SPONSORS:






User Tag List

Thanks Thanks:  0
Likes Likes:  0
Dislikes Dislikes:  0
Page 1 of 2 12 LastLast
Results 1 to 10 of 12
  1. #1
    Member
    Join Date
    May 2002
    Location
    bangalore, karnataka, INDIA
    Posts
    45
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Total Downloaded
    0

    performance of each tier in n tier architecture

    mates,
    Is there a tool to test the performance of each tier in an n tier application

    ------------------

  2. #2
    Junior Member
    Join Date
    Oct 2001
    Location
    MUNIC,BAVERIA,GERMANY
    Posts
    13
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    SilkPilot from Segue.
    For testing middleware components.

    ------------------
    Chitra Dewan
    Chitra Dewan

  3. #3
    Moderator
    Join Date
    Aug 2001
    Location
    NC
    Posts
    6,018
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    1 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    Evaluate the communications used between each tier, then make sure whatever tool you have selected will accomodate the protocols/communications mechanisms in use. For example, consider the following three multi-tier solution

    Oracle database
    <OCI>
    Application server (Java)
    <Java_RMI>
    Front end web server
    <HTTP>
    Client

    Here is how I would approach this. Test Oracle first. Dump the logs and then pull the queries generated by each business processes from the logs. Use a generic database client to "record" the SQL statements associated with each business process, parameterize and correlate as appropriate.

    One the database tier is qualified, move to the application server tier. Here is the challenging part. How to collect the calls that are moving back and forth between the web server and the app server as Java_RMI. Well, if you had a sniffer and were versed atr reverse engineering you could work it that way....but this method is very dirty and is not appropriate for most. Work with the development staff to get access to the source code associated with each business process. At least in the Loadrunner universe (not sure of others), you can pull over the same calls into a Java virtual user for load testing. (average respponse time for business process a in application server - average response time for business process a in database = average processign latency (in time) of application server.

    Next up, record the business processes from an HTTP perspective from the client. Run the load test. You now have a way to measure the latency in Web Server, in App Server, all the way to database.

    This is the ideal, rarely are companies willing to devote the time and resources required to run three seperate phases of load testing prior to deployment. Where you are most likely to see this happen is in the case of a distastrous performer in an attempt to isolate which layer is the bottleneck in the entire app.

    ------------------
    James Pulley
    nospam.jpulley@nospam.itestsolutions.com
    iTest Solutions, Inc
    704-243-2854 (voice)
    James Pulley

    Replace ineffective offshore contracts, LoadRunnerByTheHour. Starting @ $19.95/hr USD.

    Put us to the test, skilled expertise is less expensive than you might imagine.

    Twitter: @LoadRunnerBTH @PerfBytes

  4. #4
    Moderator
    Join Date
    Aug 2000
    Location
    Vancouver, BC, Canada
    Posts
    1,189
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    <BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by jpulley3:
    This is the ideal, rarely are companies willing to devote the time and resources required to run three seperate phases of load testing prior to deployment. Where you are most likely to see this happen is in the case of a distastrous performer in an attempt to isolate which layer is the bottleneck in the entire app.<HR></BLOCKQUOTE>

    We tend to see this here. I have been lobbying (unsuccessfully) for the instrumentation of the code with the ARM API ( http://regions.cmg.org/regions/cmgarmw/marcarm.pdf ). This will allow you to exactly define your measuring points and as an added benefit this can be used in production to track response times and stability. Together with your database trace and measurements this will give you a complete picture.

    What are your thoughts about the feasability of such an approach?



    ------------------
    Roland
    Roland Stens

  5. #5
    Moderator
    Join Date
    Aug 2001
    Location
    NC
    Posts
    6,018
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    1 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    My bias is to not instrument code at the performance testing phase, because one is either left with a disjoint between what is tested versus what is deployed, or one is left with extra code in the deployed app which has no benefit to the end user on performance.

    At the unit testing level, that's a different story. Lots of benefit. The trick then is getting the developers to remove all of the debug items from their source.
    James Pulley

    Replace ineffective offshore contracts, LoadRunnerByTheHour. Starting @ $19.95/hr USD.

    Put us to the test, skilled expertise is less expensive than you might imagine.

    Twitter: @LoadRunnerBTH @PerfBytes

  6. #6
    Moderator
    Join Date
    Aug 2000
    Location
    Vancouver, BC, Canada
    Posts
    1,189
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    <BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by jpulley3:
    My bias is to not instrument code at the performance testing phase, because one is either left with a disjoint between what is tested versus what is deployed, or one is left with extra code in the deployed app which has no benefit to the end user on performance.<HR></BLOCKQUOTE>

    Yes I agree on that but the ARM API is supposed to go into production as well so that performance can be continuously monitored. There are many monitoring tools out there that can receive information from the ARM API, chart the results, react on changing situations etc.

