Scalable Human Blog

In the dynamic world of cloud computing, cost efficiency and resource management are paramount. Jaime Dantas, Hamzeh Khazaei, and Marin Litoiu, researchers from York University, have developed the BIAS Autoscaler, a novel autoscaling solution that effectively integrates burstable instances to optimize cloud systems. This blog post explores their innovative approach, explained in their study “BIAS Autoscaler: Leveraging Burstable Instances for Cost-Effective Autoscaling on Cloud Systems.”

What is Amazon EC2?

Amazon Elastic Compute Cloud (EC2) is a web service that provides scalable compute capacity in the cloud. It is designed to make web-scale cloud computing easier for developers by offering a simple web service interface that allows you to obtain and configure capacity with minimal friction. EC2 gives users complete control over their computing resources and lets them run on Amazon’s proven computing environment.

Google Cloud Platform (GCP)

Similar to Amazon EC2, Google Cloud Platform offers scalable and flexible computing services. Google Compute Engine, the IaaS component of GCP, provides virtual machines that run on Google’s infrastructure. Like AWS, GCP offers various machine types tailored to different use cases, including burstable types which are ideal for workloads with variable CPU demands.

Types of EC2 Instances

EC2 offers a range of instance types optimized for different use cases:

• General Purpose: Balanced CPU, memory, and networking resources for web servers, code repositories, etc.
• Compute Optimized: Ideal for compute-bound applications that benefit from high-performance processors.
• Memory Optimized: Designed for tasks that process large datasets in memory.
• Storage Optimized: Suitable for workloads that require high, sequential read and write access to large datasets on local storage.
• GPU Instances: Specialized for graphic-intensive applications and machine learning inference.
• Burstable Instances (T2, T3, T3a, T4g): Perfect for workloads with intermittent bursts of CPU usage that occasionally need to burst above baseline performance levels.

What are Burstable Instances?

Within the EC2 family, burstable instances are designed to provide a baseline level of CPU performance with the ability to burst to a higher level when extra cycles are available. They are ideal for applications like microservices, small and medium databases, and development environments that have intermittent CPU usage.

Understanding AWS Burstable Instance Families: T2, T3, and T4g

AWS offers various families of burstable instances, each tailored to specific performance and budgetary requirements:

• T2 Instances: The first generation of burstable instances, offering a cost-effective solution for workloads that don’t use the full CPU often but occasionally need to burst.
• T3 Instances: More advanced than T2, with an “unlimited mode” that allows the instance to burst for any period of time as needed for an additional charge.
• T3a Instances: Similar capabilities as T3 but use AMD processors, offering a cost-effective alternative.
• T4g Instances: The newest generation powered by AWS Graviton2 processors, which are based on Arm architecture, providing the best price performance for burstable applications.

Performance Across Cloud Providers

The BIAS Autoscaler, as evaluated on Google Cloud Platform, demonstrates significant improvements in cost and resource efficiency. However, results may vary between different cloud hosting providers due to differences in the implementation of burstable instances, underlying hardware, and pricing models. This variability highlights the importance of choosing the right cloud provider and instance type based on specific workload requirements.

Key Concepts in Cloud Autoscaling

Transient Queuing: Refers to temporary increases in workload that require additional computing resources to maintain performance levels. Effective autoscaling must quickly adjust resources to meet these demands.

Flash Crowds: Similar to transient queuing but more abrupt, involving sudden surges in demand often due to specific events, stressing the need for immediate resource availability.

Research Findings on BIAS Autoscaler

The BIAS Autoscaler, as evaluated on Google Cloud Platform, demonstrates significant improvements in cost and resource efficiency:

• Cost Reduction: By integrating burstable instances, the autoscaler reduces costs by up to 25% compared to traditional autoscaling methods.
• Resource Efficiency: Increases resource efficiency by 42% without sacrificing service quality, thanks to its intelligent resource allocation strategies.
• Maintained Service Quality: Despite the lower resource usage, the autoscaler maintains comparable service quality, ensuring no compromise on performance.

Conclusion

The study by Dantas, Khazaei, and Litoiu provides compelling evidence of the benefits of using burstable instances within an autoscaling framework. This approach not only reduces costs but also enhances the scalability and efficiency of cloud services, though the actual performance can vary based on the chosen cloud platform.

Credit

This blog post is based on the research conducted by Jaime Dantas, Hamzeh Khazaei, and Marin Litoiu from York University. Their work provides groundbreaking insights into the use of burstable instances for cloud autoscaling.

For more detailed information, refer to their publication on the BIAS Autoscaler.

📚 Further Reading & Related Topics

If you’re interested in autoscaling and optimizing cloud resources, these related articles will provide deeper insights into managing scalability efficiently:

• Load Balancing Algorithms Every Developer Should Know – Explore key strategies for distributing traffic across cloud instances to enhance scalability and efficiency.

• Kubernetes Helm: Simplifying the Deployment of Your Applications – Learn how Helm can assist in managing and scaling Kubernetes applications, complementing burstable instance strategies.