Archive for October, 2006

Ron Bodkin talks about automated Java web application troubleshooting using Glassbox at Google TechTalk (video) on September 25, 2006.

Interesting quotes:

  • No need to change or recompile the code
  • Low performance overhead about 1% when running (about 50% on code load)
  • Configurable thresholds
  • AJAX-based web interface
  • Monitoring multiple sources
  • Uses AspectJ, 10-year old stable technology
  • Extendable trackers through configuration or simple code
  • Preconfigured tracked actions:
    • Database connections
    • Database calls
    • Thread synchronization issues
    • Failures/slowness in web services, EJBs, other remote calls
    • Things that use too much CPI
    • Error pages and exceptions in Java processing
    • Time spent at a dispatch layer
  • Future features / plans:
    • Integration with existing monitoring infrastructures companies have built
    • Monitoring by specific user/session
    • Component owner-focused monitoring, e.g. Developer, Assembler, Admin (DBA/Sysadmin)
    • System training (by monitoring the process)
    • Historical views, comparison
    • Baseline comparison (system upgrades, new releases, test/production)
    • Configurable SLA (Service Level Agreement) reporting
    • Preconfigured monitors for different frameworks - more AJAX, more web frameworks, JNDI, cache, O/R mapping tools
    • System health
    • Clustering (autodiscovery, aggregation
    • Advanced analisys
  • Originally closed source. Now Open Source (LGPL). Contributors wanted.

Definitely interesting technology and quite logical extension to Virtual Machine kind of environment.

Comments No Comments »

Google TechTalks presentation by Stewart Smith, MySQL AB (video) on April 28, 2006

Interesting quotes:

  • Variable (really) size VARCHAR
  • Significantly different and faster index add/drop (coming in 5.1)
  • User defined cluster partitioning:
    • by key
    • by range (might not be in 5.1 but in future releases)
    • by list (future, not 5.1)
  • Replicate one cluster to another cluster (note node-to-node replication)
  • much more … really for MySQL powerusersDBAs and large setups.

Comments No Comments »