incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Wendall Cada <wenda...@83864.com>
Subject Re: Tracking doc access
Date Fri, 15 Mar 2013 03:22:50 GMT
The performance of a write per read in updating the doc with a timestamp 
would be very, very poor in CouchDB.

The best scenario is create a separate stats database. Every time a doc 
in the database you are tracking for is accessed, create a doc 
describing the request in a stats database. Creating new docs in CouchDB 
is very inexpensive, so you'll not see any performance issues with this 
versus updating docs per request.

Create a new doc in the stats db like this:
{
"db": "name_of_tracked_db",
"id": "_id_of_doc_being_tracked",
"timestamp": timestamp
}

Then create a view in this database for your database that maps the 
values. You can create several view indexes to separate the data for 
whatever your needs are.

To view :
"doc_access": {
     "map": "function(doc) {
         emit([doc.db, doc.id, doc.timestamp], 1);
     }",
     "reduce": "_sum"
}

A mock query for this to see the number of times a doc was accessed over 
the entire date range would be:

http://localhost:5984/stats/_design/data/_view/doc_access?startkey=["name_of_tracked_db","_id_of_doc_being_tracked",""]&endkey=["name_of_tracked_db","_id_of_doc_being_tracked",{}]&group=true

You'd get back a result like this:
{"rows": [
{"key":["name_of_tracked_db","_id_of_doc_being_tracked"], "value": 42}
]}

If you want to get results for a specific range of dates, simply add the 
dates to the third component of the query.

This method gives you the ability to get stats for the access counts for 
an entire db, a range of docs, or a single doc for any given period of time.

The advantage of this approach 1. it's fast 2. it is extremely flexible

The disadvantage is that it takes up a ton of disk space if you never 
purge old items from the db. I've been tracking every single page 
request to our servers in this way with quite a bit of metadata in the 
docs since Dec. 2010. That database is currently 5GB compacted for ~50k 
page requests per day over this period of time. I never had the need to 
delete a single doc from this db.

I don't have any benchmarks for a comparison between the two methods, 
but I'd strongly discourage a write per read model for your accessed docs.

For an understanding about how the ordering for views works, see 
http://wiki.apache.org/couchdb/View_collation

HTH,

Wendall

On 03/14/2013 07:16 PM, Stephan Bardubitzki wrote:
> Hi Thomas,
>
> no, I need only to track read, and I need the timestamp for some charts.
>
> Stephan
>
> On 13-03-14 07:02 PM, Thomas Hommers wrote:
>> Hi Stephan,
>>
>> With 'accessed' do you mean read and write ? In case you just want to 
>> track write access i believe you could use the _rev attribute.
>>
>> Regards
>> Thomas
>>
>>
>>
>> ----- Reply message -----
>> From: "Stephan Bardubitzki" <stephan@bardubitzki.com>
>> To: "user@couchdb.apache.org" <user@couchdb.apache.org>
>> Subject: Tracking doc access
>> Date: Fri, Mar 15, 2013 08:57
>>
>>
>>
>> Hi there,
>>
>> I have a task where I need to track how often a doc is accessed. The two
>> possible ways I can think of are:
>>
>>   1. add an array to the doc and add the timestamp when it is accessed
>>   2. create a new document and add the doc._id and the timestamp
>>
>> Which one would you prefer? Or is there a better solution?
>>
>> Thanks,
>> Stephan
>>
>>
>> --------------------------------
>> Spam/Virus scanning by CanIt Pro
>>
>> For more information see
>> http://www.kgbinternet.com/SpamFilter.htm
>>
>> To control your spam filter, log in at
>> http://filter.kgbinternet.com
>>
>


Mime
View raw message