AWS EBS vs Instance Storage Patterns for Application Use Cases

When you are provisioning an EC2 instance in AWS have you seen there are two storage options available namely Instance Storage and EBS? Although Instance Storage option is available only for the high end EC2 instances, many fears in using them because of the volatility of instance storage.

What if when you stop an EC2 instance, the data stored in your instance storage is gone forever? This is one of the reasons many afraid in using instance storage over AWS EBS. However there are greater benefits in using AWS Instance storage for many different application workloads. It is also possible to deploy clusters of EC2 instances with Instance storage with high availability without loosing the data even a single EC2 instance fails or shuts down.

On the other hand there are various applications in using AWS EBS with different configuration options thats fits for various application workloads. Lets consider different configuration options using AWS Instance Storage and EBS volumes for different application use cases.

EC2 with Instance Storage

This is a common configuration pattern used for application that require high performance while reading data from a persistence resource, process and store the data in external storages such as databases or EBS volumes. One of the common application use cases is running Online Analytical Processing Workloads (OLAP) reading from a data warehouse and strong the results in a database.

It is also important to note that, while processing the data, if the instance is failed, it should able to begin from the start and process the data set again without conflicts. This needs to be handled by your application logic.

EC2 with EBS

This is one of the most common use cases in selecting storage while provisioning EC2 instances. For newbies this selection comes by default, since AWS EBS is the default choice for lower end instances. However there are more predictive behaviors in using AWS EBS with EC2 instance since the storage is persistent even after multiple stop starts of the EC2 instance which is more synonyms to your personal computer.

This is a common configuration pattern for web servers, database servers, file servers & etc which require durability and high availability in storage persistence.

EC2 Clusters with Instance Storage

Cluster configuration requires the support from applications itself, which requires reasonable knowledge and expertise in designing distributed systems. This is one of the reasons this pattern is mostly used in well established applications.

One of the popular example of this configuration pattern is for NoSQL databases which require high throughput and also durability and availability, which can be achieved through clustering.

In addition, it will also require load balancer of dns configuration for failover routing avoiding the single point of failures.

EC2 with EBS Auto Scaling Load Balancing

This is a scalable version of the EC2 with EBS pattern, where multiple EC2 instances. However one of the main differentiator is that, each EC2 and EBS units can contain the same version or different versions of the application depending on the state the application is running on. For example, upon a state while a deployment is happening, some of the EC2 and EBS might contain the new version of the application while the entire deployment happens in a scenario like rolling updates.

This configuration pattern is commonly used in Web Applications where high scalability is needed. There are different approaches to achieve this.

• Creating Snapshots of EBS and prepare Amazon Machine Images (AMIs).
• Use a base AMI and load the application from an external source like S3.

With either of the approach it is important to note that, the durability of EBS is not required since each EC2 and EBS instance should be disposable and recreated upon the need which differentiate this with single EC2 and EBS configuration.