1. The most obvious and simplest of the most convenient optimization is the SSD, the general optimization efficiency of 3 times times, (not necessarily, but explained that many bottleneck problems are storage problems)
2. The vertical expansion of the development time, short-term is the fastest, the disadvantage is that it will consume more resources, and there are bottlenecks, in addition, if the application itself has a lot of pits, the effect may not be good.
3.
A. Reduced call chain 16 steps, TPS increased from 90 to 800tps
B. Storage replaced by SSD, up to 1500
C. Remove the transaction and raise it to 1800
d. Optimize table space and increase to 2100
E. Change the UUID to sequence, convert the random number to sequential write, raise to 4000
F. Static and dynamic separation, archiving data at rest, preserving active data, and adjusting db configuration parameter write buffer to 10000
4. To visualize the bottleneck of the system and expand the intelligent Knowledge Base
5. When time is tight and resources (capital) are adequate, upgrading hardware capabilities is the quickest solution to performance. In a nutshell: 1000 the following can be achieved through application optimization and simple hardware optimization; implementing 5000+ requires a high-performance storage solution, and implementing 10000+ requires a sub-library table. Of course, time, technology allows to advance the Sub-Library sub-table rather than rely entirely on the storage performance of the upgrade. In short, the optimization of system performance depends on the specific circumstances, various constraints, can not be generalize.
Finally, to give you a performance design of the ultimate proposition: Single account entry and exit 10000+tps how to do
Notes on Java Web optimization