decentralisation - distributed computing on a massive scale
TRANSCRIPT
DECENTRALISATIONDISTRIBUTED COMPUTING ON A
MASSIVE SCALELIVING WITH DECENTRALIZATION, P2P AND THE NEW WEB
Anton Whalley Aman Kohli and thanksto Alex McMahon!
@dhigit9 @akohli
presentation on github pageshttp://akohli.github.io/presentation-
decentralised/decentralised-presentation.html
THE INTERNETHOW NON COMPUTER PEOPLE THINK
IT WORKS
HOW IT WORKS
NOT REALLY
DEVICESFrom workstations and supercomputers (thousands)
To Desktop and Laptops in offices (millions)
... and homes (tens of millions)
to Mobiles, tablets and devices (hundreds of millions)
Now lots of things
AN INTERNET OF THINGS
billions
OK, HOW IT REALLY WORKS
KINDA, SORTA
IT'S A MESH OFA MESS
WORKED EXAMPLE
5 of friends want to arrange where to meet on Saturdaynight
1. when all friends are in the same room2. when all friends are in different places
think about distributed concensus and the network
CONCURRENCY
Concurrency sets of events which happensimultaneously. The real world is concurrent, andconsists of a large number of events many of whichhappen simultaneously.Parallel happens at the same time
Concurrent (two queues and one processor) and ParallelProgramming (two queues and two processors)
Systems may be distributed over several computers: asystem of a number of independent concurrent
processes.
BEWARE THE FALLACIES
1. The network is reliable.2. Latency is zero.3. Bandwidth is infinite.4. The network is secure.5. Topology doesn't change.6. There is one administrator.7. Transport cost is zero.8. The network is homogeneous.
... thanks wikipedia
CAPTHE THREE CHOICES
The CAP speaks of access to data, that data being in theright state and it is always possible to get to it
Consistencyall nodes have the same data state
Availabilitythe system can send and receive responses
Partition ToleranceThe system can keep functioning in the face ofarbitrary network failures
SYSTEM PROPERTIES
distributed processing: a network of distributedprocessorsisolation : Two processes operating on the samemachine must be as independent as if they ran onphysically separated machines. Without sharing,everything necessary to perform a distributedcomputation must be copied.Causal ordering: If A sends a message to B then B canonly receive this message at some point in time after Asent the message.
REWIND - THEWEB
A JOURNEY THROUGH WEB 1.0 - 4.0
THE WEB 1.0 1993
A computer, to connect to others to share research
WHY DID WEB 1.0 FAIL?
IPv4 - Limited Public IP Address RangeBandwidth - Dialup/ADSLDNS - Registration Non-trivialWeb 2.0 - Centralisation became good enough
WEB 2.0 POSITIVES
Browser Wars Mozilla vs GoogleCloser Browser ImplementationsEnhanced Client Runtime - HTML 5 - CSS3Large systems led to distributed systems theorymoving out of academia - PAXOS to RAFTBittorrent 20% of North American Traffic
TO WEB 3.0
Symantic Web lack of general-purpose usefulness that prevents therequired effort from being invested
Trust - THIS IS IMPORTANT Our ability to create our own gated networks willenable the next set of webs
TRUSTReputation Based TrustZero TrustPolicy Based Trust (NO THANKS)
https://gist.github.com/pfraze/e314196dcecd4c49382d
REPUTATION BASED TRUST
Belief Ratings For StatementsTrust Rating Between Agents
WEB 4.0 INTELLIGENT PERSONAL
AGENTS
Trust is now a solved problemIPv4 - Superceded by IPv6Bandwidth - Uplink and Downlink becoming largerDNS - Still the problem
THE DNS PROBLEMNot just a registration issueAttaches content to a locationThe current URL format https://gist.github.com/pfraze/e314196dcecd4c49382dShould be : e314196dcecd4c49382d/pfraze
IPFS - FIRST ITERATION OF WEB 4.0
Removes address from content
Leverages distributed systems knowledge from 2.0
Transport agnostic so more efficient transports than
HTTP can be used
Doesn't try to solve ALL the problems
https://ipfs.io/
ALTERNATIVE WEBSfirechat – Adhoc Chat Networkkwizzi – Zero internet eductionEbola Relief - field recording
CONCLUSION