).
Unix systems at free or minimal charge will lack the performance tuning, scalability and hardware platform support to make them suitable for large commercial applications through 2002 (0.9 probability).
Linux will not displace mainstream commercial Unix versions from IBM, Hewlett-Packard, Sun Microsystems and The Santa Cruz Operation in the next five years as commercial Unix vendors shift focus to Windows NT (0.8 probability).
"Free Unix," with its open source code, will reinvigorate Unix interest, driven by widespread Web applications (0.4 probability).
The old adage "you get what you pay for" may need revision as the Web upsets the traditional paradigms of computing. Today, it is possible to configure a full Unix environment (kernel plus all programs, utilities, daemons and drivers) without incurring a license charge. The most popular "free" Unix environments are Linux, FreeBSD and the Apache Web server on Unix. "Free" (see Note 1) means downloadable source code via the Internet, or available on CDs (either free or with a minimal maintenance charge). There is a software community of thousands of developers, hackers and technical "gurus" who willingly volunteer time to update, patch and enhance public-domain Unix to promote the success of "open source code" Unix. Motives vary in nature: idealistic, adversarial (inhibit Microsoft's dominance) or recognition from the professional community. However, the question arises: Should commercial users, as typified by GartnerGroup clients, take advantage of these opportunities to cut costs and benefit from the collective knowledge of the community (rather than a single vendor), or will they wind up eventually "paying the piper"?
What Is Free Unix?
In the Unix culture, "free software" means source code available to share, reproduce and modify the program's source code. Free software has nothing to do with shareware programs, which are distributed without their source code. In 1983, Richard Stallman of the Massachusetts Institute of Technology proposed a legal document, called the GNU General Public License, creating the concept of "copyleft." This is a license granting full use, and the ability to modify software at will, as long as the original copyleft agreement is attached and no restrictions are placed on users of the modified version.
In this Research Note, we evaluate the pros and cons of installing noncommercialized Unix versions (e.g., FreeBSD and Linux, available off the Web, or versions from packagers and distributors with fees such as Red Hat, BSDI, Slackware and Caldera) compared with commercial varieties (e.g., HP-UX, IBM AIX). Decision criteria and recommendations follow.
1. Application Suitability: This factor applies to the role and function. Web servers, domain name servers, proxy servers, intranets with limited criticality, and legacy integration and application development platforms are the most suitable roles. Least suitable are roles in large and complex application workloads, databases, and data warehouses, especially ones in need of multiprocessing hardware, large storage management, significant tuning, wide-area connectivity and network management, and high-availability clustering.
2. Hardware Support: While Unix applications should be portable among hardware platforms, users will find mixed results obtaining performance and scalability, hardware support for processors (primarily available on Intel), driver support for peripherals, and distributed systems management. With hardware evolving to support larger workloads through SMP and NUMA enhancements, users can consider such operating systems as FreeBSD and Linux suitable for midrange SMP systems of up to four-way, but they will not be suitable for the high end through 2002 (0.8 probability).
3. Technical Support: Because Unix remains complicated to install and configure, users must ensure that technically proficient people are on staff to troubleshoot early deployments. Problems that commonly occur involve: plug-and-play glitches; difficulties in integration with corporate legacy applications; interoperability with existing network infrastructures, such as Novell NDS directory services and Windows and Macintosh client support; and lack of integration with system management frameworks (e.g., Hewlett-Packard's OpenView, Computer Associates International's Unicenter and Tivoli Systems' TME). Lack of support contracts may also limit it in government and defense.
4. Application Availability: Third-party support is currently restricted to nonmainstream suppliers with limited DBMSs, office software and vertical applications. DBMS vendors like Oracle, IBM, Sybase and Informix Software do not support nonbranded operating systems because of the costs. Middleware (e.g., ORBs and TP monitors) requires the user's initiative. However, Internet services, X Windows, E-mail, NFS, TCP/IP networking, Netscape Navigator and a host of other tools and plug-ins are readily available, reinforcing the role of free Unix for Internet servers.
Advice to Users Contemplating Free Unix
Red Hat Software (Linux) - free or $49.95
Walnut Creek CD-ROM (FreeBSD) - small media charge
Berkeley Software Design (BSD/OS) - $995 to $2,995
Caldera (OpenLinux Base) - $59
DBMS Database management system
NUMA Nonuniform memory access
OLTP Online transaction processing
ORB Object request broker
SMP Symmetric multiprocessing
TP Transaction processing
Bottom Line: Nonbranded versions of Unix environments, such as Linux and FreeBSD, can be suitable as inexpensive, functional Unix packages for Web servers, application development platforms, and workstation and dedicated application roles with qualified, in-house technical expertise, in place of commercially branded systems. However, these operating systems will not find widespread use in mainstream commercial applications in the next three years, nor will there be broad third-party application support (0.8 probability). Although interest is growing in their utilization as Web servers, their advantage will remain limited to moderately scalable applications (0.8 probability), and they will lack operating-system and performance tuning support for high levels of scalability (>500 concurrent OLTP users).
This document has been published by:
|Workgroup Systems Software and Hardware||4 May 1998||KA-03-2514|
|Unix & Midrange Strategies||4 May 1998||KA-03-2514|
Copyright 1997, 1998 by GartnerGroup. All rights reserved. Reproduction of this publication in any form without prior written permission is forbidden.