Self-Organizing Adaptive Networks Hari Balakrishnan MIT Laboratory for Computer Science
Some Trends Heterogeneous data types –“Multimedia”, telepresence, games Wireless and mobility –Lots of new technologies Embedded computers, devices, sensors Networks everywhere, network everything!
Imagine... Servers Services in our environment E21
Challenges Heterogeneity –Devices, hand-helds, mobiles, sensors –Wireless & wired links –Variety of applications, data and services Dynamism and spontaneous operation –Need rapid deployment and organization Highly decentralized & fault-tolerant –Must make management easy! Need self-organization and adaptation
Current View of Networks Routers DNS Hostname Address Mostly static topology & services Applications cannot learn about network High management cost Failures happen! Servers Clients
Towards the Future Providers Efficiently configure topology & match providers with consumers Learn and adapt well to heterogeneity and network conditions Efficiently configure topology & match providers with consumers Learn and adapt well to heterogeneity and network conditions Handheld Sensors People Coffee Consumers Servers Entertainment
[building = ne-43 [room = *]] [entity = thermometer] [temperature < 62 0 C] data [state = ma [city = cambridge] [region = kendall]]] [service = café] [distance < 0.25 miles] data Intentional Naming System Names are query expressions –Attribute-value matches –Range queries –Wildcard matches Names express desired properties; they are intentional [building = ne-43 [room = 504]] [entity = camera [resolution=800x600]] [access = public] [status = ready] Names are descriptive –Providers announce names
Naming Architecture [building = ne-43 [room = 510]] [entity = camera] Intentional name INR Intentional Name Resolvers form a distributed overlay Integrate resolution and message routing image Lookup camera510.lcs.mit.edu
Details Separate into inter-domain and intra- domain components Architect intentional names around virtual spaces Robust name resolution & routing based on soft state Dynamic views to track dynamic data attributes
Benefits Mobility (of people and services) Service location and resource discovery Group communication Server replication & data caching Device networking –Take efforts like Jini TM and Universal Plug & Play TM to the next level
Adaptation Overload causes congestion; wireless links often degrade performance Solution: new approaches to congestion management and wireless protocols In-Building Campus-Area Packet Radio Metro-Area Regional-Area + Asymmetry Network heterogeneity Application heterogeneity
Security and Privacy Moving from a packet-centric abstraction to a service-, device-, and people-centric one Domains define trust regions –But all name resolvers cannot be trusted –Location information may need to be private –Data has to be secure (easier) Challenge: privacy without compromising performance or function
Research Areas Intentional Naming System End-to-end adaptation framework –Congestion Manager –RadioActive networks –“Better than best-effort” networks Security and privacy protocols Lots of new applications
Applications WIND: Wireless Networks of Devices Ad hoc collaborative environments Adaptive sensor applications Self-configuring network topologies Adaptive applications (e.g., audio, video, telepresence) Service location & resource discovery Dynamic replica placement and selection Home network architectures
Self-organization middleware and adaptation protocols are key to making this a reality Self-organization middleware and adaptation protocols are key to making this a reality The Future: Networks That Learn Better application and user control over networks Dynamism, mobility, spontaneity, robustness and adaptation Integration with the real world