Replies to several comments
@ bored AC
"Seriously, are you going to use the express version for your final deployment?"
"nonetheless you aren't half making a massive rod for your own back by beta testing for MS"
Thanks for the comments. No, we won’t be using the Express version – we were using that early on (in fact, in the 2005 version) to do some rough prototyping and it happened to illustrate rather elegantly that the database file size was getting quite large, even with relatively few blobs.
And no, I don’t think we are making rods for backs. The original decision to use the 2008 stack was made precisely so that we could gain experience with 2008. You will have gathered that the download/install of Microsoft CTPs does not fill me with awe and wonder (or, rather, it does, but only in a bad way….). However all of the rest of the CTP program, actually using the software has been, as reported in the rest of this series, remarkable pain free. We have gained huge experience and both the management and the developers are extremely happy that we decided to go down the CTP route.
Had we started with 2000/2005 as you suggest we wouldn’t have gained the experience.
Kevin asks an interesting question “Who pays for the install time?” The answer is the management. Are they happy to do so? They are delighted. Why? Because they understood at the time the decision was made that working with a CTP program has associated costs. They (not the developers) decided that the benefits were likely to outweigh the costs and, happily, in this case, they have been proved correct.
@AC (possibly the same one, possibly not, it’s difficult to tell…..)
“And why worry about DB size? Overall size counts - total data doesn't get smaller if you store it outside the db; there's just as much to back up, only it's in pieces... which makes reliable backup during runs problematic, if a file is added/deleted while backup is going on. Please explain!”
It’s always a trade off. Keeping the BLOBS in the file system can be an advantage if the blobs are mainly static and can be compartmentalised. That way you can only backup the new BLOBS and it makes the backup of the database much faster because it is smaller. However I don’t want to make it sound as if these were the only considerations. We could, for example, also have partitioned the database and achieved a similar (but not identical) effect. It is easy to see these architecture decisions as simple – in fact they are relatively complex and there was some discussion and testing before we made the decision. (Indeed, that is another reason why we wanted to use the CTP, so that we could do the testing!).