Skip to Content
avatar image
Former Member

what is sent between TREX client and TREX server?

what type od data are sent between TREX client and TREX server?

for example, document search. what is transfer thru network?

thanks

Carol

Add comment
10|10000 characters needed characters exceeded

  • Get RSS Feed

3 Answers

  • Aug 28, 2010 at 04:45 PM

    Hi,

    the type of communication RFC or http(s) depends on the kind of application.

    The TREX client is send the search request to the TREX server and the TREX server sends the search result to the TREX client.

    That's all.

    Best regards

    Frank

    Add comment
    10|10000 characters needed characters exceeded

  • Aug 30, 2010 at 08:51 AM

    Hi,

    within the RFC/http request there is a XML structure for the search request (a TREX notation of the search request with the search method, the TREX attributes which should be in the TREX search result and so on) and the TREX search result is as well a XML structure with all the requested TREX attriutes and the ID of the search object and so on. But this structure can differ when you are not asking TREX for a search result. Because you can as well ask TREX for "Did You mean" results and so on.

    But anyway it is not much important the knowledge of the exact structure because you have to use the function module in ABAP which are released or the released API in Java. Even it is not supported to use the un-released TREX function modules which are available in ABAP.

    Best regards

    Frank

    Add comment
    10|10000 characters needed characters exceeded

    • Former Member

      hi Frank, thanks for your information.

      i want to make my question more clear. for example: document search in KM in EP.

      1.. user send a search request to TREX http client

      2.

      The Web server converts the requests to a TREX-internal format and then forwards them to the responsible TREX servers.

      3.

      The queue server collects incoming document,

      4.

      preprocessor loading documents

      directly or load from the repository that the URIs reference.

      this sounds like, user send (all) documents to Web Server in a search request, Web server converts to TREX format and send to TREX server. then TREX send back indexed documents back to client.

      my question is :

      is this data workflow correct?

      is this statement user send (all) documents to Web Server in a search request correct?

      Edited by: Wang Carol on Sep 2, 2010 8:42 AM

  • Sep 02, 2010 at 09:55 AM

    Hi Wang,

    I see.

    So my explanation is a very simple one and only regarding search.

    But now we go in more detail and also we have a look to indexing.

    Searching:

    The user (Portal / application based on AS Java) is sending a search request to TREX http server.

    The TREX http server converts the request in an internal format and asking the TREX Name Server which TREX Index Server it should ask for the search request. Afterwards the TREX http Server is asking the assigned TREX Index Server. The TREX Index Server is searching based on the index and the result is send back to the TREX http server and afterwards as http(s) again back to the user.

    Indexing:

    We assume that you are running your index with a queue. This is the default when using a Portal.

    The Portal is sending all commands like create a index as http(s) to the TREX http server. Afterwards the TREX http server is asking the TREX Name Server which TREX Index Server it should use. Sending this command to the index server.

    Afterwards the Portal is collecting all documents which should be indexed and sending this documents (not the content, only the RID and URL) to the TREX http server. Again the TREX http server is asking the name server which queue server it should use. Sending this documents (url) to the queue server.

    The queue server now proceed its queue. Sending a document (URL) to the TREX PreProcessor. Now the PreProcessor fetch the document content with the attributes via a URL request from the Portal and proceed further. Afterwards when the document is converted in UTF-8 / html then the PreProcessor is finished. After this the queue server regarding the queue parameter of the index is sending all this information to the Index Server and the Index Server is updating the index.

    Finish with indexing......

    Best regards

    Frank

    Add comment
    10|10000 characters needed characters exceeded

    • Hi,

      Q: It * fetch the document content with the attributes* via a URL request from the Portal and proceed further.

      So, preprocess definitely will collect content of documents and convert into UTF-8 format. right?

      A: YES. There is one URL request from PreProcessor to Portal Server (KM) for fetching the document content and the document attributes (all attributes from KM side which are marked as indexing in the KM Meta Property configuration (in the defualt the file name, description and display name).

      Q: if per day 20,000 search requests with thousands of MS office documents, there will be very large data traffic over http request/response.

      A: Yes it could. 20.000 search request per day means 20.000 http requests/response from TREX client (Portal) to TREX http server (containing the search request and search result).

      10.000 documents which are new or updated in the portal means 10.000 http messages from TREX client to the Queue Server (only containing the document URL). And 10.000 http requests from PreProcessor to TREX client for fetching the document content and attributes which are marked as indexing.

      best regards

      Frank