I'm building the system with handle million request from other vendors. The request from vendor, the system processing, call other services and then response.The problem I get that is our business rules are complex so our system down when getting too much request. Current ours can handle 20 ~ 25 request per second.
We using java vert.x, rabbitmq, memcache, mysql database, jooq flowing microservice.
Can you give some ideal to increase the request ?
The problem can be handled by building a cluster of computers and using it (queries are independent so parallel processing is useful). Such a cluster can be viewed as a private cloud. It also may be easier to use a public cloud services, get consulting how to do it and move your software to the cloud (Amazon, Google, Microsoft provide such services, for example).
No increase the scalability in terms of no of requests handle per second you have to do Mysql connection pooling in the application server then it can handle thousands of concurrent requests at a time.
The problem can be handled by building a cluster of computers and using it (queries are independent so parallel processing is useful). Such a cluster can be viewed as a private cloud. It also may be easier to use a public cloud services, get consulting how to do it and move your software to the cloud (Amazon, Google, Microsoft provide such services, for example).
It's pretty clear that you need to scale the MYSQL backend as it seems to be the bottleneck, so I suggest you take a look at what other people have done to do this. One example is Facebook, which supposedly handled 60 million queries per second with MYSQL. I put a GigaOM link that refers to this below, although a key link in it seems to be dead. I've also included a link at highscalability where a MYSQL guy describes how to scale up to 200 million queries per second. It seems you'll need to so some kind of sharding to get paralllelize the MYSQL backend. This stackoverflow question and answer that I have provided a link to gives some pointers on sharding, as well as mentions the highscalability web site, a good place to see descriptions of website that have really scaled.
In the longer run, you might consider going NOSQL on the backend, but it seems that might not be an option in the shortrun.
https://gigaom.com/2011/12/06/facebook-shares-some-secrets-on-making-mysql-scale/
http://highscalability.com/blog/2015/5/18/how-mysql-is-able-to-scale-to-200-million-qps-mysql-cluster.html
http://dba.stackexchange.com/questions/33578/how-to-scale-when-database-gets-huge
Dear Nhan
Unfortunately I can not help. from what I understand you want the best system. It involves knowledge of algorithms and networks.
I hope that others can contribute.
Hugs
Fabio
personally i would rethink my architecture (distributed computing, concurrency), maybe a actor based computing system is a viable alternative. Check out akka.io (or frameworks integrating it..lagom,play,spray...), with this approach you can scale out horizontally.
You can handle millions of requests if you have server with proper configuration. by default mysql create 151 connection and it exceeds upto 100000. so my opinion is that put your application on cloud hosted server.
- Badges
- Science topic