Application Performance Monitoring One Approach - PowerPoint PPT Presentation

1 / 68
About This Presentation
Title:

Application Performance Monitoring One Approach

Description:

none – PowerPoint PPT presentation

Number of Views:290
Avg rating:3.0/5.0
Slides: 69
Provided by: regio6
Category:

less

Transcript and Presenter's Notes

Title: Application Performance Monitoring One Approach


1
Application Performance MonitoringOne Approach
  • John Slobodnik
  • April 18, 2006
  • 130 p.m.
  • CMG Canada

2
Introduction of Product Suite
  • ServerVantage
  • ApplicationVantage
  • ClientVantage
  • VantageAnalyzer
  • VantageView

3
ServerVantage (SV)
  • Collects server level data.
  • Multiplatform Windows, LINUX, UNIX, etc.
  • CPU, memory, disk, network out of the box.
  • Collects application level data.
  • Oracle, SQL server, WebLogic, IPlanet, LDAP, etc.
  • One SV agent installed on each client server.
  • Runs most of the time.
  • Customized counters (metrics) can be written.

4
ApplicationVantage (AV)
  • A sniffer.
  • Agent-based application analysis of packet level
    communications.
  • Gathers all traffic that passes through the
    Network Interface Cards (NICs).
  • Can merge the data together from multiple
    servers.
  • Can trace, for example, SQL server traffic.
  • One AV agent is installed per client server.
  • Turned on when required.
  • Most often in firefighting mode.

5
ClientVantage (CV)
  • Gathers data on the performance of your
    application.
  • Done through timings of synthetic business
    transactions on CV workstations (robots).
  • Scripting of business transactions done with a
    tool called QARun.
  • We are doing active monitoring.
  • There are two other options available here now
  • Passive monitoring using CV
  • A hardware-based solution

6
Vantage Analyzer for J2EE (VA)
  • J2EE (Java) based tool to help pinpoint exact
    locations of code-level performance problems.
  • Locates slow methods, SQL statements and
    transactions.
  • The VA agent runs inside your Application servers
    JVM obtaining performance metrics using Byte Code
    Instrumentation.
  • Data sent in real-time to nucleus server where it
    is stored and distributed to VA performance
    consoles.
  • Supports WebLogic, Websphere, etc.
  • Customized component
  • Allows a transaction to be followed in
    VantageView.

7
VantageView (VV)
  • Web-based portal for viewing SV, CV, AV and VA
    data for monitoring and reporting.
  • Accessing information from the Vantage suite of
    tools, VantageView users check the status of
    clients, servers and networks from their intranet
    while also providing a near-time service level
    perspective on application availability and
    performance.
  • The flexibility of VantageView enables different
    levels of users to view pertinent information for
    easy problem determination and resolution.
  • Customized counters (metrics) can be created in
    the VV database.

8
(No Transcript)
9
A Few Easy Setup Steps
  • A summary of the steps to implement the solution
  • Install the agents.
  • Complete Administration
  • Set Preferences
  • Management
  • Create tasks and apply blackout schedules.
  • Create monitoring views.
  • Create reports.
  • Optional steps taken
  • Create dashboards.
  • Create custom counters (metrics).

10
Install the Agents
  • This is a quick, procedural task that is quick to
    complete.
  • A script is run to do the install followed by any
    applicable patches.
  • The product keeps track of the level of agent
    installed on each server in a central repository.
  • SV and AV agents are installed on each server
    (Window, UNIX, LINUX)

11
Administration Configure Databases
Set up the online database(s).
12
Configure Historical Database
Define the historical database. We keep 3 months
of data online. All else goes to historical
database.
13
Control Server Configuration
Set up the control servers.
14
Define Users
Define VV user profiles.
15
Preferences
Business applications 4 applications.
Business locations Various Canadian cities.
Business transactions An application (29
transactions) is broken down into 3 transaction
groupings (14, 7, 8).
Server groupings Production, pre-production,
support, third-party, etc.
16
Management - Create Tasks
Create a new task.
17
Create Tasks
Select the type of server Windows, UNIX, etc.
18
Create Tasks
Select the counters you wish to see.
19
Create Tasks
Add a rule for alerting.
20
Create Tasks
Set up alerts if you want them. For example
System Thrashing, TCP Connectivity lost from WL
to WL layer, CPU gt 90, etc.
21
Create Tasks
Alert notification via pager, email, SNMP,
etc. Different audiences for different tasks,
DBAs, App. Support, etc.
If you can do it from a command line it can be
automated here shell scripts, bat
files. Perform an action based on a threshold
being breached. (1) Kick off a WL thread dump
based on a WL counter below a certain level. (2)
Send an alert based on an ASCII pattern
match. (3) Previous problems can be proactively
addressed with this type of instrumentation. We
examine WebLogic logs
22
Create Tasks
Select the appropriate data sampling
interval. Key to the size of your database.
23
Create Tasks
Start the task.
24
Management Blackout Schedule
  • Apply a blackout schedule, if applicable.
  • ServerVantage agents do not run when the
    application is down daily.
  • Client Vantage robots are also set up to run on
    a blackout schedule.
  • Implemented through CV which uses the Windows
    scheduler.

