The document compares the performance of Linux containers and virtual machines (VMs) in terms of scalability and efficiency, highlighting virtualization technologies and their benefits. It concludes that containers significantly outperform VMs, being 22 times faster in scaling and better suited for application deployment, though potential security concerns exist for business-critical data. The discussion includes different virtualization methods, their advantages, and a detailed performance testing setup using Docker and AWS EC2.

1. Performance Comparison Between
Linux Containers and Virtual
Machines
Venue: 2015 International Conference
on Advances in Computer Engineering
and Applications

2. Agenda
• What is virtualization?
• Virtualization technologies
• Comparing the technologies
• Conclusion

3. What is virtualization?
• virtual machine (VM) is an emulation of a
particular computer system.
• VM is a common way to run different services in
the cloud.
• Main benefits of virtualizations are:
– Portable deployment (fast application delivery)
– Scale and deploy with ease
– Hardware independence
– Isolation
– Secure user environment
– ….

4. Virtualization technologies
• Hypervisor-based virtualization
– Kernel-based VM (KVM), VMWare
• Container-based virtualization
– Docker, LXC
• Hybrid virtualization
– ZeroVM, OSv

5. Hypervisor-based virtualization
• Operates at the hardware level
• Isolates VM from the underlying host system.
• Benefits
– Abstracts hardware
– Support running VMs independently and
isolated from the hardware (windows VM can
run on Linux host)
Problems
• A full OS runs on the VM=>
emulation of VM has huge
overhead.(big images and long
start-up time)

6. Container-based virtualization
• Implemented in Linux kernel via
– control groups(resource management)
– namespaces(restricted access to resources)
• Benefits
– A light weight alternative to hypervisor-based
– implement isolation of processes at the OS level
– Disk images are smaller
– A higher density of virtualized instances
Problems
• Window container cannot run on a Linux host
• Resource isolation is not as accurate as
hypervisors (issues for multi-tenant security)

7. Comparing the technologies-
performance comparison
• In this paper, Docker is used for running
containers.
• The set up:
– Two physical servers (OS: Ubuntu14.04, RAM: 2GB) for hosting
docker and VM.
– front-end= application server hosting Joomla
– Back-end= database server hosting Postgres
– Jmeter is used for stress testing Joomla by sending
simultaneous requests to Joomla.

8. Application performance comparison

9. Comparing the technologies-
Scalability comparison
• For container scalability test, use Kebernetes cluster manager to scale the
the Docker host on the 2 physical servers.
• For VM scalability test, The application is deployed in an AWS EC2 with
auto scaling enabled.
• The set up:
– AWS EC2 for cloud and a physical server to host docker.
– Application = a load balanced WordPress application
– Jmeter is used for stressing and trigger the scaling

10. Scalability comparison
The time to scale for deploying another container (~8
sec.) is much less than the time to scale for deploying
another virtual machine (~3 mins).
=> 22x times faster scalability

11. Conclusion
• Containers out performed virtual machines
both in terms of scalability and performance.
• Containers are better for application
deployment.
• However, using containers for applications
with business critical data may reveal security
breaches.