Talk:FindMe
CS290F Fall 2006 - UCSB Computer Science - Thorsten von Eicken
Project Comments
Guys, looks pretty good. A couple of comments. First, where's the dream functionality? No, actually you don't really need any. This is sufficiently advanced as-is. Next I'm going to lay out what data models I think you should have:
- users: obvious
- locations: again, duh.
- user_locations: this is a relational data model that connects users and locations with some additional metadata describing the association (type of association, etc.)
Actually, that's about it. There's not need for preferences because that can just go in the users table. Does a user have anything other than exactly one set of preferences? Is there a reason it needs a separate table? Also, the session data is manged by rails, and will not live in the database. In project four, you will use memcached to keep it memory resident and available to multiple Rails instances.
Stefan 15:38, 27 October 2006 (PDT)
- Can't get to your site. I only get the Fedora Core Test Page... Please email me when fixed.
- Careful when storing user object in session: need to invalidate if anything changes in user object.
- The formatting of your SQL analysis pages is pretty screwed-up. I also don't see much in the way of analysis...
TvE 00:11, 18 November 2006 (PST)
- Please create a page for "/" on port 80 or an appropriate redirect.
- Creating an account generates an application error.
- Looks like I can't test much...
TvE 17:49, 26 November 2006 (PST)
- Good description of your setup & good start at performance analysis!
- Time to reorganize your information into multiple pages...
- Please shrink the graphs a bit, I can't see the whole graph on my screen!
- What is the X axis of your graphs? I suspect it's "connections started per second".
- I don't get much out of your graph. It just shows that as you increase load the queueing delay goes up. How many requests/sec are delivered? Look at a couple of other projects for inspiration on better presentation. Also, you do not explain your test setup enough.
- You mention DB optimizations: where do you describe them?
- BTW, how big is your database (rows/bytes)?
TvE 00:11, 29 November 2006 (PST)
- Regarding your initial multiple-machines results, it does indeed look like you need to work on the database some more, I can't explain the lack of linear scaling otherwise.
- Be sure to start the connections/sec low enough to catch the "ramp-up".
- You should try and fix the bottleneck before running things on more machines.
TvE 20:57, 4 December 2006 (PST)
