Freeing the IP Internet Architecture from Fixed IP Addresses

Abstract

Abstract-The IP Internet architecture is such that applications must bind fixed IP addresses and ports before any other operations can be executed. These early bindings cause bottlenecks, reliability issues, and force applications and protocols to manage complex lower-layer issues. This poses a big challenge to the future of the IP Internet, given the large and growing numbers of nomadic Internet users, the shift in Internet usage from centralized servers to peer-to-peer content sharing, and the popularity of service replication and virtualization. To address these issues, we introduce and evaluate HIDRA (Hidden Identifiers for Demultiplexing and Resolution Architecture), a novel architecture that creates indirection between layers of any network stack. HIDRA enables sockets and protocols to evolve with the IP Internet by hiding all mobility, multihoming, and multiplexing issues from applications, does not induce significant overhead in the protocol stack, preserves backwards compatibility with today's Internet and applications, and does not require or preclude any additional identifiers or protocols to be used in the protocol stack.

A Simple Solution to Scale-Free Internet Host Mobility

“…Indeed, part of the justification of recent ICN architectures is the argument that using addresses is inherently limiting in supporting the mobility of services or content. However, as we have argued recently [45], [46], [47], the limitations of prior proposals stem from the assumption that the identifiers used in the communication protocols operating in the data plane must be the same as the identifiers used within the protocol stack of a host or a router to pass information across layers of the stack. We have demonstrated that this does not have to be the case and that indirection within the protocol stack can be done efficiently to allow names and addresses used in communication protocols to differ from the identifiers used inside hosts and router to refer to resources, connections, or remote processes.…”

“…First, our prior work on "hidden identifiers" in the protocol stack [45], [46], [47] demonstrates that hosts and routers can denote resources, connections and remote processes with identifiers that need not be the same as those used as part of communication protocols. Second, updating the mappings from names to addresses need not be done solely through a common directory service, and indeed communicating hosts or routers can update one another directly if done properly.…”

Named-Data Transport

“…The use of end-to-end connections between specific end-point addresses prompted the development of many approaches to cope with mobility in the context of the IP Internet [9,33,43,65,82,86,91,93]. However, this prior work does not address eliminating connections for reliable end-to-end communication, and some can be used in NDT.…”

A deployable identifier-locator split architecture

“…This is an important consideration, especially because the average number of connections that a host has at any given point in time significantly outnumbers the number of separate hosts with which it is communicating, at a ratio of at least 4 to 1! [8] This has profound implications for the scalability of transport-layer solutions. Figure 13 illustrates control message growth for a single handoff as we vary the number of corresponding hosts from 1 to 100.…”

“…In response to the inherent problems with host-identifier proposals, recent host-locator proposals [5], [2], [6], [7], [8] split identifiers from locators above the network layer at end hosts. These approaches propose addressing datagrams to the current network location of a host (i.e., the locator), and updating the IP address bindings made by higher layers (typically at the transport layer) of end hosts by way of additional signaling or protocol options exercised when a host experiences mobility.…”