NonIntrusive RealTime Instrumentation for Distributed Systems and Enterprise Applications Prepared f

1
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsPrepared forBorland
ConferencePresented byJorly Metzger,
703.368.6107x235, jorly.metzger_at_progeny.netProdu
ct ManagerGary Sikora, 703.795.4684,
gary.sikora_at_progeny.netSales ManagerStewart
Bittel, 703.772.3512, stewart.bittle_at_progeny.net
14 September 2004Document P000XXX, Rev -
TM
2
Agenda

• Background
• What is CodeEyes
• How CodeEyes Works
• How Variables are Selected
• The Visualizations Views
• The Data Buffer
• CodeEyes Benefits
• Customers
• Questions

3
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsBackground
4
Background Instrumentation Approach

• Typical approach
• Application code modification
• Execution environment interruption
• CodeEyes approach
• Accessing memory via the Operating System (OS)
  independent of the application

5
Background Application Code Approach

• Methods
• System outputs to the console window
• Embedding instrumentation code
• Special interfaces
• Result
• Special code affects the execution and requires a
  recompile between iterations
• Consequently increasing development costs

Coded System Outputs
Displayed System Outputs
6
Background Execution Environment Approach

• Methods
• Breakpoint to view variables
• Single-step and jump through code
• Integrated Development Environment provided
• Result
• Extremely powerful in simple, single-threaded,
  isolated applications
• Unusable when coupled threads and applications
  continue execution while a selected one was halted

Breakpoint
Variable Watch
7
Background Memory Space Approach

• Methods
• Accessing memory via the OS independent of the
  application
• Buffer data for post analysis
• Monitor and control data real-time
• Result
• Does not effect the application code or execution
  environment
• Powerful for coupled threaded and multi process
  system and enterprise applications

Project Compiler
Native Symbol File
Graphical
Memory
Project Source Code
Command Line
or
Buffer
Special Compiler
Special Symbol File
8
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsWhat is CodeEyes
9
What is CodeEyes Overview

• An open architecture software instrumentation
  framework built for complex, distributed,
  heterogeneous, real-time applications found in
  todays systems and enterprises

• Enabled for the first time are synchronous views
  of application data distributed across .Net,
  Java, Web Services, CORBA and legacy
  architectures
• Non-intrusive approach The applications code
  or execution environment are not modified or
  interrupted when instrumentation data is accessed

10
What is CodeEyes Usage

• Development
• Real-time, synchronized access to software
  variables across disparate, distributed
  applications decreases development time and
  increases product quality
• Testing
• Complex measurement and recording of product
  performance and functionality provides increased
  test repeatability and thoroughness, and
  streamlines the certification process
• Deployment
• Non-intrusive monitoring of system or enterprise
  applications provides invaluable insight that
  helps avoid failures while the product is in
  production

Support many phases of product life-cycle not
just development
11
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsHow CodeEyes Works
12
How CodeEyes Works Symbol Files

• Natively produced symbol files are used to map
  application variables to memory locations
• Symbol files produced by compilers
• No special compilation is required
• Instrumentation synchronization
• Variables originating from different applications
  with different native symbol file formats can
  coexist on the same graph
• The current symbol file formats

13
How CodeEyes Works Synchronized Example

• The first application is written in Visual Basic
  and runs in a .Net framework
• The other is written in C and runs in an
  embedded system
• The associated symbol for each, PE/COFF and
  Elf/DWARF 2 are used
• Channels are setup to the variable via agents
  running on the host
• Both variables, A and B, are synchronized in the
  presentation layer views

14
How CodeEyes Works Architecture

• The architecture is comprised of four components
  three that make up the Model View Controller
  (MVC) and a fourth Agent component which is
  responsible for accessing memory locations
• Agent interface implements a Common API which
  allows new Agents to be developed, extending the
  instrumentation capability without affecting the
  core MVC components
• New Agents can be developed for additional OSs,
  embedded systems, even for non-memory sources,
  such as databases and hardware

15
How CodeEyes Works Architecture

• The architecture is comprised of four components
  three that make up the Model View Controller
  (MVC) and a fourth Agent component which is
  responsible for accessing memory locations
• Agent interface implements a Common API which
  allows new Agents to be developed, extending the
  instrumentation capability without affecting the
  core MVC components
• New Agents can be developed for additional OSs,
  embedded systems, even for non-memory sources,
  such as databases and hardware

