Linux Replaces Unix as the Operating System of Choice for Supercomputers

In the latest TOP500 ranking of supercomputers, Linux-based systems occupy 476 spots, representing a staggering 95.2%. While Linux now dominates the supercomputing world, this wasn't always the case since its inception.

In June 1993, the first-ever global TOP500 list of supercomputers was released, around the same time that Linux was just beginning to gain traction. In fact, it wasn’t until 1993 that the first successful Linux distributions—Slackware Linux and Debian Linux—were introduced.

Subsequently, as a forthcoming report from the Linux Foundation highlights, analyzing 20 years of data from the TOP500 organization regarding computational performance shows that since Linux first appeared on the TOP500 list in 1998, it has consistently ranked among the top ten operating systems for supercomputers over the past decade. Moreover, starting from June 2010, Linux-based supercomputers have accounted for more than 90% of the TOP500 list.

Prior to Linux's dominance in the supercomputing realm, Unix was the primary operating system used in supercomputers. However, starting in 2003, there was a significant shift: Unix systems, which once occupied 96% of the TOP500 list, were replaced by Linux systems occupying the same share. By 2004, Linux had fully taken over the supercomputing domain.

According to the Linux Foundation, "Linux has become the driving force behind breaking computational performance barriers and has spurred innovation in computing research and technology. In other words, Linux has successfully claimed its position as the dominant operating system in the supercomputing field, with most supercomputers using Linux. It effectively assists researchers in overcoming and resolving computational limitations and bottlenecks."

The foundation attributes this phenomenon primarily to two reasons. First, because most of the world's top supercomputers are superscalar architecture research machines designed primarily to handle specialized tasks, each supercomputer is an independent project with unique characteristics and optimization needs. Developing a custom operating system for each system is impractical. However, with Linux, research teams can easily modify and optimize it once, enabling breakthrough designs suited to modern supercomputers.

Another critical factor is that "for a customized, self-supported Linux distribution, the licensing cost remains the same whether you're using 20 nodes or 20 million nodes." Additionally, "by leveraging the vast open-source Linux community, all your projects can receive free technical support and development resources, thereby balancing or even reducing the costs associated with developers and operating system expenses."

As a result, supercomputers continue to grow faster. From the total computational peak perspective, based on the Linpack benchmark, the maximum computational performance of a supercomputer roughly doubles every 14 months, far surpassing Moore's Law (which states that the number of transistors embedded in chips will approximately double every 24 months). In high-end products, supercomputers operate even faster. For instance, the fastest supercomputer on the TOP500 list has doubled its peak performance, increasing from the CM-5’s 59.7 gigaflops in 1993 to 33.86 petaflops in 2013, matching the speed of Tianhe-2 today.

Therefore, the Linux Foundation concludes: "By separating computational peaks and operating systems through 20 years of global TOP500 data, it becomes clear that Linux not only powers the majority of today's supercomputers but also significantly contributed to the disproportionate growth in supercomputing performance over the last decade. Moving forward, Linux will continue to drive progress and innovation in the computational field, helping to explore the mysteries of the universe and solve some of the toughest real-world problems."

We cannot help but wholeheartedly agree.