为高容量共享媒体服务器PHP / mySQL重命名文件的实用方法

Question

Ok I am in the midst of developing a shared system/service of sorts. Where people will be able to upload there own media to the server(s). I am using PHP and mySQL for the majority of the build, and am currently using a single server environment. However I need this to be scaleable as I do intend on moving the media to a cluster of servers in the next 6 months leaving the site/service on its own server. Anyway thats a mute point.

My goal, or hope rather is to come up with an extremely low risk naming convention that runs little possibility ever of running into a collision with another file when renaming the file upon upload. I have read to date many concepts and find that UUID (GUID) is the best candidate for my over all needs as it has a number so high of possibilities that I dont think I could ever reach that many shared images ever.

My problem is coming up with a function that generates a UUID preferable v3 or v5 (I understand they are the same, but v5 currently doesn't comply 100% with the standard of UUID). Knowing little about UUID and the constraints there of that makes them unique and or valid when trying to regex over them later when and or if needed I can't seem to come up with a viable solution. Nor do I know which I should really go with v3 or v5. or v4 for that matter. So I am looking for advice as well as help on a function that will return the desired version UUID type.

Save your breath I haven't tried anything yet as I don't know where to begin currently. With that, I intend on saving these files across many folders to offset the loads caused by large directory listings. So I am also reducing my risk of collision there as well. I am also storing these names in a DB with there associated folders and other information tied to each image, so another problem I see there is when I randomly generate a UUID for a file to be renamed I don't want to query the DB multiple times in the event of a collision so I may actually want to return maybe 5 UUID per function call and see what if any have a match in my query where ill use the first one that doesnt have a match.

Anyway I know this is a lot of reading, I know theres no code with it, hopefully the lot of you don't end up down voting this cause theres to much reading, and assume this is a poor question/discussion. As I would seriously like to know how to tackle this from the begining so I can scale up as needed with as little hassel as possible.

Answer 1

If you are going to store a reference to each file in the database anyway .. why don't you use the MySQL auto_increment id to name your files? If you scale the DB to a cluster, the ID is still unique (being a PK, it must be unique!), so why waste precious CPU time with the UUID generation and stuff? this is not what UUIDs are made for.

I'd go for the easiest way (and i've seen that in many other systems, though):

1. upload file
2. when upload succeded, insert DB reference (with the path determined by 3.); fetch auto_incremented $ID
3. rename file to ${YEAR}/{$MONTH}/${DAY}/{$ID}
   (adjust if you need a more granular path, when too many files uploaded per day)
4. when rename failed, delete DB reference and show error message
5. update DB reference with the actual actual path in the file system