InfoQ Homepage Presentations Tardis: Affordable Time-Travel Debugging in Managed Runtimes
Tardis: Affordable Time-Travel Debugging in Managed Runtimes
16:39
Summary
Earl Barr, Mark Marron discuss building time-travel debuggers for managed languages, implemented with Tardis, and enabling developers to investigate what happened prior hitting a bug.
Bio
Earl Barr is a member of the System Software Engineering Group and the Centre for Research on Evolution, Search and Testing (CREST) at the University College London. Mark Marron is Senior Research Software Development Engineer @ Microsoft Research (Redmond Lab) in the RiSE Group.
About the conference
SPLASH — the ACM SIGPLAN conference on Systems, Programming, Languages and Applications: Software for Humanity — embraces all aspects of programming, languages, and software construction and delivery. It incorporates OOPSLA, Onward!, the Dynamic Languages Symposium, and a host of workshops, panels, tutorials, demonstrations and invited talks.
Feb 04, 2015
-
-
From Docker to Kubernetes: Container Networking 101 (By O’Reilly)
Related Sponsored Content
-
-
Related Sponsor
Community comments
the focus should be on behavior and less so on state (heap)
by William Louth /
the focus should be on behavior and less so on state (heap)
by William Louth /
Your message is awaiting moderation. Thank you for participating in the discussion.
I've never used a traditional debugger because when I started out developing profilers it was not possible to run multiple agents within the same process but from looking over the shoulders of others I rarely see heaps as low as 50MB within a managed runtime and this was positioned as some kind of high average in the presentation.
The problem with debuggers as well as profilers is that they view all data and behavior with equal measure when in reality tools as well as people can only observe and understand the essence of each. Less is more but you have to add smarts into the runtime and tooling as piggybacking on the runtime will only get you so far. Before attempting to record software create an agent that is able to determine what it is that makes a software truly tick (perform) in the process eliminating noise. This is the approach taken with Simz and Stenos via the Satoris adaptive metering runtime.
When you have this time travel is just one of many amazing steps a technology can take.
A Time Lord for the Java and JVM Universe
www.autoletics.com/posts/a-time-lord-for-the-ja...
Software Regulators! Mirror Outwards, Simulate Inwards.
www.autoletics.com/posts/software-regulators-mi...
Beyond Metrics and Logging with Metered Software Memories
www.autoletics.com/posts/beyond-metrics-logging...