Wily Introscope APM
Wily Introscope APM
Robert Dring
Technical Consultant EMEA
Rega, Latvia
WebSphere Nordic User Group
22nd September 2006
eMail : [email protected]
Tel : +44-7917-046945
Agenda
• Who is Wily?
• Product Overview
• Introscope In Depth
— Architecture / How / What
• A new division of CA
• Wily manages more Java applications than all of our competitors put
together
— Leading contributors to World-wide Java forum
• Leaders in the Gartner Group “Magic Quadrant”
• Certified:
• Corporate relationships:
Siebel
Network Load
Firewall Balancer Portal Mainframe
Router Database
Switch
Web
Applications
End User Servers
Web Services
Introscope Databases
• 24 x 7 monitoring
• Rapid triage
Real-time view of
Application performance
Early indications of a
problem
Online Purchasing
Additional Components
• Web Services
• MQ Series / Tuxedo
• Portals (WPS) 5.02 and above.
• Integration (WBI, WLI, SAP Enterprise Portal)
• Web Servers (Apache, IIS, etc.)
• And others including: (z/OS, Oracle Databases, etc.)
Database
App Server
To solve this… Web
Configuration?
Too Many Application
EJB Calls? Server Is it JDBC?
HTTP Server
Connectors…
Workflow…
Mainframe
Authentication
Inside-to-Outside
Transaction MQ Series
Request
Third
Parties
Response
Content
Web Services
APPLICATION SERVER INTEGRATION
JVM Databases
Introscope Probes
Inserted
CLASS
AutoProbe Application 1010010
Integration
with
Object Code 1111010
1010010
0101010
JVM ClassLoader (.jar files, etc.) 0101000
10110
Introscope
Agents
Enterprise Portal
Alerts, Events &
Agen
Performance Data
t
Portal Server
JVM
Agen
App Health, SLA, t
Six Sigma, Non-
Java
Business Impact Historical Systems
and Capacity Data like MQ
Planning reports Series
Introscope
EPA
• Advanced Metrics
— Concurrency
— Stalled Methods
— Instance Counts
Connectors…
Workflow…
Mainframe
Authentication Top-to-Bottom
End-to-End
Transaction MQ Series
Request
Third
Parties
Response
Content
Web Services
APPLICATION SERVER INTEGRATION
JVM Databases
Slow transaction
Oracle specific
data
View detailed
information about
errors as soon as
they happen and
improve end user
experience
You need the ability to detect and diagnose the causes of serious errors,
which can prevent individual users from completing web transactions
Copyright © 2006 CA and Wily Technology, Inc. All Rights Reserved. 22
Analyse Problematic Transactions
- Transaction Tracer
If App’s Average Response • Isolate problematic individual transactions
for Most Users is Fine … • Analyze and pinpoint bottleneck for each
• Improve overall app performance
• From Browser to Back-end
HELP
HELP
Records 100% of performance data for problem solving, trend analysis and capacity planning
Copyright © 2006 CA and Wily Technology, Inc. All Rights Reserved. 26
IT to Business…
Technical Data & Business Views
Connectors…
Workflow…
Mainframe
Authentication
Transaction MQ Series
Request
Third
Parties
Response
Content
Web Services
APPLICATION SERVER INTEGRATION
JVM Databases
IT to Business
Copyright © 2006 CA and Wily Technology, Inc. All Rights Reserved. 27
Alerts and Framework Integration Aid in
Collaborative Application Management
• Flexible/Powerful Event Manager
Console Alert Message
Alert Emails
Events to Systems Framework
Console
Shell Script Actions
Any combination of Events for either
Threshold
Connectors…
Workflow…
Mainframe
Authentication
Transaction MQ Series
Request
Third
Parties
Response
Content
Web Services
APPLICATION SERVER INTEGRATION
JVM Databases
Development
Staging
QA Production Hosting
Introscope
Explorer &
Web or Swing Alerts, Historical Scope &
Customizable Blame System
client Data & Reports Control
Dashboards
Monitor Improve Patented Agent Highly
JMX & PMI AutoProbe Transaction
Availability Technology scaleable,
Support integration Tracer
(No coding) Low overhead
Introscope Extensions
Control Performance Brower Response
Leak Hunter SNMP Adapter
Time Adapter
C
3o
om
Authentication CICS
Connector
C
3o
om
Authorization
MQ /
Portlet
Servlet EJB Tuxedo
Portlets Container Connector
Building Models
C
3o
om
JDBC
Driver
Page Loading
Page Rendering J2EE APPLICATION
Engine
Thread Pool EJB Pools JDBC Pools
C
3o
om
3Com
C
3o
