Skip to main content

ATG Search - performance decisions in Enterprise search

Read more about the basic design decisions for a commerce search setup @ http://tips4ufromsony.blogspot.in/2012/03/basic-design-decisions-for-commerce.html

Read more about search engine tuning @ http://tips4ufromsony.blogspot.in/2012/02/atg-search-search-engine-tuning.html

Following items will affect search performance:  number of items to be searched, quantity of content for each item, number of meta tags, number of facets, Search Merchandising settings, constraints, number of returned results,... You could find a detailed analysis below :

1. Decide how many search engines you need : Based on the performance requirement of the ecommerce site, create the Search Project @ BCC - Search Admin and add the search engine hosts. Read more about search project creation and search engine host set up @  http://tips4ufromsony.blogspot.com/2012/01/atg-search-how-to-create-search-project.html.

2. Decide the search indexing properties : Search indexing time and search response time are dependent on the index file size. As part of verification process, the irrelevant attributes from search index definition file should be removed. As a result, users will view relevant results with less number of attributes and also will get improvements in search indexing time.

3. Decide how many search projects you need : If your application is a multi locale application, you could create multiple search project for each locale. This will give you more options for each locale and a separate search engine can support a locale. Read more about search project creation @  http://tips4ufromsony.blogspot.com/2012/01/atg-search-how-to-create-search-project.html.

4. Decide whether you need to do parallel search indexing : If you have multiple search projects, you could do parallel search indexing and improve search indexing performnace.

5. One search engine per box : Search engines should always run on their own machine, with no other applications running.  This is critical for machines responding to search queries to ensure proper and consistent user experience.

6. Indexing engine box : The search engine used for indexing should not be run on the same server as an engine used for query response.  The search engine used for indexing is preferred to be run on its own machine (with no other applications), but can be run on the CA/BCC.  If this is done, the CA machines should be deployed with extra power (CPU speed and cores, RAM) so that indexing does not affect BCC user experience.


Labels:

Comments

Post a Comment

Popular posts from this blog

JBoss - know more about the JBoss directory structure

Fundamentally, the JBoss architecture consists of the JMX MBean server, the microkernel, and a set of pluggable component services - the MBeans. The JBoss Application Server ships with three different server configurations. Within the <JBoss_Home>/server directory, you will find three subdirectories: minimal, default and all. The default configuration is the one used if you don’t specify another one when starting up the server. If you want to know which services are configured in each of these instances, look at the jboss-service.xml file in the <JBoss_Home>/server/<instance-name>/conf/ directory and also the configuration files in the <JBoss_Home>/server/<instance-name>/deploy directory. JBoss 4.0 features an embedded Apache Tomcat 5.5 servlet container. conf --> The conf directory contains the jboss-service.xml bootstrap descriptor file for a given server configuration. This has the jboss-log4j.xml file which configures the Apach...
Post a Comment
Read more

ATG Search and Search engine activity log

We could use the SearchEngineActivity log files to get the request/response to the search engine from a commerce instance. This folder is located in each commerce instance or the instances from which the call to the search engine is done. The SearchEngineActivity log file folder can be configured @ SearchEngineService component ( /dyn/admin/nucleus/atg/search/routing/SearchEngineService). To get the log files for the search engine calls, you need to specify the SearchEngineService.dumpingRequests as true. Then you need to specify the engineActivityPath as the folder in which you need the SearchEngineActivity logs. Below you could find my SearchEngineActivity log folder. Each successful call to the search engine from the commerce instance will create 5 files in the SearchEngineActivity folder : namely  request, response, search engineinfo, stack trace and response row . Each file name start with a specific sequence. You ...
1 comment
Read more

How to simulate Browser back button

When someone asks how to simulate a back button, they really mean to ask how to create a link that points to the previously visited page. Most browsers tend to keep a list of which websites the user has visited and in what order they have done so. The DOM window object provides access to the browser's history through the history object. Moving backward and forward through the user's history is done using the   back(), forward(), and go() methods of the  history  object. To move backward through history, just do window.history.back() ; This will act exactly like the user clicked on the Back button in their browser toolbar. Find below a sample html code: <html> <head> <script type="text/javascript"> function goBack(){  window.history.back() } </script> </head> <body>    <input type="button" value="Back" onclick="goBack()" /> </body> </html>
Post a Comment
Read more

ATG Search - high level overview of product-catalog-output-config.xml and XHTMLs

The definition file format begins with a top-level item element that specifies the repository and item descriptor to use, and then lists the properties of that item type to include. The top-level item element has the is-document attribute set to true. This attribute specifies that an XHTML document should be generated for each item of that type (in this case, each user item). Property values that come from standard JavaBean properties of the RepositoryItem object (rather than dynamic bean properties) are specified using a dollar-sign ($) prefix. The item element has an is-multi attribute for specifying multi-value properties. If a property is an Array, Collection or Map  you should set this attribute to true. Eeach ATG Search document is uniquely identified by a URL (typically the path name of the file on the file system). In the XHTML documents that the ATG platform generates from repository items, meta properties are represented by meta tags in the head of the document, whi...
Post a Comment
Read more

Search engine shutdown call from the Estore instnace

 When the estore or commerce instances are restarted the routing system service can send shut down requests to stale engines, this can be caused because of any of the following. 1. Some other instance marked the engine as stopped in the DB (This can be caused because the machine could not reach the host running the engine) 2. The eStore instance is using a different search schema that has data about the search engine host and is marked as stopped in that DB. To avoid this overrride \atg\search\routing\RoutingSystemService.properties by setting cleanUpStrayEng=false.
Post a Comment
Read more