My team has just launched an iphone application with the server backend. The architecture is something like this; user interacts with the iphone app and the iphone app sends the request to the server. server performs some calculation on the request and sends back the response. the server is involved in each interaction of the user. There is no caching has been done on the server, neither there is any session handling..
My question is that, do we need to configure a load balancer for such type of application, if the active user of the application is ranging from 30,000 to 50,000? currently the cloud server has been used and the database and apache server are on the same machine. the size of the cloud server is 2GB. Will it be sufficient if the cloud server has been re-sized from 2GB to something greater 10GB or 15GB instead of using the load balancer?
Another question is that if there are any good tools available for ubuntu which could help in montioring the memory, request/response of the server?
Thanks
There are at least two reasons you might need a load balancer:
You can (and should) answer the first question by performing a load test. Have a look at the Apache JMeter project for a great load testing tool.
If you cannot tolerate an outage, you need at least two web servers behind a load balancer that can individually handle all of the traffic (or if you have, say, 3 web servers, each should be able to handle about 60% of peak traffic to void the risk of losing a single web server).
Regarding monitoring, you really should ask separate questions separately. I personally use Zabbix for monitoring. It is open source (not “crippled community” and “enterprise” like some solutions). The learning curve is slightly steep (and some things changed with the recent 2.0 rollout that are not well documented) but overall it is a solid solution. Nagios is also very popular and worth having a look at.