om
C
3o
om
C
3o
om
Centralized C
3o
om
Point for
Accessing
Applications
• Long delays in
addressing poor portal
Low Portal
performances results in a
Poor Portal decline in portal usage
Adoption
Performance
Servlet EJB
1 3 4
JSP JDBC
SQL MQ/CICS
LDAP
Security
User Profile Portal
Database
Page Loading Page Rendering
Database
Database Application
Server
HR Siebel PSFT SAP
Servlet EJB
JSP JDBC
Class JVM
Portlet Container
Model J2EE
Authentication
Building
Authorization SAP HR
People
Soft
J2EE
Siebel Services Appl
Portal User MQ /
Portlet
Portal Framework
Servlet EJB Tuxedo
Portlets Container Connector Mainframe
• Java Components – monitoring Java component real time including Servlets, JSP’s, EJB’s etc.
• Gateway Servlets – The entry point of all incoming transactions.
• Authentication – Response times to logins, stalled login requests including WPS authentication feature.
• Authorisation/Access Control – Components that authenticate users and roles including back end
systems such as LDAP.
• Portlets – Determine which individual portlets are causing a problem, including postback data and
rendering.
• WebSphere Portal Engine – Responsible for building models, loading pages and rendering.
• Portlet Container – Managing portlets and performing portlet actions, stalls here caused by particular
portlets.
• Portlet Services – The Credential Vault and Content Access services are provided here. If content is
slow then specific URL’s must be identified.
• Back-End System Connectors – There are many disparate systems feeding the portal such as
databases, mainframes, MQ-Series etc. Isolation of problems from these external sources is critical.
E-mail
CRM HR Services News Feed
ERP Mainframe
Portlets
Portlet
Engine
Gateway
Authentication Servlet Container Portal & Application
Transaction User Login Model Databases
Request Building Portlet Perform
Portal Action
Users Authorization Page
Portlet
Load User & Roles Loading
Services Back-end Servers
Load User Group
Credential Vault
Load Resource Page
Rendering Content Access LDAP
Page Results
Help! ?
• Quickly isolate problems in
? individual portlets
CRM
? HR
? ? Custom
News
• Correlate performance to
? J2EE
connected systems
?
I can’t access my pension ? • Improve control over portal
benefits on the portal. ERP ? services delivery
SAP ? email
Manageability of
the entire portal Problem isolation of
workflow. individual portlets.
Router
Firewall Load
Switch
Network Balancer Web Database
Tier Mainframe
Tier Application Tier
Tier
Customers Mirrored Port for XX Application Web
Production VLAN (i.e. Cisco SPAN Port)
Out-of-path
Monitor Interface
No IP Address
Management Interface
Requires IP Address
Management
VLAN
TIM
Transaction Impact
Monitor (TIM)
Management Interface
Requires IP Address
Bob Wright
Process Order Out-of-path
Slow Transaction Time = 22.8s Monitor Interface
06.01.05 8:59:48am No IP Address
CEM Incident:
CEM • Process Order
• Slow Time > 5.0s
• 256 Customers, 22 Premier
Identifies and Measures Customer Transactions • Time: 06.01.05 8:59am
• Duration: 35minutes
Behavior and Content Defect Detection • Incident cost: $278,00
Evidence Collection:
• Network, Web, App, DB,MF
Correlation of Defects into Incidents
Business Prioritization and $$ Impact Analysis
Evidence Collection of Infrastructure State
(performance, utilization, configuration, logs)
Incident Forwarding into Existing Systems and Procedures
Unique Customer Transaction Repository
(Current and Historical Reporting, Scorecards, Analytics)
Copyright © 2006 CA and Wily Technology, Inc. All Rights Reserved. 52
CEM - Features
Open API
• Simple deployment and operation CE
M
– Appliance-like installation
CEM Platform
– One-time setup for universal visibility (Transaction Impact AnalysisTM)
NOTE: In most J2EE applications a Business Transition and a Transaction are the same
• CEM stores copies of the web server SSL keys more securely than web
servers
– No IP address on the monitored network
– Encrypted using AES
– Does NOT terminate sessions
• Prioritisation of incidents
• SLA compliance
Black Box
Proactively Monitors Detects Problem
Production Applications and Notifies
Any Questions ?