16
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsHow Variables are Selected
17
How Variables are Selected Watch Profiler

• Select variables across applications and nodes to
  control and monitor
• Drill-down across classes and methods
• Watch profiles are displayed in tabbed panes in a
  tree and tabular hybrid structure

18
How Variables are Selected Watch Profiler
(cont.)

• Menu selected applications and hosting node
  agents
• Watch profiles can be saved and opened to enable
  testing repeatability and support formal software
  certification and maintenance activities
• A hybrid tree and tabular variable representation
  intuitively displays complex aggregate data
  structures and their associated values and
  attributes
• Values can be displayed in various engineering
  units to ease measurement relationship to other
  artifacts of a software applications life-cycle
• Variables can be modified in real-time, providing
  the ability to apply what-if scenarios to see
  how the change affects the rest of the system or
  enterprise

19
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsThe Visualization Views
20
The Visualization Views Types

• There are three visualization views
• Strip Chart Temporal graph of variables
• Spectral Spectral analysis of variables
• Statistical Statistical analysis of variables
• The open architecture is extensible so new views
  can be added

21
The Visualization Views Common Features

• Simultaneously graph multiple variables with each
  plot differentiated by color
• Multiple graph instances are used to organize
  related data and provide swift navigation between
  groupings
• Scroll bar controls provide stepped zoom in/out
  control with skew offset control, a magnitude
  optimization control based on variable values,
  and a temporal slide view window
• Run controls are provided to reset, start, pause
  and resume the graph

22
The Visualization Views Strip Chart

• Cursors provide the capability to measure
  variable magnitude and temporal values and delta
  values between two points
• Scroll bars provide a capability to track new
  data or review past data while the graph is
  running
• Graphed variables temporal values can be exported
  to the data buffer for post analysis

23
The Visualization Views Spectral

• Cursors provide the capability to measure
  spectral magnitude and frequency values, and
  delta values between two points
• Scroll bars provide a capability to optimize the
  frequency axis based on cursor position or
  maximum magnitude
• FFT window function and number of points can be
  applied in real-time

• Averaging and persistence controls for a trace
  history view
• Magnitude axis can be set to absolute or relative
  with logarithmic or inverse scale
• Sampling window can be set to a free running
  sliding window or sequenced with percent overlap

24
The Visualization Views Statistical

• Signal mean and standard deviation are
  continuously calculated and displayed
• Truth can be a static value or a reference to a
  real-time variable

• Bias error, mean, standard deviation and RMS
  error are continuously calculated and displayed
• Every measurement can selected to be displayed in
  the waveform
• A signal distribution thumbnail graph provides a
  quick view

25
The Visualization Views Statistical

• Signal mean and standard deviation are
  continuously calculated and displayed
• Truth can be a static value or a reference to a
  real-time variable

• Bias error, mean, standard deviation and RMS
  error are continuously calculated and displayed
• Every measurement can selected to be displayed in
  the waveform
• A signal distribution thumbnail graph provides a
  quick view

26
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsThe Data Buffer
27
The Data Buffer Capability

• Record variables free-running or based on an
  event
• Recorded data can be viewed in tabular form,
  exported to spreadsheet standard formats, or
  replayed through a visualization view

28
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsCodeEyes Benefits
29
CodeEyes Benefits

• Maximizes Return-on-Investment (ROI) through
  reduced specialized code or execution
  interruption during the development,
  certification and maintenance stages of product
  life-cycle
• Real-time debug with variable watches and
  modifications reduces coding and compilation
  iterations
• Visualization results can negate the need for
  specialized code to support product certification
  and maintenance
• Real-time synchronized visualization of variables
  across system and enterprise distributed
  applications
• Multiple simultaneous visualization views
  including strip chart, spectral and statistical
• Data buffer that records data for playback and
  export to spreadsheet standard outputs
• Extensible open architecture enables integration
  of new data sources, including non-application
  data sources such as hardware and network
  interfaces, and integration of new views

30
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsCustomers

• Joint Strike Fighter
• Northrop Grumman Radar Group
• Northrop Grumman Distributed Aperture (DAS)
  System Groups
• Lockheed Martin Software Development Station (SDS)

31
Non-Intrusive Real-Time Instrumentation
forDistributed Systems and Enterprise
ApplicationsQuestions
Come See Us