What's the quickest and many stable non-sql database to keep large data and process 1000's demands throughout your day (it's for traffic exchange service)? I have found Kdb+ and Berkeley DB. Could they be good? Exist other available choices?

More particulars...

Every day server processes > 100K visits. For every visit I have to read corresponding stats from DB, write log to DB increase stats in DB, also known as 3 procedures with DB per visit. Visitors are continuously growing. Thus DB engine ought to be fast. From one for reds DB is going to be handled by demon written on C, Erlang or other low-level language. From another side DB is going to be handled by PHP scripts.

The file system is faster and much more stable than almost other things. It stores large data effortlessly and effectively. The API really is easy.

You are able to store and retrieve in the file system very, very effectively.

As your question is a touch thin on "needs" it's difficult to express a lot more.

Wrong with SqlLite? Because you did clearly condition non-sql, Berkeley DB derive from key/value pairs that might not suffice to your requirements if you want to grow the datasets, even much more, how does one make that dataset connect with each other using key/value pairs....

However, Kdb+, searching in the FAQ online is really a relational database that may handle SQL via their programming language Q...bear in mind, if the necessity to migrate seems, there might be potential hitches, for example incompatible dialects or perhaps a query that utilizes vendor specifics, hence the possibility to obtain locked into that database and the inability to migrate whatsoever...something to keep in mind for afterwards...

You have to be careful that which you decide here and check out it from the lengthy-term perspective, future upgrades, migration to a different database, how easy will it be to up-scale, etc

One apparent entry within this category is Intersystems Caché. (Well, apparent in my experience...) Bear in mind, though, it isn't cheap. (However I don't believe Kdb+ is either.)

How about Redis?

http://code.google.com/p/redis/

Haven't check it out yet did find out about it also it appear to become a fast and stable enough for data storage. Additionally, it gives you a good anti-single-point-failure solution, so far as I realize.

Berkely DB is attempted and examined and hardened and it is in the centre of numerous mega-high transaction volume systems. An example is wireless company infrastructure which use huge LDAP stores (OpenWave, for instance) to process a lot more than 2 BILLION transactions daily. Scalping strategies also generally have something similar to Oracle within the mix too for time recovery, however they use Berkeley DB as duplicated caches.

Also, BDB isn't restricted to key value pairs within the simple feeling of scalar values. You are able to store anything you like within the value, including arbitrary structures/records.

MongoDB may be the quickest and finest nosql database. Take a look at this performance benchmark.