One of the most important faces of Computing is coding. We all tend to code, code and code for the list of growing requirements all the time. And there's developmental testing coupled with unit, function, integration and regression tests to add to the load. If we're involved in a huge project requiring thousands of lines of code, the coding phase grows not just to the desktop phase but on a larger version. We eventually get into a place where a data-center (read a centralized database) and a web-server comes into foray of managing things. Things are getting complicated.

Switch to Server Computing. All work is now zoomed into the server as an epicenter with the changes leading to alterations (or repercussions) in the business logic. That's when two important layers of Computing can lend a helping hand - Scalability and Performance Tuning. Often, these are ignored in a small company due to extreme workloads, but the bigger ones, take them very seriously. We've seen dedicated teams of huge sizes for a single process handler, haven't we?

Scalability. Means of making servers handle more user/process requests from a meagre count of a thousand to upwards infinitium, process more tasks and manage databases in parallel. This leads to a bigger number of machines to operate. But the situation goes out of hand when there are too many processes on a single machine - computations, database interactions and request handling jobs, all working in parallel. If one process takes an extra millisecond to operate, the rest of the dominoes fall, server queue length overflows, web-server stop to respond, continuous incoming requests fall back loosing data and the system breaks. Scalability can be a huge factor where, the solution lies in expanding servers based on tasks. Segregation is a big relief to the main guy where he is off loaded with the entire process architecture. All he might be doing is load balancing the requests based on the request type and leading that task to the server handling that particular type of request. Database segregation is also mandatory - split the databases on a live record base based on a prefixed date boundary (live-caching) and transfer the older data to an archived database (records that are not that "useful" for the current objective, but are required for the business logic in longer terms), cluster databases for handling fetch requests and modularize based on growing time for insertions. Web-servers can also follow similarly - one handling complex computations for fetch requests asynchronously (read higher thread management), another processing computations for record-insertions synchronously. One is managing these two and returning responses back to clients directly. Dedicated load balancers can switch databases/servers based on load factors, maintenance schedules, replications and unaccounted down-times. The system grows in size, but is easily managed by a few. The swarm architecture, in short.

On the other hand, keeping in mind, the crunch in "costs" to the company which cut down on supplies, the challenge lies in the way of making a single (or two) servers, smarter - optimized process handling, reduced task overheads, dedicated job management and a scheduled auto-maintenance. The beauty of performance tuning might not be observable but can have seriously good results. Your server begins to get more robust, handles things better than traditional methods and lives up to it's life span in full. A simple example was of implementing an HTTP server to a Django project. The time taken to process incoming requests, zoomed to almost ten times compared to Django's pathetic inbuilt web-server handlers. Or take the case of database indexes. A secondary index on a table made life much easier by faster process of incoming POST requests. A factor of 20. Of course, there are trade-offs in these as we need more storage space but these can be minimal.

Balancing these two can be a challenge in itself. Of course, testing all of the above with some peer consultations and expert opinions can be a daunting factor. The kind of opinion you go with, can even lead to your rise or fall of your business, unless you pick more and more reading of them.