25
Create Monitoring Views
  • Monitoring views contain all data points.
  • Flexible you can plot many different metrics on
    the same chart.
  • Business metric vs. server performance.
  • Application metrics vs. server metrics.
  • TeeChart Editor gives you Excel chart type
    functionality to modify the look of the chart.

26
Monitoring View
Saved as a permanent monitoring view report.
27
Monitoring View
28
Monitoring (ad hoc)
Can drill into data point.
29
Drill into IDP
Intelligent Data Point (IDP)
30
Create Reports
  • Reports contain different levels of data
    summarization.
  • From all data points to daily average.
  • We have created 12 hour, 2 day, 1 week and 1
    month views of all reports.
  • Flexible you can plot many different metrics on
    the same chart.
  • TeeChart Editor gives you Excel chart type
    functionality to modify the look of the chart.

31
Create Reports
Select the metric source.
32
Create Reports
Select the metric(s) desired.
33
Create Reports
Select the time range.
34
Create Reports
Select the display format.
35
Create Reports
Schedule the report.
36
Create Reports
Save the report.
37
Reports
38
Reports
39
Reports
40
Then the Business asked
  • How can we prove that the API calls are
    performing better?
  • Custom program installed on WL servers.
  • Gathers API call response time data, converts it
    to a local CSV file, FTP to VV database.
  • API Response Time report created, queries VV DB.
  • APIs split between internal vs. outsourced (for
    reporting purposes).
  • There are a number of activities within each bean
    conversation.

41
API Response Time Report
Sample bean conversation report.
42
Then Management said
  • We need to have a some different dashboard views.
  • Each level of dashboard gets more detailed.
  • Special dashboard for outsourced infrastructure.
  • Dashboards were created using the integrated
    VISIO (Vantage Visualizer) piece of the product.

43
Management Dashboard
44
Drill down to Application Availability
45
Application Availability (bottom)
46
Drill down to Heat Chart
47
Drill down to CNS report
48
Drill down to Application Scorecard
49
Application Scorecard (bottom)
50
Drill down to Transaction Scorecard
51
Drill down to Performance Summary
52
Drill down to Orders Report
53
Drill down to Session Current Count report
54
Drill down to WL Serviced Requests report
55
(No Transcript)
56
Geographic Dashboard
57
We asked ourselves
  • How can we make this easier to support?
  • Customized metrics can be created in VV or SV.
  • Make non-standard types of metrics available.
  • Samples of some of the customization created
  • Disk usage of SV logs files directory.
  • Automated removal of SV log files.
  • Automate push of patches to all agents.
  • Send a command to run on a server and return the
    result.
  • Count the number of SV datafiles.
  • Agent restart.
  • Gather SV log files.

58
More Customization
  • TCP Connection test from WL layer to WL layer.
  • Number of Orders.
  • SQL query to xml to csv to VV DB.
  • ASCII file pattern match in WL logs (3).
  • Automatic thread dumps WebLogic.
  • Average Elapsed Time
  • Customer purchase at store experience.
  • Individual transaction timings is CV, adding them
    up is custom.
  • Network Test / TCP Connection Test
  • Traceroute response time for up to 10 hops
    alert.
  • API Response time monitoring.
  • Average, max, min, std dev

59
Network Connection Test
60
Vantage Analyzer
  • Installed on production WebLogic servers during
    the peak annual sales period.
  • Now in the pre-prod environment.
  • So bugs can be found before promoting new code to
    production.

61
J2EE JavaScape Paints a landscape view of your
J2EE environment. This view displays component
interactions between JSPs, Servlets and Web
services, Session, Entity and Message-driven
Beans, as well as database usage.
62
Transaction Explorer The tree is organized by the
largest consumers, from top to bottom. The tree
can be sorted by the CPU or Transaction time
period.
63
Transaction Scope Gives a detailed view on each
individual transaction which runs through your
application.
Stalled Threads Shows thread-level detail of a
transaction.
64
Method HotSpots Identifies the biggest consumers
in your application. The view can be sorted by
Transaction or CPU time.
65
SQLyzer HotSpots Lets you pinpoint the largest
SQL consumers.
SLA Monitoring This view displays pre-configured
SLA rules and when they were last violated.
66
Memory HotSpots Locate memory leaks as well as
memory allocation hot spots to help assist with
server availability and performance.
67
Summary
  • Management extremely pleased.
  • Customized dashboards, peak period success, want
    more applications instrumented.
  • Business application ran almost 99.9
    availability during peak processing period of the
    year in large part due to this solution.
  • Now instrumented to be more proactive than in the
    past.
  • Being used as a model for the rest of the
    enterprise.
  • Support teams have embraced the solution because
    it makes their lives easier.
  • DBAs, application support, system
    administrators, performance and capacity
    planners, etc.
  • Significantly less time wasted determining whose
    problem it is (you know, 6 teams in a room)
    during fire-fighting.

68
  • John.Slobodnik_at_bell.ca
  • (905) 282-3342
Write a Comment
User Comments (0)
About PowerShow.com