January 2015 Summaries

The blog post discusses the development and practical applications of a tool called "spectrogram," which visualizes real-time latency in operating systems using ANSI coloring in a terminal, allowing users to monitor system calls and their latencies. Developed during a weekend of experimentation with sysdig chisels, the spectrogram categorizes latency into buckets and uses color to represent the frequency of system calls, providing insights into system performance. The post details a practical use case involving a comparison of different AWS storage options on an EC2 instance—local SSD, magnetic EBS, and SSD EBS volumes—highlighting how each storage type handles varying loads and their impact on latency. The findings suggest that local SSDs may introduce unexpected latency under moderate loads, while remote SSD and magnetic options show similar performance under intensive workloads. The blog emphasizes the importance of measuring performance in real-world environments and encourages users to try sysdig spectrogram for better decision-making regarding storage options, while noting the availability of Sysdig Monitor and upcoming events for further exploration.

Sysdig 0.1.94 introduces a range of updates and improvements, including bug fixes and performance enhancements, alongside new features designed to bolster its capabilities. Notable changes include an increased sysdig-probe ring buffer size to 8 MB per CPU, improvements in libscap's live capture performance, and corrections of syntax errors in various chisels and tools. The release also adds new and updated chisels, such as the spectrogram chisel for real-time OS latency visualization, and updates filter fields and event parameters to improve accuracy and functionality. Community support and further resources are available through the sysdig mailing list and GitHub for bug reporting.