Scalability is when additional resources to handle the load and increase the yield investment (ROI) can be obtained when the system capacity.
First, the planning scalability
Second, in order to win time extensions
Optimize performance
Stronger performance of the hardware purchase
Third, the scale-up
Fourth, scale-out
Can be extended outwardly (or sometimes also referred to as a lateral extension extended horizontally) policy divided into three parts: copy, split, and a data slice (sharding).
The most common and easiest method is extended outwardly by copying the data to be distributed across multiple servers, the library is then prepared for reading query.
Another relatively common method is to extend outwardly workload across multiple "nodes."
1), a functional split
2), data slice
3) Select the partition key (partitioning key)
4), a plurality of partition key
5) across slice inquiry
6), data distribution, and fragmentation node
7), deployed on the node slice
8), fixed allocation
Allocate data to slice There are two main methods: fixed allocation and a dynamic allocation. Both methods require a partition function, such as the partition key value of the input line, the return line memory fragmentation. Fixed allocation using the partition function depends only on the partition key value.
9), dynamic allocation
Dynamic allocation, mapping each data element to be a fragment.
10), and the mixture dynamically assigned fixed allocation
11), explicit assignment
12), data slice rebalanced
13), generating a globally unique ID
14), slicing tools
V. multi-instance extension
Sixth, through the cluster expansion
Seven, inward extension