In many Enterprise System architectures, the size of the equipment becomes necessary in accordance with concurrency and workload requirements. Basically, product manufacturers will provide their own equipment calibration sheets, in which you simply plug in metrics, and they will expose details of the number of servers required by RAM, etc. However, when a system is developed using open source software components, such recommendations are not available.
I need to do a hardware calibration of the system. The system is horizontally scalable, so the growing volume of traffic can be supplemented by adding more nodes. To begin with, the system will have two HAProxy load balancing nodes. Behind the load balancers are web servers serving HTML pages and javascripts. For RESTful web services, there are more nodes hosting the servers. There are several Cassandra nodes on the server. We have a rough idea of traffic or concurrent users.
What should be my approach to the size of the hardware? Are there general recommendations?
, , . , , .
, , , .., , . , HTML, , OLTP-.
. , ASP.NET 12 /CPU ( config), concurrency, 3000 / , Ethernet .. , , ..
, . , .
, , POC , .
, - , cpacity - .
, - .
, 1000 , , . , , . VM, , P2V, . , , , , , para / , , , , , . , , . , . 1000 , , , (: DR, , : - ) , , . , // VM- .
, , , , .
, , Xeon, Nehalem, - .