1. Performance Comparison of Virtual Machines and Containers with Unikernels
Nagashree N
Suprabha S
Rajat Bansal

2. AGENDA 1 2 3 4 5 6 7 8 Introduction Unikernel: How is it different?
Key Advantages 4
Architecture 5
Types of Unikernels and Ecosystem Elements 6
Applications/ Use-Cases 7
Performance Analysis - Results 8
Limitations
CONFIDENTIAL

3. Introduction 3
CONFIDENTIAL

4. Unikernels
“Unikernels are specialized, single-address-space machine images constructed by using library operating systems.”
Traditional OS
CONFIDENTIAL

5. Key Advantages Performance Security Scalability No GOS
Less CPU usage and memory footprint.
Security
The application is the kernel. Unikernels provide extremely tiny and specialized runtime footprint that is very less vulnerable to attack.
Scalability
- The image size is much smaller (less than 100MB), enabling flexible deployment on a large scale.
No GOS
Flexible, speedy and versatile for cross-platform environments, big data analytics and scale-out cloud computing.
Extremely tiny, specialized runtime footprint.
Less vulnerable to attacks. Minimal Attack Surface.
Massive scaling
Performance
Saves money for the customer.
CONFIDENTIAL

6. Unikernel: Architecture
Source: Wikipedia
CONFIDENTIAL

7. Unikernel Operating Systems
Unikernel Project
Language
Platform
Use case
MirageOS
OCaml
Xen Project
Safety & Security
HalVM
Haskell
LING
Erlang
Performance, Security, read-only filesystem
ClickOS
Network devices
Speed
IncludeOS
C++
KVM
N/A
RumpRun
POSIX compliant
Compatibility with Legacy Applications
OSv
General Purpose
KVM, Xen, VMware, VirtualBox
CONFIDENTIAL

8. Ecosystem Elements JITSU Rump Kernels Xen Project Hypervisor UniK
“Just-in-Time Summoning of Unikernels”
Rump Kernels
provides the modular drivers from NetBSD in a form that can be used to construct lightweight, special-purpose virtual machines
Xen Project Hypervisor
Its para-virtualization capabilities allow unikernels to have a very small and efficient footprint interfacing with devices.
UniK
utilizes a simple Docker-like command-line interface, making developing on unikernels as easy as developing on containers.
CONFIDENTIAL

9. Applications/ Use Cases
Banking Services
Sessions as a unikernel service
Secure transaction
Defense Systems
Resource Consumption
Network Firewall
RAM requirement
Boot up time
IoT
Every event served as unikernel service
CONFIDENTIAL

10. Performance Analysis - Results Requests/Second (Higher is better)
Apache Web Server
Performance Metric VM
Container
Unikernel
Requests/Second (Higher is better)
16.45
30.12
28.12
Time/Request (ms)
Total Test Time (s)
54.23
31.230
28.53
CONFIDENTIAL

11. Performance Analysis - Results (contd…)
Maximum Time Taken to Connect, Process Request & Wait for response
X: time(ms)
Y: metric
CONFIDENTIAL

12. Performance Analysis - Results (contd…)
Time Taken to Service the Requests
X: service (%)
Y: time(ms)
CONFIDENTIAL

13. Performance Analysis - Results (contd…)
MySQL: Read-Only and R/W Requests per Second
X: Requests (/s)
Y: metric
CONFIDENTIAL

14. Performance Analysis - Results (contd…)
Time Taken to service 95% of Read-Only and R/W Requests
X: service (%)
Y: time(ms)
CONFIDENTIAL

15. Performance Analysis - Results (contd…)
Performance Analysis of Applications with VM, Container & Unikernel
Scenario
Application
Requests/Second (Higher is better)
95% Service Time (ms) VM
Container
Unikernel
Database
MySQL
(sysbench)
(R), (RW)
(R),
(RW)
(R),
(RW)
2.05 (R),
3.63 (RW)
1.91 (R),
4.03 (RW)
0.72 (R),
4.14 (RW)
Web Server
Apache
(ab)
16.45
30.12
28.12
CONFIDENTIAL

16. Limitations Single Process Single User Impoverished Library Ecosystem
CONFIDENTIAL

17. Conclusion Performance Analysis
Applications demanding Predictive Performance – Unikernels offer direct access to hardware.
Specialized Implementation – drivers, libraries and interfaces.
DR, Transient Micro Services
CONFIDENTIAL

18. Thank You

19. OSv Application Build Process
CONFIDENTIAL