Howdie stackoverflow people!
So I’ve been doing some digging regarding these NoSQL databases, MongoDB, CouchDB etc. Though I am still not sure about real time-ish stuff therefore I thought i’d ask around to see if someone have any practical experience.
Let’s think about web stuff, let’s say we’ve got a very dynamic super ajaxified webapp that asks for various types of data every 5-20 seconds, our backend is python or php or anything other than java really… in cases such as these obviously a MySQL or similar db would be under heavy pressure (with lots of users), would MongoDB / CouchDB run this without breaking a sweat and without the need to create some super ultra complex cluster/caching etc solution?
Yes, that’s basically my question, if you think that no.. then yes I know there are several types of solutions for this, nodeJS/websockets/antigravity/worm-hole super tech, but I am just interested in these NoSQL things atm and more specifically if they can handle this type of thing.
Let’s say we have 5000 users at the same time, every 5, 10 or 20 seconds ajax requests that updates various interfaces.
Shoot ;]
OK, so to get this right, you’re talking about 250 to 1000 writes per second? Yeah, MongoDB can handle that.
The real key on performance is going to be whether or not these are queries, updates or inserts.
For queries, Mongo can probably handle this load. It’s really going to be about data size to memory size ratios. If you have a server with 1GB of RAM and 150GB of data, then you’re probably not going to get 250 queries / second (with any DB technology). But with reasonable hardware specs, Mongo can hit this speed on a single 64-bit server.
If you have 5,000 active users and you’re constantly updating existing records then Mongo will be really fast (on par with updating memcached on a single machine). The reason here is simply that Mongo will likely keep the record in memory. So a user will send updates every 5 seconds and the in-memory object will be updated.
If you are constantly inserting new records, then the limitation is really going to be one of throughput. When you’re writing lots of new data, you’re also forcing the index to expand. So if you’re planning to pump in Gigs of new data, then you risk saturating the disk throughput and you’ll need to shard.
So based on your questions, it looks like you’re mostly querying/updating. You’ll be writing new records, but not 1000 new records / second. If this is the case, then MongoDB is probably right for you. It will definitely get around a lot of caching concerns.