This chapter will first key-value stores, in fact, this is the principle of correspondence stored in the relational database schema based, general nosql is to use this, there is no obvious feature is the schema, is a great addition to the key lump of value, do not need to have any neat format.
Speak database, natural cap must mention the theory, that is to say Consistency Available and Partition tolerance, namely consistency (C), availability (A) and partitions fault tolerance (P) make up the impossible trinity, can only meet two.
Cassandra then spoke architecture, Cassandra write operation more convenient, written memtable (memory), the sstable thrown into the disk inside, searching after the memory is full, then using bloom filters, before studied this thing false positive, but nothing great relationship. Then copy sstable is not merge it to be consistent, of course, when inconsistency is likely to be queried.
Cassandra is to ensure that the A and P database that is fast, good scalability, but consistency is not so strong.
Specific can see here https://www.cnblogs.com/logo-fox/p/8927067.html .
Corresponding also says the hbase, but also nosql completely different principles and Cassandra, I basically did not actually listen to loud noise, then you can take a look at the review here https://www.cnblogs.com/bonelee/p/6279248.html . hbase C and P is guaranteed.
Then talking about the time consistency, it is how different the system time consistent ah, say the two algorithms, crstian and ntp, I totally heard. . . . The main hearing is behind the talk of lamport time stamp, this principle is that we do not care about the specific time, only care about the order before and after the occurrence of different systems in each transaction, the method is what happened to each process time-stamped (1, 2, 3 and the like), then the order can be drawn back and forth between a lot of different system processes the transaction by logical deduction, of course, there are many things to push it out, it would need to arrange artificial rules. Details here speaks good https://blog.csdn.net/aigoogle/article/details/38823175
This course feeling much better amount of content, information and too close together. . .