Jacoby Thwaites
Inventor, designer and implementer at Google
Google Inc
Since 1990 Jacoby Thwaites ran his own company in the UK architecting large-scale networks and enterprise messaging systems. Around 2000 the company got into writing and selling insurance systems in the London Market. He then co-founded a Silicon Valley Web 2.0 startup which folded a couple of years later, and most recently joined Google in 2008 to set up the London AdSense engineering team. He has spent the last 6 months inventing, designing and implementing this state machine as his Google 20% project.
Jacoby Thwaites is Giving the Following Talks
A prototype state machine “MadCloud” for distributed applications
Distributed applications often use client-server techniques across a TCP/IP infrastructure. This talk presents a prototype infrastructure that abstracts the client out of programmers' code and can optimize inter-server traffic across datacenters. It is written in Erlang with an admin UI that is quite pretty because HTML5+CSS3+XSLT+JQuery rocks actually.
Whatever. The infrastructure is very simple, and works by sequencing calls on application-level servers following some initial event. It does this by collecting sets of fields, sending subsets to servers and collecting replies where appropriate, until no further work can be done. The result is, application backends that are declarative.
My feeble hope is that this opens up pathways including applications constructed by search, mutable applications that change many times a second, large-scale caching of server operations, applications that self-render and other things I’ve not thought of.
The target Audience is anyone dealing with distributed applications and enterprise network and/or messaging infrastructure; including architects, system reliability engineers and programmers.
Whatever. The infrastructure is very simple, and works by sequencing calls on application-level servers following some initial event. It does this by collecting sets of fields, sending subsets to servers and collecting replies where appropriate, until no further work can be done. The result is, application backends that are declarative.
My feeble hope is that this opens up pathways including applications constructed by search, mutable applications that change many times a second, large-scale caching of server operations, applications that self-render and other things I’ve not thought of.
The target Audience is anyone dealing with distributed applications and enterprise network and/or messaging infrastructure; including architects, system reliability engineers and programmers.
