Obviously, for example, ali cloud disk I/O no

CodePudding user response:

. Lz refers to improve disk I/O? If from raising disk, namely raid nas, SAN and core is divided, with many blanks to improve I/O
Or direct use of SSDS provide large capacity ram cache cache algorithm to improve the throughput, lz said ali should adopt the conventional equipment, slicing algorithm scheduling algorithm part should be no big problem, main or network latency and equipment itself the limitations of personal opinion

CodePudding user response:

Ali would be to use multiple disks (logical disk), concurrent access to data; Before is the raid technology, etc. Is now a cloud storage, eventually I/O is similar approach should be; Do store choice is often encountered time (how long data is actually written to disk) and memory space (kernel or automatically to cache);
High-performance storage devices, za not met, storage devices, for the purpose of this company do ordinary disk throughput between the theoretical value is 120 MB to 180 MB/S, the actual use of time, is to reach the effect, unless the cache space is large enough; So the total feeling cannot be ruled out in double tenth caton, the disk I/O problems;