~/topics/Federated Architectures
Updated Feb 26, 2021

📐 Federated Architectures

The recurring theme of federation in distributed system

Some editorial choices about this deck:

• Entertaining > Precise
• Cool > Useful
• Future > Present

(Presenter notes available in the source of these slides)

IRL Federation

• United Federation of Planets 🖖
• United States 🇺🇸
• Most companies (teams, franchises, branches) 🏭

Federation: organization pattern involving a combination of centralized decision-making and as peripheral autonomy

Hybrid of top-down (hierarchical) and peer-to-peer (decentralized)

Federation in Computer Systems

The abstract pattern of Federation is everywhere.

Often in disguise, or called with by different (more precise) names

• DNS
• E-Mail
• Kernel
• CPU architecture

Federation in Distributed Systems

Concept of federation is broad overlaps with many concepts we encounter everyday:

• Gateway, Router, Facade patterns
• Orchestration pattern
• Microservices & decentralized systems

Federation is abstract and overlaps with these and more.

In the rest of this deck: practical examples of federation in action

Data federation

GraphQL & Apollo

“GraphQL is a query language for APIs and a runtime for fulfilling those queries with your existing data”

Below the API, query federation.

Apollo stores no data, it federates the request to other systems.

Data layout, location, provider can change without affecting users.

Search federation

Maps Search

Different queries require very different handling, separate indices, ranking, spell-correction model, and more.

• “Starbucks”
  • Nearby locations of a specific chain
  • Or maybe coffee shops
• “1200 Main Street”
  • An address (that exists in hundreds of towns!)
  • Show the closest?
• “1600 Pennsylvania Ave”
  • A nearby address?
  • Or a very famous landmark? (the White House)
• “Paris”
  • Paris, France?
  • Paris, Texas?
  • (many other)

Search federation

Maps Search Architecture

A collection of small specialized search engines.

• Category search
• Address search
• Contacts search
• Businesses search
• Landmark search
• Geographical entities search
• …

And auxiliary services

• Spelling suggestions
• User metadata and history
• Driving time estimation
• …

Federated storage

FileCoin

Planetary scale decentralized file storage system. Utilizes the blockchain to track blocks and reward people for storing them.

Mine (get paid) to offer storage capacity Pay others to store, serve, replicate, …

Similar: Threefold (P2P compute grid), LivePeer (video encoding grid), and many more

Federated infrastructure

GAIA-X

European Union next generation cloud.

Similar to AWS in that it offers a variety of services, from bare metal machines, all the way up to high level services.

AWS is centralized, GAIA-X is federated through and through. Only offers the basic protocol for providers and consumers to find each-other and negotiate terms.

• Federated resource catalogs: compute, storage, bandwidth, …
• Federated services catalog: high level functions, services and API, …
• Federated data catalogs, with varying degree of access, licensing, …
• Federated identity and trust: to regulate and delegate access to data and resources

GAIA-X Use case example

Medical records

Each citizen has a virtual medical record.

It is stored safely in the cloud.

• The patient can grant access to a new doctor
• The doctor can add to the medical record (and retain access to the newly added data)
• the patient can delegate access to a parent or guardian
• Research institution may have limited access to aggregate data for statistics and research purposes
• Insurance, prospect employer, etc may request partial access

Federated Service Platforms

Uber, AirBnb & co successfully connect demand and supply at scale.

But their primary driver is profit. Above all they need to appease shareholders and investors.

Making users happy is just a secondary goal. A mean to an end. The real goal is: make money.

Re-imagine these platforms without the shareholders (or making users the shareholders). Same service, different goal: optimize for the collective good of all users.

More similar to public utilities than to for-profit companies.

Built on open-source (trust is paramount).

Still collect a fee on transactions, but use it to:

• Pay for infrastructure costs, insurance
• Compensate employees (support, developers, legal, …)
• Reinvest surplus into making the platform better or give fractional dividends to all participants

Examples: Eva, FairBnb, WindingTree

Federated Social Networks

Facebook, Reddit, Twitter, Instagram, etc. Are great for staying in touch with friends, discuss current events, and more.

But satisfying user’s need for connection is just a mean to an end.

The goal of these platform is to make money to shareholders. They are not shy of using dirty tricks to get users to:

• Reveal private information, which they can harvest and monetize
• Get people addicted, enraged, misinformed, anything required to keep them engaged

User’s interests (such as privacy) are only considered if ignoring them is damaging to the platform.

A new generation of federated social platform are emerging: Diaspora, Mastodon, Mobilizon, and more.

Federated Communications

Matrix

WhatsApp, Telegram, Discord, iMessage, GMail, … provide a great service, but at a (hidden) cost. Your data is being mined and sold (end-to-end encryption is a marketing ploy).

Matrix is a federated, decentralized answer to this problem. An open-source protocol for instant messaging (and, communications in general).

Built ground-up to be federated. Small groups of people can operate their own homeserver, sharing the cost.

Through the homeserver, they can connect to the rest of the world (or not).

Further reading:

• Beyond distributed and decentralized: what is a federated network?
• The Ultimate Guide to Federated Architecture & Decentralized Social Networks
• Decentralized Social Ecosystem Review