Managing Diversity to Manage Technological Transition

Technological transition is one of the hardest questions in architectural studies. This paper examines how technological transitions take place in the Internet. In the first half of this paper, we sort out the patterns of technological transition in the networking industry, and hypothesize that diversity is requisite to cope with an “unplanned” technological transition. At the same time, however, excessive diversity would be harmful since it could cause chaos or uncontrollability. In the latter half, we measure some diversity trends of Internet Architecture, and observe a phase mismatch between layers. One can assume that such a phase mismatch helps to avoid excessive diversity as a whole, and helps the sustainability and evolvability of the Internet. 1. Patterns in Technological Transition 1.1 Difficulties of Technological Transition Technological transition is basically hard. There are a number of reasons for this. First, successful technology generates a positive feedback loop, which promotes successful dissemination even more. For example, the more the number of users, the more benefits of the users increase with less cost. Also an eco-system would easily be established due to a bandwagon effect. Second, people using a paradigm recognize the environment and conditions as a matter of course, and put an implicit preconditions. It makes them put effort solely for the improvement or extension to the existing technologies and make it difficult to find new technical possibilities. Classic books give us a perspective on these difficulties: “The scientific paradigms preceding and succeeding a paradigm shift are so different that their theories are incommensurable — the new paradigm cannot be proven or disproven by the rules of the old paradigm, and vice versa.” [1] “What all sustaining technologies have in common is that they improve the performance of established products, along the dimensions of performance that mainstream customers in major markets have historically valued. Most technological advances in a given industry are sustaining in character. An important finding revealed in this book is that rarely have even the most radically difficult sustaining technologies precipitated the failure of leading firms.” [2] “To be truly innovative and competitive in today’s world, the team that created and built a presently successful product is often the best one for its evolution — but seldom for creating its replacement.” [3] And needless to add, it is easy to find such examples in this industry, i.e. various lock-in effects to the existing systems, including the difficulty of migration to IPv6. Having said that, regardless that people hope or not, technological transitions happens. Technologies do vary across the ages. 1.2 Patterns in Technological Transition In this chapter, we sort out the patterns of technological transition in the networking industry. To start with, we may need to separate “planned” transition from “unplanned” transition, since the designing/engineering methodologies should be different. In a “planned” transition, a community has a common objective, predefines a value of the posttransition status, defines the scope of the system, and thus applies what amounts to a set of top-down system design/engineering methodologies. Examples of such “planned” transition include 3GPP cellular/mobile systems. 3GPP unites 6 telecommunications standard development organizations (ARIB, ATIS, CCSA, ETSI, TTA, TTC), and constantly publishes technical specifications called a “release”. The 3GPP technologies from these groups are constantly evolving through Generations of commercial cellular/mobile systems. [4] On the other hand, it is sometimes hard for a community to have a common objective, to predefine a value of the post-transition status or the scope of the system. In that case, “planned” transition rarely happens. The latter case is the case for the Internet. It is difficult to “plan” the transition, due to its bottomup emergent culture and its architectural principles. “Nobody owns the Internet, there is no centralized control, and nobody can turn it off.” “Its evolution depends on rough consensus about technical proposals, and on running code.” “Engineering feed-back from real implementations is more important than any architectural principles.” [5] “A grass-roots solution seems to be the only means to the success. Top-down mandates are powerless.” [6] So we tried to identify “unplanned” transitions happened in the industry, and observed the following patterns. (1) Swing-over between Opposing Concepts: An underlying construct for a preferable technology of the period tends to swing-over between the opposing concepts, for example: • Stateless <-> Stateful • L3 <-> L2 • Distributed <-> Centralized • Connectionless <-> Connection Oriented • Overlay <-> Hop by Hop • Virtual <-> Physical • Dynamic <-> Static • Separated <-> Integrated • Loosely-coupled <-> Tightly coupled • Etc.. (2) Conflict, Co-existence and/or Selection: There may be multiple technologies that achieve similar objective. Such technologies could compete, co-exist, or be selected over time. Examples include: • X.25 -> Frame Relay -> ATM • ATM -> Switched Ethernet • GOSIP -> TCP/IP • ISIS, OSPF • AppleTalk/DECnet/IPX -> IP • VXLAN, NVGRE, STT • DS-Lite, 4RD, MAP-E.. • Etc.. (3) Disruption: Disruptions happen from outside of the community. In 1995, Toshiba Research started work on what eventually was published as RFCs 2098 and 2129 – a cell-switched router, which used a basic ATM infrastructure to connect an IP network, and optimized it with dedicated cutthrough VCs for large data flows. The following year, Ipsilon commercialized the idea. Cisco and IBM proposed an alternative model called “Tag Switching”, which was eventually standardized as MPLS. The effect was to give large operators the ability to traffic engineer their networks, which IP Routing was not designed to do, and the way IP Routing works in those networks fundamentally changed. The recent Openflow boom could also be seen as a disruption. • IP Switch/MPLS • Openflow/SDN Figure 1 shows these technological transition patterns found in the networking industry. Figure 1: Patterns in Technological Transition Unplanned( ((1)( (Swing0over(between(( opposing(concepts((