Database queries, even ones that return like 30,000 rows, generally only take a fraction of a second. With a proc and RAM like that, it's probably slowing down at IO. I'd be willing to bet that most of the time is being spent on writing the file(s).

That being said, I would suggest having all the data up front. At least then, you're reading from memory. I'm not sure if MySQL loads everything into memory with a pointer to the current row, or if it only loads the current row.

With a minute processing time your dataset doesn't sound large enough to require partitioning of the result set.

So with that assumed, loading it all into memory would be fine.

However, mysql caches the query anyway, so successive duplicate queries shouldn't be a large issue either.

It does sound like IO write to file on disk is your slow point. I don't know much about how styles and formating works for xls files but I'm pretty sure you could reduce a lot of pointless writes by reducing formatting or organizing it so that you are not doing any in-line formating, if thats even possible.

The best thing I would advise is optimizing your query. That is usually the first optimizing place I look with this type of issue. If you could edit your post and include the sql I could have a look. (feel free to snip&strip sensitive information)

I would suggest profiling your PHP code to find out where the most time is spent. You can find a step-by-step tutorial about using XDebug to profile PHP conde on my blog. Then try to optimize the parts which consume the most time / are called the most frequently. One of the best optimizations you can do is to replace PHP methods by built-in functions if possible, because they are orders of magnitude faster.