Getting the most from your cores

Strong Core


CPU utilization charts can be very useful because they visually indicate how heavily a core is utilized. If the utilization is fairly high (close to 100%) but then drops low for a noticeable period of time, something is causing the core to become idle. The reasons for CPU utilization to drop could be from waiting on I/O (reads or writes) or because of network traffic from one node to another (possibly MPI communication).

In a general sense, CPU utilization provides an idea of how well an application is performing and if it is using the cores as it should. Remember, the first two letters in HPC stand for "High Performance."

The Author

Jeff Layton has been in the HPC business for almost 25 years (starting when he was 4 years old). He can be found lounging around at a nearby Frys enjoying the coffee and waiting for sales.

Buy this article as PDF

Express-Checkout as PDF
Price $2.95
(incl. VAT)

Buy ADMIN Magazine

Get it on Google Play

US / Canada

Get it on Google Play

UK / Australia

Related content

  • Determining CPU Utilization

    CPU utilization metrics tell you how well your applications are using your processing resources.

  • Processor and Memory Metrics

    One goal of HPC administration is effective monitoring of clusters. In this article, we talk about writing code that measures processor and memory metrics on each node.

  • Monitoring Storage with iostat

    One tool you can use to monitor the performance of storage devices is iostat . In this article, we talk a bit about iostat, introduce a Python script that takes iostat data and creates an HTML report with charts, and look at a simple example of using iostat to examine storage device behavior while running IOzone.

  • Top Top-Like Tools
    Admins solve problems ranging from slow servers to failing applications. The first tool I reach for when I need to check on a server with shell access is Top.
  • Process, Network, and Disk Metrics

    In the continuing story of monitoring HPC systems, we look at code that measures process, network, and disk metrics.

comments powered by Disqus