Using MySQL as a NoSQL - A story for exceeding 750,000 qps on a commodity server

Neat. I didn't know you could do anything like that with MySQL. Does Postgres have an equivalent? I wouldn't move from an SQL DB to a NoSQL key/value store without a significant performance improvement. Seems like you can tune MySQL far more than I realised...

Interesting that bypassing the SQL part of MySQL would increase performance by that much. In the comments he says that prepared statements would still be slower but it would be useful to see some numbers on his hardware to compare with.

So, if you're bypassing most of the bits of an RDBMS, what do you actually gain? Is the InnoDB storage engine that much superior compared to other products? Or is it about the management and admin features, e.g. sharding, replication?

There was a little discussion when this article was posted here three weeks ago: <a href="http://news.ycombinator.com/item?id=1807527" rel="nofollow">http://news.ycombinator.com/item?id=1807527</a><p>(OT: under what circumstances are duplicate submissions allowed? There's nothing in the FAQ or the Guidelines.)

Pretty cool. It seems this kind of hack would also integrate well into the Drizzle project as it's designed for light weight pluggable storage. I'm curious if this has any impact on thread buffers and the number of clients MySQL can accept. It's great to be able to perform 750k QPS, but if you're still limited to a few thousand clients, other solutions may be preferable.

Cool project. It's great to see people doing things like this. On the other hand, I don't know that it would be useful for many people. Note this caveat near the end.<p>&#62; No benefit for HDD bound workloads<p>Which is most workloads.<p>The author cites the expense of memory as the drawback to memcache. It's true, a large pool of dedicated memcache machines can be quite expensive. However, the given technology only works when the database fits in memory. If the database fits, then a cached subset of it certainly will fit as well. So memcache will only use a single server. One server is not a major expense compared to overhauling an application to marry it to a custom MySQL extension.<p>Also consider what happens if the database ever does grow beyond memory capacity. If you're using memcache, you might just need to add more cache servers. With HandlerSocket you'd have to... switch to memcache. Then add more cache servers.