How to monitor clients that access your blob storage?

Some time ago I wrote about the monitor and logging support available for Windows Azure Storage. In this post we will talk about how we can use this feature to detect what clients are accessing storage.
Let’s assume that we have a storage that is accessed by 100.000 users. At the end of the month we should be able to detect the users that downloaded a specific content with success.
What should we do in this case?
(Classic Solution) Well, in a classic solution, we would create an endpoint that will be called by the client after he download with success the specific content. In this case we would need to create a public endpoint, hosted it, persists the calls messages, manage and maintain the solution and so on.
In the end we would have additional costs.

What should we do in this case?

(Windows Azure Solution) Using Windows Azure Storage, we can change the rules of the game. Windows Azure Storage offer us out of the box support for logging mechanism. All the request that are made to our storage will be logged.

Based on this idea, we activate this feature and log all the access to the storage. In the following example you can see how the logs look like for a file called ‘m.txt’ under a container named ‘container’.

1.0;2013-11-12T13:03:23.4277012Z;GetBlob;AnonymousSuccess;200;7;7;anonymous;;radudemo;blob;"http://radudemo.blob.core.windows.net/container/m.txt";"/radudemo/container/m.txt";98ebc2d2-249f-47d2-8a3b-e7bbcb5e3c59;0;86.124.100.155:21131;2009-09-19;379;0;284;7;0;;;"0x8D0ADBEAF6D6A3C";Tuesday, 12-Nov-13 13:03:15 GMT;;"Mozilla/5.0 (Windows NT 6.3; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/30.0.1599.101 Safari/537.36";;

As you can see, we have all the necessary information to identify the client, inclusive the client IP. But we want to do more than that. We want to be able to detect clients based on aware own identification mechanism.
To be able to do something like this, we have to options:
Query parameters
We can add a custom query parameter that represent the client unique ID. For example we can make the following requests
http://radudemo.blob.core.windows.net/container/m.txt?clientID=123
In this case, the client will receive the file and in the logs we will have:

1.0;2013-11-12T13:03:36.3117012Z;GetBlob;AnonymousSuccess;200;14;14;anonymous;;radudemo;blob;"http://radudemo.blob.core.windows.net/container/m.txt?clientID=123";"/radudemo/container/m.txt";c90e928f-f3e7-4c4f-859b-4bb0051f0271;0;86.124.100.155:21131;2009-09-19;392;0;284;7;0;;;"0x8D0ADBEAF6D6A3C";Tuesday, 12-Nov-13 13:03:15 GMT;;"Mozilla/5.0 (Windows NT 6.3; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/30.0.1599.101 Safari/537.36";;

In the client URL we can find the clientID that can be filter, processed and so on.
User Agent
Another option is to set a custom user agent. Using a custom user agent it will be very easily for us to identify witch client made the requests.

1.0;2013-11-12T13:03:36.3117014Z;GetBlob;AnonymousSuccess;200;14;14;anonymous;;radudemo;blob;"http://radudemo.blob.core.windows.net/container/m.txt";"/radudemo/container/m.txt";c90e928f-f3e7-4c4f-859b-4bb0051f0271;0;86.124.100.155:21131;2009-09-19;392;0;284;7;0;;;"0x8D0ADBEAF6D6A3C";Tuesday, 12-Nov-13 13:03:15 GMT;;"Mozilla/5.0 (Windows NT 6.3; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/30.0.1599.101 Safari/537.36 ClientID/123";;

Both solutions are similar, we are only sending the unique client id in different locations. This files can be very easily processed and analyzed using different solutions. A good solution when you have a lot of logs file is by Hadoop.
The advantages of using such a solution is the out of the box support. You don’t need to create and manage another system for this. Also you eliminate the case when the client is able to download/access the content but is not able to confirm the download.
The only additional costs that are added to a classic solution is the storage cost for storing the logs data itself, plus the transactions costs. This costs are very low in comparison with the case when you would create all the infrastructure for logging. In both cases you would need to pay the storage costs of logs.