Last week, a friend of mine saw my post about MongoDB and got in touch with me. He is working on an IoT product. The company that helped build the software recommended using the load balancer for higher availability during peak usage.
But their AWS billing was too high. So, he was curious to know whether it was normal and wanted to validate it. We got onto a call; I tried to understand the application and did a short inspection.
Their MongoDB Atlas deployment had 0 indexes in the DB. I navigated to the page where MongoDB Atlas recommended the simple and compound indexes and told him to monitor it for a week.
Previously where the server utilization was a minimum of 40-60%, it is now just 15 to 20%. And yesterday, when the server traffic went high, the maximum server utilization was 35%. The load balancer did not create an additional instance at run time.
So the issue was on the database side and not really at the EC2 instance. Without those indexes, the DB used to take more time to process those queries on a huge database. Hence, servers were not able to process so many requests at a time.
Adding indexes in DB will save 50-60% of his cost.
The load balancer has become a buzzword these days for tech solution providers. However, it is a great feature and does make sure high availability.
But if there are some issues in my underlying configuration, and if I choose a load balancer as a solution, then I am not solving the real problem.
0 Comments