Cassandra

Cassandra is by default an AP (Available Partition-tolerant) database, hence it is “always on”. But you can indeed configure the consistency on a per-query basis.

my questions

how did cassandra ensure data security
- https://cassandra.apache.org/doc/latest/cassandra/managing/operating/security.html
- TLS/SSL encryption for client and inter-node communication
- Client authentication
- Authorization

challenges

full multi-primary database replication
global availability at low latency
scaling out on comodity hardware
linear throughput increase with each additional processor
online load balancing and cluster growth
partitioned key-oriented queries
flexible schema

basic knowledges

keyspace (database)
table
partition (primary index)
row
column

storage engine

logging data in the commit log
writing data to the memtable
flushing data from memtable
storing data on disk in SSTables

commit log

memtable

SSTables

SSTables are the immutable data files that Cassandra uses for persisting data on disk. SSTables are maintained per table.

data.db - contents of rows
partitions.db
rows.db
index.db
summary.db
filter.db - bloom filter
CompressionInfo.db
statistics.db
digest.crc32
TOC.txt
SAI*.db

summit keynote

messaging summary

each node starts a gossip round every second
1-3 peers per round
3 messages passed
constant amount of network traffic

pratical implications

who is in the cluster?
- gossip with a seed on startup
- learn all peers
- gossip
- lather, rinse, repeat
how are peers judged UP or DOWN
- what does UP/DOWN mean
  - local to each node
  - determined via heartbeat
- failure detector
  - glorified heartbeat listener
  - records timestamp when heartbeat update is received for each peer
  - keeps backlog of timestamp intervals between updates
  - periodically check all peers to make sure we’ve heard from them recently
- UP/DOWN affects
  - stop sending writes (hints)
  - sending reads
  - gossip
  - repair/stream sessions are terminated
when does a node stop sending a peer traffic
when is one peer preferred over another
- dynamic snitch to rank all peers’ latency
when does a node leave the cluster

Query Structure

cql

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17


-- Create a keyspace
CREATE KEYSPACE IF NOT EXISTS store WITH REPLICATION = { 'class' : 'SimpleStrategy', 'replication_factor' : '1' };

-- Create a table
CREATE TABLE IF NOT EXISTS store.shopping_cart (
userid text PRIMARY KEY,
item_count int,
last_update_timestamp timestamp
);

-- Insert some data
INSERT INTO store.shopping_cart
(userid, item_count, last_update_timestamp)
VALUES ('9876', 2, toTimeStamp(now()));
INSERT INTO store.shopping_cart
(userid, item_count, last_update_timestamp)
VALUES ('1234', 5, toTimeStamp(now()));

misc

configuration

cassandra.yaml

source code

to hack the source code locally, we had to use ant (to build the project), also to generate files by using command ant generate-idea-files, after this, the IDEA starts to learn how to source the libraries.