3 ms·

I'd worked on a couple of Mongo DB projects a year and half ago. It could not scale for performance, it was supposed to be fast datastore for simple lookups and

I'd worked on a couple of Mongo DB projects a year and half ago. It could not scale for performance, it was supposed to be fast datastore for simple lookups and heavy updating (multiple 10s of updates/sec - spiking to hundreds of updates) (no matter what we did), eventually had to switch out to other tools (either inbuilt custom ones or off the shelf). I won't claim that we did everything under the sun to optimize it, we could have probably optimized it to work eventually. It was just easier to switch and not deal with the headaches of maintaining/optimizing the mongoDB setup after days/weeks of headaches.

danpalmer 12y ago
I'd agree that MongoDB has its fair share of performance issues, but there are many companies using it at scale in production. Trello comes to mind as a good example of a high traffic web application that is doing pretty well on it so far from what I've read. This suggests there are things that can be done to improve performance enough, and in that case, maybe switching would be a bad idea. There is certainly a very high cost in terms of development time, learning and dev-ops to switching between two databases.
- thedufer 12y ago
  I work at Trello. I believe we're seeing ~1000 updates/sec and ~10000 reads/sec at peak. That said, our data is split across 7 shards. Total documents count is in the low billions.
MichaelGG 12y ago
>heavy updating (multiple 10s of updates/sec - spiking to hundreds of updates) In what world is 100 updates/sec "heavy"? I don't mean to come off rude, but even with a full disk write per update that fits into normal magnetic disk IOPS without even batching transactions. For comparison, with zero work on our end, and running off a 2-disk RAID 1 array with a 10 or 15K HDD, it was trivial to pass 10K/sec updates.
- gaius 12y ago
  It is heavy for MongoDB. It's a weird phenomenon you see these days, people talk about 10 updates a second on a million records as some sort of intense load that needs "sharding" and God knows what else, when a DBA from 20 years ago wouldn't even blink at it.

danpalmer 12y ago

I'd agree that MongoDB has its fair share of performance issues, but there are many companies using it at scale in production. Trello comes to mind as a good example of a high traffic web application that is doing pretty well on it so far from what I've read. This suggests there are things that can be done to improve performance enough, and in that case, maybe switching would be a bad idea. There is certainly a very high cost in terms of development time, learning and dev-ops to switching between two databases.

thedufer 12y ago
I work at Trello. I believe we're seeing ~1000 updates/sec and ~10000 reads/sec at peak. That said, our data is split across 7 shards. Total documents count is in the low billions.

MichaelGG 12y ago

>heavy updating (multiple 10s of updates/sec - spiking to hundreds of updates) In what world is 100 updates/sec "heavy"? I don't mean to come off rude, but even with a full disk write per update that fits into normal magnetic disk IOPS without even batching transactions. For comparison, with zero work on our end, and running off a 2-disk RAID 1 array with a 10 or 15K HDD, it was trivial to pass 10K/sec updates.

gaius 12y ago
It is heavy for MongoDB. It's a weird phenomenon you see these days, people talk about 10 updates a second on a million records as some sort of intense load that needs "sharding" and God knows what else, when a DBA from 20 years ago wouldn't even blink at it.