Generally speaking, I'll allow storing of small images (such as small badge photographs) in a database and store large binaries in a file-path reference only.

In all practicality, I'll usually only use the file-path reference method even for small binaries for two reasons.

1. The data normally originates in a file. If we use the file-path reference method, you can avoid the bit of pain it takes to load a binary into the database. This is especially important if changes to the binary are made. No database action is required, really. Just overwrite the file and you're done.

2. Since the data normally does originate as a file, most people end up wanting to keep the file even if they don't use it for other things. It seems a bit wasteful to me to have the data stored in two places.

3. Of course, there's the front end and related security aspects. It's sometimes a whole lot easier on both if the binary data is stored in the database. Of course, if the binary data will never change, it seems a waste to constantly backup potentially voluminous static data over and over. It also makes restores take much longer should a restore need to be accomplished. If the binary data were stored in files instead of the database, restores would be much quicker. It will also distribute the I/O to possibly more than one machine (a file server and SQL Server) which would usually enhance performance.

So the answer is and although I lean toward the file-path reference method, "It Depends". Every situation is a bit different and I wouldn't rely on any rule of thumb here. Sit down and make a list of requirements vs advantages/disadvantages of each method and then make a decision.