so I have an application that is processing a large text document which is composed of many logical documents. Say for eg it has 200,000 logical documents. Right now I read one logical document and do the transformation and write logical document out. This works well because nothing is stored in memory so the wor
开发者_JAVA技巧kflow ends up finishing fast but the number of IOPS end up being high! Is there any recommendation on how I can reduce the number of IOPS ?
If this is a single pass through the document, increase your read buffer size (i.e. 2 reads with a buffer size of 1024 becomes 1 read with a buffer size of 2048, which means less overhead).
If you want more detailed suggestions, you'll need to provide more details in your question.
继续阅读:architectureperformance
精彩评论