Every cluster builder wants to know how fast his or her computer is. After all, speed is the primary reason to build a Linux cluster — aside from the gains in data capacity and resource redundancy. But how do you measure speed? CPU clock speed is one thing, but using those cycles to actually get work done is quite another. And in a cluster, the network connections between nodes can quickly become a crippling bottleneck.
Every cluster builder wants to know how fast his or her computer is. After all, speed is the primary reason to build a Linux cluster — aside from the gains in data capacity and resource redundancy. But how do you measure speed? CPU clock speed is one thing, but using those cycles to actually get work done is quite another. And in a cluster, the network connections between nodes can quickly become a crippling bottleneck.
Theoretical peak performance is easy to calculate. However, this number is almost meaningless: it’s the unachievable upper bound of performance (measured in operations per second) for a given computer system. On distributed memory systems, like Beowulf clusters, few computational problems scale linearly, so most applications achieve a decreasing fraction of theoretical peak performance as the number of processors increases.
What’s needed is a standard benchmark for measuring system performance. Unfortunately, there are lots of “standard” benchmarks.
However, for high performance computing, a single package has been the gold standard for gauging achievable computational performance for many years: Linpack, a set of algorithms for solving linear algebra problems. Every hardware vendor quotes Linpack results. Great. So you run Linpack and compare your results against others’, right?
But hold on a minute! Who cares about how fast a computer system runs someone else’s software package? Unless you bought or built your machine to run Linpack day in and day out, your system’s performance on Linpack isn’t all that important. What is…
Please log in to view this content.
Not Yet a Member?
Register with LinuxMagazine.com and get free access to the entire archive, including:
