Brain Storm to sustain load by Software Systems

I am always been bothered by one of the Questions. How would one scale is j2ee application to cater to more and more load/user requests.

I suppose these kind of thoughts would pave paths to Design patterns.

In this context let me tell you some thing about different scaling techniques.

The Basic one: 

• Use the Firewall, Loadbalancer and couple of servers and manage to run your applications. Handle the static content to web-servers or Loadbalancers. DB servers, Internal cache mechanism
• If situation worsens then run profilers, change GC algorithms.
• still not in good form? resort to next  technique

Vertical Scaling:

• Add power to your machines in terms of CPU, RAM. This would incur more cost to the business. 
• if situation worsens then implement above steps and also think on mechanism to cache static content by using more load balancers , install squids with white and black lists etc.,
• DB servers scale up add power to them, so the full table scan is possible seamlessly with proper page size available for the DB query results. I am not convinced by the question people ask me ""Why can't we use Horizontal scaling for DB servers?". I am not sure, how can one scale the MYSql horizontally as these traditional DBs are not designed for that. Apache Casandra is kind of DB which perimits this horizontal scaling distrubutedly.
• DB partition. I will focus on DB partition in next blog. We can discuss on Row level , Vertical partitioning and Column level , Horizontal DB partition.
• still not in good form? resort to next technique

Horizontal scaling:

• Add more Machines/Servers to process requests.
• DB partition into READ/ WRITE SANs.
• Optimize code. 
• Partition you server prototype to different geo graphics. 
• Use User clusters to deploy the features of the app relevant only to those users
• Manage a Master DB Table to route which user belongs to which cluster.
• Still not in good form? resort to CLOUD the future.