International telecommunication union
Use of ICN function state migration
Download 0.91 Mb.
|
- Navigate this page:
- 6.2.4 ICN removes host endpoint abstraction from application data
- 6.3 Overview of network softwarization
6.2.3 Use of ICN function state migrationICN is a good technology choice for function state migration and execution context migration in future 5G networks131415. Content Centric Networking (CCNx) facilitates process migration while enabling many desirable features such as strong checkpointing and data de-duplication. Not all migration techniques require strong checkpointing, and in those cases CCNx offers a faster and weaker naming technique that allows pages or blocks to go dirty in a checkpoint. CCNx offers an intuitive naming of resources that are part of a service or execution context migration and building checkpoints around those resources for consistent state transfer. De-duplication is a technique where only one copy of data exists and it is shared between multiple instances. CCNx allows resources to be de-duplicated both within and between virtual machine instances. For example, in the previous discussion about using hash names for resources, if two disk blocks, for example, have the same hash value they will refer to the same Content Object. Only the block index in the CCNx manifest will be different. A VM hypervisor may also share blocks between VMs. When generating the names used to fetch a checkpoint, the source migration agent running in the source hypervisor could use a name like {/nyc/host7, hash = 0x63223…} so any instance or any component can share the same data. Assume that the memory page size and the disk block size are the same. Then that name for hash 0x63223… could be both a disk block and a RAM page of the same data (e.g. a shared library code section). Because the manifest can point to different name prefixes for each hash and can indicate the virtual resource of that hash, we can have the same physical bytes used for many purposes. This approach may also be applied when page and block sizes are not the same by using smaller units of naming. Gap analysis
6.2.4 ICN removes host endpoint abstraction from application dataIn a traditional IP-based network, the Internet Protocol adds a level of abstraction to communications endpoints by assigning them a location-dependent name. When applications wish to communicate, they must rendezvous those host addresses – such as with DNS or SIP or some other well-known means – before communication can take place. Because the rendezvous is done outside the network layer, the rendezvous protocol must employ its own means to determine locality – such as with ping triangulation – to determine which replica to use. Some applications use IP anycast addressing to move rendezvous back in to the network layer and realize those benefits. CCNx, as an ICN protocol, naturally keeps rendezvous on addresses within the network layer so all applications can benefit from localized services without needing to add on additional rendezvous layers with their own localization protocols. ICN may be used as a de-abstraction layer for virtualized functions: using direct function naming in ICN means the network can move functions and change routing without needing to update intermediate abstractions of endpoint identification. For example, a single host IP address might hide many virtualized functions, so it may not be possible to directly move an IP address. One would need orchestration to inform components of a new socket endpoint, which could result in service interruption during the time when a function has finished migration and the time when an existing prior service is notified of a new service endpoint. With CCNx, the orchestration does not need to inform prior components of a new service address, it only needs to update the named routing to the new location. Because CCNx, as an example ICN protocol, is not tied to the P-GW identity – such as for the source endpoint address – it means that CCNx is well suited for multiple P-GW egress. Service frameworks, such as Mobile Edge Computing, could realize significant simplification by using a CCNx approach for multiple P-GW egress without needing to assign the UE multiple identities or using layers of address translation.
6.3 Overview of network softwarizationNetwork Softwarization is an overall transformation trend for designing, implementing, deploying, managing and maintaining network equipment and/or network components by software programming, exploiting the natures of software such as flexibility and rapidity all along the lifecycle of network equipment / components, for the sake of creating conditions enabling the re-design of network and services architectures, optimizing costs and processes, enabling self-management and bringing added values in network infrastructures. The terminology, Network Softwarization, was first introduced in Academia, NetSoft 2015, the first IEEE Conference on Network Softwarization, to include broader interests regarding Software Defined Networking (SDN) and Network Functions Virtualisation (NFV), Network Virtualization, Mobile Edge Computing, Cloud and IoT technologies. Directory: ITU-T -> focusgroups focusgroups -> International Telecommunication Union focusgroups -> Innovation of ict in Developing Countries focusgroups -> International Telecommunication Union focusgroups -> Telecommunication focusgroups -> Focus Group on Smart Grid focusgroups -> Telecommunication focusgroups -> Telecommunication focusgroups -> Telecommunication focusgroups -> Innovation of ict in Developing Countries focusgroups -> Innovation of ict in Developing Countries Download 0.91 Mb. Share with your friends:
|