Which is better? Extracting images from directories or from database?

Question

Good day,
I would like to start a discussion on extracting image (binary data) from a relational database. Although some might say that extracting image from directories is a better approach, I m still sceptic on that implementation.
My argument towards this is based on the reasonings below:
1. Easier maintainence. - System Administrator can do backup from one place which is the database.
2. High level of security - can anyone tell me how easy it is to hack into a database server?
3. image is not dependent on file structure - no more worries about broken links because some one might mistakenly change the directory structure. If there needs to be a change, it will be handle efficiently by the database server.
The intention of my question is to find out :
1. Why is taking image from a directory folder which resides on the web server is better than using the same approach from the database?
2. How is this approach (taking image from directory) scalable if there is thousands of images and text that needs to be served?
If anybody would be kind enough to reply, I would be most grateful.
Thank You.
Regards
hatta

Answer 1

Databases are typically more oriented towards text and number content than binary content, I believe. If you carry images in the database you will need to run them through your code and through your java server before they are displayed. If they are held in a directory they will be called from hrefs in the produced page, which means that they are served by your static server. This is quicker because no processing of the image is required. It also means the Database has to handle massively less data. Depending on the database this should be far quicker to query.
It is worth noting that it is also quite difficult to actually change mime-types on a page to display a picture in the midst of HTML- the number of enquiries on these pages about this topic should be enough to illustrate this.
If you give over controls of all the image file handling to your java system (which I do when I write sites like the one you describe) then the actual program knows where to put the images and automatically adds them to the database. The system administrator never needs to touch them. If they want a backup they save the database and the website directory. The second of those should be a standard administrative task anyway, so there is not a huge difference there. The danger of someone accidentally changing the directory structure is no greater than the danger of someone accidentally dropping a database table- it can be minimised by making sure your administrators are competent. Directory structures can be changed back, dropped tables are gone.
The security claim is slightly negated because you still have to run a webserver. Every program you run on your server is vulnerable to attack but if you are serving web pages you will have a server program that is faster than a database for image handling. You are far more at risk from running FTP or Telnet on your server or (worst of all) trying to maintain IIS.
The images in directory structure is more scalable because very large databases are more likely to become unstable and carrying a 50k image in every image field rather than 2 bytes of text will make the database roughly 25000 times larger. I have already mentioned the difference in serving methods which stands in favour of recycling images. A static site will be faster than a dynamic site of equivalent size, so where you can, take advantage of that.