Virtualization
From FIpedia
The Disagreement
The abstraction paradigm finds its origin in object-oriented programming. Applied in the context of distributed/networking environments, it has lead to the so-called virtualization concept and its multiple variants: virtual nodes/networks, virtual links/paths, virtual resources, etc. and their combination in order to e.g. "overlay" functionalities on top of current network layer. Similarly to object-oriented programming, the driver was to abstract actions (associated to the control/processing) and data structures (associated to the information exchanges) to allow offering desired processing and messaging capability without incurring complexity to entities external to the system (abstraction exposition).
Nevertheless, when applied in distributed environments, factoring out details of the individual components requires an additional network-wide level of indirection in order to abstract reachability of/accessibility to the components of these virtual entities. For this purpose, an identifier-to-locator resolution (or distribution) like-mechanism isrequired to allow communication between "users" and virtual entities as well as between virtual entities (safely assuming here that actual forwarding is still performed on locators).
Food for thought: "Any problem in computer science can be solved with another layer of indirection... but that usually will create another problem." —- David Wheeler
On the other hand, there is still no well identified Internet challenge for which virtualization would be a suitable answer or solution: this concept would enable resource aggregation (composition) as well as re-usability and re-location of dedicated resource shares (slicing). Note: this leads to the side of how a given set resource units can be shared and dedicated during the same time interval ?. Both aggregation and slicing find since so far limited applicability in large-scale networks such as the Internet that assumes that any resource (unit) is continuously shared (What is the actual value of virtualization ?). For instance, past experience in overlaying capabilities for multicast traffic distribution on top of the Internet were not successful because of dynamic state processing/mainteance as well as processing of host-triggered messaging.
Discussion points:
- What is the future of so-called virtualization (and other overlay-based routing) beside experimental environments that benefits from (global) resource scheduling ?
- Which Internet challenges are addressed by so-called virtualization ?
- What is the value/utility for the end-user/hosts ?
- How to perform rate x state (performance) benchmarking when positioning virtualization as THE enabler of excellence ?
Viewpoints:
"Virtualisation of networked service infrastructures: Virtualisation is an important driver enabling the delivery of core infrastructure services as a 'utility-like service' to the business layer independently from the underlying platform. Virtualisation techniques will provide an illusion of infinite capacity of the networked service infrastructure and "elasticity", i.e. the capacity can be scaled instantaneously to the demands."
... would certainly deserve discussion and debate ! "an illusion of infinite capacity..."
Viewpoint 1
Summary
Viewpoint 2
Summary
Moderator
Dimitri Papadimitrou
