MongoDB Is Web Scale

Attention: the "humor" tag.

In conclusion. We came to the conclusion that MySQL is an excellent database for our site. Questions?

Yes, I have a question. Why didn't you use MongoDB? MongoDB is a horizontally scalable database, it does not use SQL or JOINs, and therefore has high performance.
')
This is a great question. We studied a few NoSQL databases and realized that all the options are still immature for use on working projects. MySQL is a proven database that is used throughout the world and has all the functions we need.

But it does not scale. Everyone knows that relational databases do not scale because they use JOINs and write to disk.


Scalability is a complex issue and it is wrong to make a general statement like this.

Relational databases were not conceived horizontally scalable. MongoDB supports horizontal scaling. You can turn it on and the database is automatically scaled.

You may be surprised that there are a lot of very visited sites that use relational databases, and in particular MySQL.

Relational databases have interface packaging.

I think you meant mismatch .

MySQL slower turtle. MongoDB can run around MySQL because MongoDB is horizontally scalable.

MongoDB shows impressive results on tests, but they do interesting things to achieve these numbers. For example, when you write in MongoDB, you actually do not write anything (the text was written a long time ago, when it was by default - note. Lane ). You are postponing the data record for later. If there is a problem with writing your data, you're in a hmm ... unpleasant situation. Does this seem to be a good architectural solution?

If this is what you need to achieve such stunning performance, then this is a great architecture.

If you are so stupid as to completely ignore reliability just to show beautiful results, I advise you to send the data immediately to / dev / null. It will be very fast.

If devnull scales horizontally, I will use it. Devnull scaled horizontally?

Are you kidding me, huh? I was just joking. In a sense, if it suits you to write to a database that does not give you an understanding that your data is really written, just to get good performance, why not immediately write to / dev / null. He's damn fast.

Does devnull support sharding?

Tree sticks. For your own comfort, I will assume that you are only trolling me, and not really mentally retarded. Do you even know what a shard is?

Shards are a secret ingredient in a horizontal scaling recipe. They just work.

Please tell me that you are not making a living in IT.

I am a web programmer.

Everything, I officially refuse to continue to work as a developer and leave for the collective farm as an agitator for pigs and mainly for anal candles for sick horses, because it is a thousand times easier to endure than to work in the same field with such receptions as you. You have read the last post on Habré and now think that you, damn it, you can design Google, repeat like parrot phrases like “horizontal scaling” and “sharding”, but don’t understand what you are talking about. You will soon doom a project to failure, because you have an unhealthy riser when you play with programs like rubber dolls from a sex shop or are seriously trying to win a volcano in a casino. Relational databases have been around since the 70s and this is one of the most mature technologies you can find. And yet there are clever men who need to re-invent everything, because Google and Amazon have published a couple of studies. If you need to build a global distributed search engine that manages petabytes of data, well, create your own database. But, if you are the same as 99.9% of the companies, then MySQL will probably be enough for you and maybe another memcache.

Redis will kick ass memcache. It is so fast and scalable.

Okay, now I imagine how fun it will be to castrate the first bull there on the collective farm. I can't wait to try to cut off the testes of a half-toned angry bull as he tries to blow my head.

MongoDB is a document-oriented database that does not need a JOIN. It uses map / reduce.

Now I’m cleaning a cistern of chicken droppings, but I’m not complaining because it’s better than listening to NoSQL fans quoting a list of features from their favorite structureless database.

Using the mapping of files to memory, MongoDB can improve the write speed 10 times.

What the hell? Let's forget about transactions, consistency, reliability, get all the critical data and give them such a shake-up that they will never forget, not forgetting to invite the funeral band. In terms of who cares what we really save, as long as we do it quickly. Oh, stop. For sure. I'm on the farm choking on the stench of thousands of gas-emitting cows. But it is a pleasant aroma, because I am far from this idiotic conversation.

MongoDB uses atomic modifiers for data consistency without loss of performance.

So I picked up some deadly contagion, cleaning the cow shed from manure and bleeding. I will die soon, but this is a welcome relief. After all, I will not be witnessing the collapse of the global economy, when NoSQL evangelists will persuade financial institutions to leave well-functioning data stores because they do not support fucking distributed map / reduce.

MongoDB stores files of any size, without requiring the addition of new technologies.

Thank you for your questions. This presentation is over and I went to buy a train ticket to the collective farm to start a new career.