What is Auto-Scaling in SaaS?

Auto-scaling is a feature that adjusts a SaaS application’s resources up or down based on current usage. When there are more users, auto-scaling increases available resources – on the flip side, it decreases them during quiet periods. Kubernetes’ Horizontal Pod Autoscaler, for example, checks usage every 15 seconds by default.

SaaS companies need auto-scaling because they often don’t know when traffic is going up or down. While it’s sometimes predictable (e.g., a product launch), sudden global events and unexpected spikes are where auto-scaling should be deployed.

Reasons to use auto-scaling in your SaaS business include:

• Performance: Use auto-scaling so that your site or app works optimally during busy periods; without it, you could encounter service disruptions.
• Availability: Auto-scaling supports the reachability of sites, products, and services as demand fluctuates, while a load balancer manages traffic distribution among auto-scaled servers.
• Cost Optimization: Use auto-scaling to scale up and down based on real-time usage – you risk needless expenses if you use an alternative method.

Auto-scaling is triggered by cloud performance metrics, checked at intervals (every 15 seconds by default in Kubernetes), such as:

• CPU Utilization: The amount of CPU being used at any given time.
• Memory Usage: How much memory an app or website is using in real-time.
• Network Traffic: Data received and sent over the network.
• Request Latency: Response times to user requests; this partly depends on how many users are on the same server.
• Queue Length: Pending queue requests; auto-scaling up is triggered when this is higher.

Different auto-scaling policies work in varying scenarios. Here’s a list of some common ones and when you should use them:

• Target Tracking: Moves resources based on a metric like average queue length; AWS strongly recommends it for EC2 Auto Scaling (AWS).

• Step Scaling: Scales within thresholds you set, in steps sized to the breach; use when target tracking is not enough.

• Scheduled Scaling: Scales based on scheduled times; use if you know daily or weekly traffic patterns.

Examples of SaaS companies that use auto-scaling based on demand are:

• Shopify: Runs microservices on Google Kubernetes Engine, which Google says scales during traffic spikes such as Black Friday (Google Cloud).

• Slack: Autoscales its web app tier quickly when workers become saturated (Slack Engineering, 2020).

• Netflix: Auto-scaling parameters are set to manage traffic when more viewers are online (e.g., evenings, as a Netflix patent on predictive scaling filed in 2013 describes).