Scalability: How to Infinitely Scale Software

Categories
Aug 2018
comment-big-icon.png   0
Share
Author : Pete Peranzo
Scalability-How-to-Infinitely-Scale-Software.png

Your awesome web app or website is working so well that more people are using it. Now, it’s becoming slow. Users aren’t as happy with the service and therefore aren’t recommending it to others. If you want to keep growing, it’s time to scale.

What it Means to Scale
Scalability is developing software that is capable of growing as needed, usually in terms of data and bandwidth. That is how many users access the system and how much data is moving through the system (in and/or out). Consider the typical scenario where a new application starts with a few thousand records and a handful of users accessing that data, then going to millions of records and hundreds of users within a few months. After a year, that number could be tens of millions of records and millions of users. Can your system handle the growth?

Perhaps use the famous Twitter scale problems and the challenges faced by IoT where instead of a million users over a few years of operations to millions of devices from day one (such as we had at Reveal). Every company needs to build applications with scale in mind.

If your service is slowing down, it’s probably time to scale. At Imaginovation, we build apps that scale as needed from day one. But if your program was not built with scaling in mind, you may find challenges when doing so.

 

Methods of Scaling
How you scale a program depends on what it does, what you’re trying to do, and how it was built. There are many approaches to scaling. The one you or your team chooses depends on the problem. So the first step is to find out what’s causing your application’s slow down. Once you figure that out, you can choose a method of scaling that addresses the specific problem. Selecting the wrong approach means adding unnecessary complexity.

At Imaginovation, we may use any of the following to scale your application:

  • Horizontal Scaling – This refers to using a load balancer, which will manage memory, sessions, cache, and more in such a way to handle traffic in parallel — with multiple requests at once. This is the best approach to scaling and is proven by giants such as Facebook and Google.
  • Caching – One straightforward way to ease the burden is to cache responses to the most frequent or slowest loading queries. Chances are, there is a list of most commonly requested items. If your data rapidly changes, this might not work, but for many projects, this is the obvious first step. Be sure, too, to cache static resources for web applications.
  • Duplication – One way to scale is to make an exact copy of your program. You often see this with websites and web services. In those cases, developers will create the same site on several servers and use a load balancer to manage traffic.
  • Create subsets – If you’ve heard the term “sharding,” that’s a way of scaling. It refers to creating subsets or splitting data that is the same. One of the easiest examples of this is with postal codes or letters in names. One such set in a database is called a shard.
  • Splitting functionality – Depending on the application, it might be scalable by dividing it into different services, so you end up with two programs, each with a separate function.
  • Combination – You can also scale by mixing some of these approaches. For example, you might split functionality into microservices that perform specific tasks, then duplicate each of those to handle the volume.

You might also hear about vertical scaling. To scale vertically (at least in software terms) means to add more hardware. This might mean more RAM or more nodes. However, this is a Band-Aid, and we don’t rely on this method for our projects.

Tips for Scaling

If you’re working on scaling your application, consider the following tips to expand your app efficiently:

  • Write clean code – Watch out for spaghetti. Keeping your code as clean as possible will make it easier to scale now — and again in the future.
  • Future-proof your database – Use shard-safe IDs and design schemas that are not limited (ensure they can be split up). Instead of direct access to databases, use an API so that database connections can easily be converted to cache copies as the need arises. Make your software architecture in such a way that it can scale by duplication. In other words, simply by adding more hardware, your software can take advantage without the need for changing code.
  • Consider your cloud provider – Some hosting providers offer auto-scaling features and other tools to help.
  • Update – This goes along with maintenance, but be sure to update every part of your system as needed.

 

If you need to scale, talk to our team about the best method.

 

Write your comments

Your email address will not be published. Required fields are marked *