Linux & Open Source

Our clients demand and we believe that open-source is the way forward.  Indeed, our GRIDBYTE™ founders have participated in open-source projects dating back to the 1980’s.  We see open-source as the way forward because a high percentage of our growth is related to optimization of open-source-based compute infrastructures.  However, open-source also has significant room for improvement; i.e., R&D.  In many environments, we find that the open-source tools and operating systems are poorly optimized for performance.  The reasons for poor optimization far exceed configuration details.  Open source also has high support costs especially when the components required are not in popular distributions.

       The lack of optimization, judging from data on several projects completed by GRIDBYTE™, can account for up to 50% of uncaptured performance.  Said differently, some businesses lose 50 cents on every compute infrastructure dollar spent due to lack of OS tuning, lack of software (application and virtualization) tuning, and a lack of mission-mode tuning involving the actual productivity applications.

       Considering the open-source flagship, Linux continues to suffer from many of the same problems as it did in the early 1990’s.  With desktop/laptop computers, we still struggle with the same key issues every time:

       Video drivers: Are they available and do they work as well as the Windows® Drivers?  Do they drive internal LCDs at maximum resolution; e.g., 1280×800 (widescreen on Sony VGN T250P notebook)?  Can they simultaneously drive external monitors at maximum resolution; e.g., 1600×1200?  Do we have all of the video switching options; i.e, internal LCD + external display, internal LCD only or external LCD display only?  Can the video configuration be changed without restarting the X-server or, worse, a reboot?

       WiFi drivers: Do we have the right drivers for respective hardware?  Do they work reliably?

       NIC drivers: Do they work?  Do they work well?

       Audio drivers: Do they work?  Do they work well?

       Does the performance solution for a given set of problems require yet another custom-compiled kernel?  Does the new kernel, introduce compatibility problems in other areas?

       The reasons for these shortcomings are long and varied.  The popular one is that open-source must necessarily lag behind cutting-edge technologies because vendors do not distribute resources to enable the community to create drivers early.  However there is more to it—and we, as a community, must move beyond excuses if Linux is to be a serious contender and not just an alternative.

       Beyond price (open-source is “free”) and lifestyle (an alternative to proprietary software), open-source allows us to innovate freely, contribute to the technical community and create companies like Redhat.  Open-source continues to levelize the competitive space by forcing proprietary systems to open-up; e.g., Solaris.

       Open-source must also deal with high “hidden” support costs.  With respect to operating systems, we cannot simply assume that open-source Linux will work well in the wide array of compute environments.  Open-source requires higher degrees of specialization to optimize various layers of computing software and hardware.

       As we show in different case studies, open-source sometimes encourages decision-makers to select poor compute infrastructure choices because the strategic cost of an open-source based project can be several times higher than proprietary software solutions.  The cost is high because there is often no single-point of responsibility for glitches and problems; thus many project teams routinely find major bugs or inadequacies for which they must create custom solutions.  In our experience at GridByte™, these realities divert creative energies from the core mission of many business process projects.