• Multi-purpose
  MoSKito can collect data about almost everything: Threads. Memory. Caches. Storages. Services. Averages. ExecutionTime. RequestCount. CPU Usage. IOStats. Registrations. Payments. Conversion. Partner Performance. Online Users. ScamScores. Load distribution. User-dependent load (guest/member). Call traversal. Path-dependent measurement. ... whatever you need!
• Non-Invasive
  
  • Plug-able in any application
  • Plug-able via Filters, Webapp-loaders
  • Annotations, Proxies, Wrappers, AOP or direct calls
  • Measures ‘around’ the measured code, without changing it.
• Interval based
  
  • The behavior of a system depends on hour, weekday, weather, holidays and good karma.
  • Large amounts of collected data make monitoring nonsensitive to anomalies.
  • Inspection of short intervals offers more understanding about system’s behavior
  
• Monitoring System Kit
  
  • Provides the infrastructure to collect stats.
  • Useable to build up whatever monitoring system required.

Collect - Store - Analyze.

• Collect
  
  • Support for different types of collectable data (request-oriented, size-oriented, business data etc).
  • Further types plug-able.
  • Integration via proxies, annotations, inheritance, aop, class loader or plain calling.
• Store
  
  • Remote storage with MoSKito-Central in a storage system of your choice.
  • Local storage with log appenders.
  • Attachable snapshot archivers.
  • Direct XML / JSON / CSV feed from the WebUI.
• Analyze
  
  • Instant analysis of a web-application with the MoSKito WebUI.
  • Service load analysis with logs.
  • Post-Release analysis with snapshots (each monday 10 p.m.)
  • Call flow analysis with monitoring sessions and use-cases.

• Memory leak detection in storages (maps etc) with put/get ratios, efficiency checks with get/ put/remove/miss ratios.
• Cache/Proxies hit rates in warmed and cold states.
• Error monitoring, which components produces (handled or unhandled) errors
• Concurrent load on components (number of parallel requests). Lock contention optimization
• Capacity planing aka What do I have to buy to handle 5.000 more concurrent users?
  
  • Memory per user.
  • Requests per user.
  • CPU Time per user.
  • Separation by domain (multiple sites), guest/ member/paying member traffic, male/female traffic.
• Post release analysis - what have changed?
• Network traffic optimization.
• Call trace over multiple VMs.

Whats in the Pipe?

• Server side stats cumulation
  To save RAM MoSKito only holds the last (and the current) value of a stat object in memory. The WebUI has no server side support for time-based analysis. Instead the user had to rely on stat-archiving, moskito-central or archivers to accumulate cumulated stats over a time period (like last 20 5m values and so on).
  Now, large MoSKito installations have thousands of producers and ten thousands of stats and keep them all over a longer period of time will likely blow up any RAM limits, but the demand for graphs, and therefore for chrono-data was so huge, that we decided to go with a compromise.
  We allow the user to specify in a special config file, or on site, which values of which producers should be cumulated and how long. This way we don't blow up memory but allow the user to get graph data for at least most needed stats.
• Reverse call-tree tracking
  MoSKito can show you how much time has been spent in a method of a class, how often it got called, how many errors it produced, but not who was calling it. Of course you can manually start a monitoring session and try to find out a path to your class. You will never get all. Probably. Reverse call-tree tracking tries to address this issue. If turned on (and it has to be turned on manually) it will track all incoming calls, and as soon as it detects that a call has passed a monitoring point that should be watched, it will record the call.
  At the end of the day you will find all (at least all actively used by users) pathes in your application which lead to a call on a method.
  We do not know yet for sure, but we assume that this feature will come with a performance penalty if enabled (hence manual turning on and off), but we believe it's worth the penalty. Especially for systems you are not THAT familiar with (yet).
• ... more