Unikernels are specialized virtual machines compiled from application code and only necessary operating system components. They provide benefits like reduced memory usage, faster load times, and a smaller attack surface compared to traditional virtual machines. Several unikernel implementations exist like MirageOS, Rumprun, and IncludeOS which are compiled from different languages and have varying boot times, image sizes, and hypervisor support. Unikernels are being applied to applications such as proxies, edge computing, and IoT gateways.
3. Methodology
• I am collecting
• Related articles indexed by Google Scholar for word unikernel (~170)
• Related videos indexed by Youtube for word unikernel from technical
conferences and summits (~821)
• Blog posts from the unikernel.org website.
• Papers that have cited the unikernel key paper (~115)
• Unikernels: Library operating systems for the cloud (ACM SIGPLAN 2013)
4. Unikernels: Library operating systems for the
cloud
• Library operating systems is a more general term
• Minimizing OS image to libraries that are needed to run an application.
• Lib OS has been used for parallel and distributed OS (Exokernel-1995)
• Lib OS has been used for jvm in a virtualized execution environment (Libra-2007)
• …
• The novelty of unikernel is that Lib OS is used for cloud deployment.
5. Definitions
• Unikernels are a technique to compile specialised applications that
run directly on a hypervisor instead of requiring an intervening guest
operating system.
• Unikernels are specialised, single-address-space machine images
constructed by using library operating systems.
• Unikernels are constructed by combining application code with only
the operating system components necessary for that application to
run.
• Unikernel is a highly specialized, single-purpose application which can
be deployed directly to the cloud or onto IoT-like devices.
14. Metrics to compare unikernel with other
technologies
• Memory footprint (B)
• Need for disk space (B)
• Computational burden (B)
• Schedulers (B)
• Code size (B)
• Attack surface (B)
• Development time (P)
15. Comparing unikernel with other
virtualizations
• Network Function Virtualization
• A comparison of containers and virtual machines for use with NFV (Master
thesis 2015)
• Networking latency and throughput and CPU utilization (Master
thesis 2016)
Extracted metrics
Comparisonpapers
17. Comparing unikernel implementations (no
paper)
• Year
• Boot time (Osv less than a second, ClickOs 30milliseconds, )
• Hypervisor (Xen, KVM, bare-metal)
• Language (Ocaml, C, C++, Erlang)
• Number of applications per hypervisor
• Image size (ClickOS 5MB, )
• Process isolation
• …
18. Applications
• Jitsu: Just-In-Time Summoning of Unikernels
• A power-efficient and responsive platform for hosting cloud
services in the edge network.
• SCANDEX
• Service Centric Networking for Challenged Decentralised Networks
• On-the-Fly TCP Acceleration with Miniproxy
• Miniproxy: a Xen unikernel, is as fast as state-of-the-art GNU/Linux-based
proxies, re- quires only 6 MB of RAM to run and can boot in just 12ms.
• The enterprise micro data center: small cell networks, IoT gateways
and edge computing servers
19. Organizations and projects and people
• Cambridge university
• IBM Linux Technology Center
• ALAN TURING INSTITUTE, oxford
20. Conferences to publish
• SIGCOMM
• IEEE/ACM International Symposium on Cluster, Cloud, and Grid
Computing
• USENIX
Notas do Editor
By combining the familiar tooling and portability of Docker with the efficiency and specialization of next-generation unikernel technology, organizations have a flexible platform to build, ship and run distributed applications without being restricted to a particular infrastructure.
Unikernels reduce software complexity by only including code that is required, resulting in portable applications with much smaller footprints and fast boot times.
Because workloads that reach the data center today are on a spectrum from physical machine to container to hypervisor, only the Docker platform can further widen the scope and provide more flexibility for orchestrating hybrid applications.