    I found that in our environment it is not so much a challenge to collect performance information during testing but when the application is in production. Any monitoring in production is deemed taboo because of the influence it has on the performance. Therefore we are missing an important piece of the puzzle.



    ------------------
    Roland
    Roland Stens

  7. #7
    Moderator
    Join Date
    Aug 2001
    Location
    NC
    Posts
    6,018
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    1 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    Have you considered any SLA/Performance management tools which use the same sampling method as in test? I am not sure abotu all of the vendiors on the left, but a Loadrunner virtual user can be rolled over the wall from test into production through the use of TOPAZ (Mercury Product) or TAPM (Tivoli Product). This avoids the instrumented code approach and it also provides a common sampling method from test to production - no wasted effort explaining the response time devations due to sampling differences.
    James Pulley

    Replace ineffective offshore contracts, LoadRunnerByTheHour. Starting @ $19.95/hr USD.

    Put us to the test, skilled expertise is less expensive than you might imagine.

    Twitter: @LoadRunnerBTH @PerfBytes

  8. #8
    Moderator
    Join Date
    Aug 2000
    Location
    Vancouver, BC, Canada
    Posts
    1,189
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    <BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by jpulley3:
    Have you considered any SLA/Performance management tools which use the same sampling method as in test? I am not sure abotu all of the vendiors on the left, but a Loadrunner virtual user can be rolled over the wall from test into production through the use of TOPAZ (Mercury Product) or TAPM (Tivoli Product). This avoids the instrumented code approach and it also provides a common sampling method from test to production - no wasted effort explaining the response time devations due to sampling differences.<HR></BLOCKQUOTE>

    Yes, but our IS security folks will not allow me to do this. It's a good idea but it doesn't fly for me.

    Thanks.



    ------------------
    Roland
    Roland Stens

  9. #9
    Moderator
    Join Date
    Aug 2001
    Location
    NC
    Posts
    6,018
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    1 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    I see your challenge. Certainly there is some level which is deemed non-intrusive, once every 24 hours, once every 12 hours, once every hour, once every 15 minutes for a short duration, read only transaction or something as non-intrusive as long/logout? ARM would provide continuous intrusion into every section of code so altered, one active sample every (insert your interval here) would only intrude once at the specified interval and only for the business process under consideration. It is the goal of SLA tools to be as non-intrusive as possible.

    There are client side solutions that watch the work effort of the client and traffic going to and from the client. However, the response times generated from these Passive solutions can almost never be compared to an active generated solution based upon a load testing tool like LoadRunner, since all they do is either patter match or watch for pauses in a conversation and brand thes patters or traffi between pauses as "timing records" or transactions abnd because the nature of the traffic flows is highly variable, never do you have a consistent repeatable sample that is being used for SLA purposes.

    How do the secutiy folks monitor it, reactively based upon number of calls to the support center? Do they use a college intern to log in "every so often" and collect results? Maybe there is a way to mimic their groups methods for assessing performance proactively using a tool

    ------------------
    James Pulley
    nospam.jpulley@nospam.itestsolutions.com
    iTest Solutions, Inc
    704-243-2854 (voice)
    James Pulley

    Replace ineffective offshore contracts, LoadRunnerByTheHour. Starting @ $19.95/hr USD.

    Put us to the test, skilled expertise is less expensive than you might imagine.

    Twitter: @LoadRunnerBTH @PerfBytes

  10. #10
    Moderator
    Join Date
    Aug 2000
    Location
    Vancouver, BC, Canada
    Posts
    1,189
    Post Thanks / Like
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Total Downloaded
    0

    Re: performance of each tier in n tier architecture

    I am not responsible for monitoring the performance in production nor are our IS security people. But since I signed off on the Performance in Test and because for my work I need "real" data to compare my results to, I have been pushing for production monitoring.

    Adhoc monitoring is done, but by the System support people who are basically just interested if there servers are still up and running. Any major errors (time outs, database errors etc) are reported to our helpdesk.

    The company is now embarking on an effort that will greatly add to the complexity, size and targeted public. Currently the full appreciation for the complexity of the plans and what is needed to manage this properly, is lacking.

    I appreciate your thoughts on this. You certainly gave me some ammunition.

    Thanks

    ------------------
    Roland
    Roland Stens

 

 
Page 1 of 2 12 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
Search Engine Optimisation provided by DragonByte SEO v2.0.36 (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 10.00%
vBulletin Optimisation provided by vB Optimise v2.6.4 (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
User Alert System provided by Advanced User Tagging v3.2.8 (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
vBNominate (Lite) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Feedback Buttons provided by Advanced Post Thanks / Like (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Username Changing provided by Username Change (Free) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
BetaSoft Inc.
Digital Point modules: Sphinx-based search
All times are GMT -8. The time now is 11:41 PM.

Copyright BetaSoft Inc.