Heap structured Patricia Trie - topling/PureOld-TerarkDB GitHub Wiki
Specific optimization for highly intensive writing work-load
Writing Perspective :
Name | Writing Time Cost | Writing Throughput |
---|---|---|
Patricia Trie | 17.47 sec. | 55.08 MB/sec. |
Skip-list | 72.58 sec. | 13.23 MB/sec. |
Reading Perspective :
Name | Get Time Cost | Seek Time Cost |
---|---|---|
Patricia Trie | 220.49 sec. | 285.22 sec. |
Skip-list | 241.74 sec. | 245.63 sec. |
**Hence Patricia Trie's seek operation contains sub-operation with quantity of active tries, the seek throughput rate of Patricia Tire were being counted as average and documented, per se.
The basic idea behind this optimization is not complicated in any perspective, frankly it's quite primitive. By deviding the single task into multible tasks, we have been able to scale down the task and make them working concurrently, resulting the leap of writing speed during heavy writing work-load.
From the reading prespective, as it's not the major feature during this situation. We merely remain one-one like radio as the original skiplist memtable does.