What’s all this flat-file talk? You lose a lot of good features with that.
Flat files of 50 posts each? What if I do a search of anything? The application is going to have to search through possibly gigabytes of flat files. Relational with indexes and foreign keys does that fast.
Replication by shipping files? If you’re think of a master/slave setup (one server writes and replicates the data to the others, which can be read from), any active thread is going to result in shipping old data. On average, you will ship 1/2 the average file size unnecessarily times the number of active threads. MySQL can be set up so the slaves periodically (even every minute) access the log on the master and read in only the new data.
If you’re thinking of a cluster (all servers reading and writing), forget it with flat files, it’s just not practical.
And most importantly, with flat files you lose enforced data integrity below the application level. It’s dangerous.
In case you haven't noticed, Search is completely broken, and has been for months now, because the server can't take the CPU load of searching.
We don't need searching, we need an active, robust, healthy server which the Nut Roots can't take down.
Right now, as things stand, the server is taking ITSELF down because it can't withstand the load.
A month from now, when the Nut Roots start pointing their attacks at this site in earnest, the server is going to melt down and set the server room on fire.
And that doesn't even address the perilousness of the single point of failure caused by having all the servers at the end of the same pipe - all the Nut Roots have to do is flood the pipe with PINGS [or other ICMP traffic], and the pipe will be hopelessly clogged.
Again, these people committed a federal felony in breaking into Sarah Palin's private email today - if you think for a microsecond that they wouldn't hesitate to take down Free Republic [and put an end to original reporting like Kristinn's], then you're insane.