Internet Mapping : from Art to Science

2008

Value to me : * The tuple space idea is interesting 
			  * Co-ordination of data is key 
			  * Lacking any this Africa (to my knowledge and to this scale, only 1 ark node :/ ) 
			  
			  
Paper's Angle : Architecture of a distributed system that performs monitoring and taking measurements designed to deal with
				things like parallization, distribution and co-ordination so that researchers concerned about the measurements
				themselves don't have to concern themselves with those issues. 
				
General : 

-> Dependence on the Internet (yadada) 
-> Comparision of the internet to a trillion dollar ecosystem (interesting, I kind of like the idea) 
-> Some data points suggest doubt in the ability for the internet to act as the worlds communication medium 
-> Lack of data sharing due to distributed ownership 
-> Project to support large scale active measurement studies of the internet
	-> secure
	-> analysis/processing 
	-> annotation
	-> topology generation
	-> interactive visualization 

Archipelago 

-> is Caida's new measrument infastructure -> skitter based active measurement infrastructure
-> 31 "Arks" (only one in africa ...) 
-> Plans to deploy more in africa/south america
-> most monitors are in academia (plans to have some organization etc)
-> Promotes rapid and easy dev -> high productivity
	-> more "risky" experiments/measurements -> possibly better results
-> Scamper 
	-> steesr measurements
	-> general purpose management engine 
		-> handles parrelized traceroutes
		-> talk about other stuff specific to their platform
		
Dynamic and Co-ordinated measurements

-> At simplest a measurement infrasture executes pre-configured set of measurements to a static set of targets
-> Some measurements have dynamic components and co-ord among measurement nodes
	-> prefix monitoring to detect for prefix hijacking (as an example) 
-> Major focus on co-ordination 
	-> planning, co-ordinating and executing a series of computations
	-> allows for measurement components to work well together 
-> New implementation of Ark for co-ord -> Miranda
	-> tuple space co-ordination model (ref to orginal paper is there)
		-> basically distributed shared memory with simple ops
		-> stores tuples and retrival of tuples through pattern matching (sounds quite strange ...) 
		-> supports 1-1 and many-many communication
		-> decentralize measurements (those at actual nodes) will automagically communicate with centralized control 
		   when needed and more cause further measurements dependent of results
		-> Simple to use + don't have to worry about network faults etc in deployment
		-> Discuss how tuples arent addressed to a particular recipreant (and alot of other complex stuff) 
		-> Above allows to break complex measurements into phases

Measurement Services

-> Section mainly discusses ways that ark provides services 
-> XML-RPC and SOAP are traditional 
-> Tuple space allows for same sort of thing (see the ping example, pretty neat idea)
	->rest of section discuesses advantages of this approach

Macroscopic IP Topology 

-> Measures IP Level paths to dynamically generated list of IP addies covering all /24 prefixes in routable ipv4 space
-> Due to distribution its possible to get all the traceroute info for 13 monitors to probe 7.4 million /24's at 100pps
-> Do lots of random probing of /24 space
	-> avoid possible routing probelms 
	-> avoid complaints
-> Make use of scamper for trace routes
-> ICMP Paris Traceroute (something to look into) 
-> Use a bulk DNS Lookup service due to the number of lookups required (Scapy) 
	-> Two datasets
		->Simple IP-Hostname map
		->Raw DNS Query and the responses (clever...)

Alias Resolution 

-> is reconstructing the router topology from trace routes
	-> requires grouping ip's belonging to the same router together 
-> CAIDA IFFinder tool 
	-> Paper describes how this tool works ( it is rather hectic)
-> Should add more here when I review paper

-> The rest of the paper deals with AS internet topology maps, while interesting isn't very relevant