I could be wrong BUT...
> 1) Does increasing the number of replication slaves increase query
> latency on the master? We're considering tiering the replication if
> it might help - replicate the master to two slaves, each of which
> replicates to ten clients.
The slaves should only be pulling from the log file, not querying the
master data directly. But yes, I guess I could cause an additional load
on the server if there are many many slaves. But with < 10,000 updates
a day (that is 8 per minute, this shouldn't be much of a load at all.
> 2) Is there a chance that the insert latency is coming from the fact
> that the table is growing so long? At a certain point, even with
> indexes, I imagine that the engine is going to have to do some linear
> searching.
Well, back to answer 1. Replication is about log's, not querying the
data.
You mentioned updates, but what about querying the data. Do you run a
lot of queries against the data on the master server? We have a
database with 50M rows in it and we have a complicated replication
strategy for the reader just so we can take 99% of the load off the
master. We have a slave'd database just to run reports from (actually
we have a load balanced cluster of them). The master received inserts
about 20 records/s
Also, what type of database are you using? INNODB? MyISAM? If you are
running MyISAM then things can get slow on updates.
