Sure, I'll reply the emails in the other mailing list. Willem Jiang <willem.ji...@gmail.com> 于2018年8月15日周三 下午8:44写道:
> Hi Lance, > > I already already introduced the background of performance tuning here[1]. > We could improve the performance by leverage the redis cache without > looking up the events with the complex SQL query. > As the mailing list have some trouble to display the picture that you want > to use. > Maybe you can share a link of the document which the rough design you have. > > > [1] > > https://lists.apache.org/thread.html/df24000621c903d1847cca5b273d38b69e7aa6e44d1c2dbd9d0687ab@%3Cdev.servicecomb.apache.org%3E > > > Willem Jiang > > Twitter: willemjiang > Weibo: 姜宁willem > > On Wed, Aug 15, 2018 at 7:44 PM, Lance Ju <juzhenat...@gmail.com> wrote: > > > Hi, gang, > > I've made some tests and looked into the implementation of saga > > recently. Currently, the alpha-server relies on some complex SQL queries > > for compensation and status maintenance. I'm thinking if it is possible > to > > make some optimizations on this. We can replace the database with > redis(or > > any other mem db) to have a try, if it works, we can introduce a mem > cache > > between alpha-server and database, to improve the performance. > > > > We can store the Saga transactions with a key pattern and a hash > value > > in redis, scan the keys with pattern rather than complex SQL query, and > > maintain a status of the transaction. Here is a pretty rough design(any > > comment is welcome): > > > > [image: saga-redis.jpg] > > >