Introscope's architecture supports seamless integration with existing IT infrastructure through the use of open-standard approaches. Its compatibility with major application servers, various JVMs, and a wide range of platforms like AIX, HP-UX, Linux, Solaris, and Windows, ensures that it can work within diverse IT environments. Integration is further facilitated by the cross-support of various business portals and systems, and an open API allows for the integration of Introscope data with existing event management and logging systems .
Introscope facilitates proactive management by providing real-time monitoring and alerts that enable the identification and resolution of potential bottlenecks before customers report issues. It actively monitors various metrics and employs analytical tools to diagnose problems early in the application lifecycle. The system's ability to capture detailed transaction paths and performance data supports proactive incident management, minimizing downtime and improving overall application availability .
Introscope utilizes an industry-standard byte-code instrumentation approach, which allows it to instrument Java applications without modifying the source code. This is possible because the Wily Agent is compatible with any JVM later than JDK 1.2 and performs instrumentation 'on the fly.' This approach ensures that application object code, such as .jar files, is instrumented and loaded directly into memory, thus enabling performance monitoring without source code changes .
Introscope's SmartStore Historical Repository captures and stores all performance metrics over time, functioning as a 'flight recorder' for Java applications. This repository allows for the analysis of historical data through the Workstation Client and reports, providing comprehensive feedback on JVM heap usage, CPU and other resources, as well as AppServer resources and Java code performance. By maintaining a complete record of performance data, Introscope assists in solving performance issues, conducting trend analysis, and planning for capacity needs .
CEM is designed as a low-risk solution due to its architecture, which involves a mirrored port of the production traffic rather than real-time intrusion into the application paths. This ensures that there is no application latency or load impact, as it analyzes copies of transaction data without affecting the network. Furthermore, CEM does not require agents or application changes, and securely manages data by storing only metadata instead of sensitive personal information, thereby maintaining compliance with security standards .
CEM provides real-time monitoring of all customer HTTP/HTTPS transactions, rather than sampling, and measures the success and satisfaction of these transactions. It identifies defects, correlates them into incidents, and calculates their business impact, which helps in efficient incident management by triggering immediate actions to resolve issues. Additionally, CEM does not impact application performance or require modifications, operates with zero insertion risk, and ensures secure handling of transaction data without storing personal information .
The ability to offer both real-time and historical data views is critical for comprehensive application management. Real-time data supports immediate performance monitoring and incident response, enabling quick identification of live issues. Historical data, on the other hand, aids in developing informed strategies for capacity planning, performance optimization, and predictive analysis, while also providing insight into past trends and recurring problems. This dual capability ensures a holistic approach to both short-term and long-term application performance management .
Introscope ensures low overhead by using byte-code instrumentation, which incurs minimal additional CPU usage (1-3% in CPU Utilisation). This approach requires no changes to the source code and operates continuously in production, QA testing, and development environments. The lightweight nature of this method, combined with the ability to instrument 'on the fly,' allows for real-time and historical performance data monitoring without significant impact on system resources .
LeakHunter enhances memory management by identifying potential memory leaks within production applications, focusing on pinpointing leak sources in components and methods. It monitors the lifecycle of all Java objects in memory and provides tools to isolate and address leaks during stress testing. This preemptive identification and resolution of leaks reduce the frequency of forced restarts and improve the stability and efficiency of applications throughout the production lifecycle .
Introscope provides an end-to-end visibility framework that captures live transactions across different tiers, from the browser to the database tier. It employs tools like the ErrorDetector and Transaction Tracer to analyze failed transactions and pinpoint bottlenecks. This helps isolate problematic transactions and trace their path through various components, such as JDBC connections, EJB calls, and parameters related to application server configurations. Through this multidimensional approach, Introscope assists in identifying performance issues across the application stack .