Principles for Coping with the Evolution of
Computing
Computers collaborate in the Internet much the way cells collaborate in multicellular organisms and the way organisms compete and collaborate in ecologies. What are the parallels and what can we learn from them?
Single cell organisms evolved into multicellular organisms a billion years ago. Computing has made a similar transition in just a few decades. In 1975, few computers communicated directly with others and there was no Internet. Today tens of billions of computers around the world exchange information at Internet speeds. The role that interacting computers play has changed dramatically as their costs dropped and their numbers exploded. They entertain us, help us shop, help manage the lights in our homes, help us communicate with and befriend each other, enhance our memories, recognize our faces, 'understand' our speech, and 'talk' to us. Yet computers know not what they do, nor 'know' who they do it for, or why!As the digital world inexorably becomes more complex it encounters problems common to all complex systems – problems already solved in the evolution of living systems. This website explores the challenges of complexity and discusses architectural solutions for the problems inherent in increasingly complex systems.
In the late 1960s a handful of computers in universities and research labs began to be connected together in a persistent high-speed network called ARPANET. A descendant of that network eventually became the Internet. In 1990 Tim Berners-Lee put the first Web page up on the open Internet. The web was born and it grew rapidly -- today there are over 1.5 billion web pages to choose from. An isolated computer is an oddity. Computers surround us, they are in our pockets or purses, are on our wrists and in our cars, houses and offices. Google, Amazon, Microsoft, Yahoo, Baidu (China's Google-like equivalent) and many other less well-known organizations spider/crawl the web for various purposes. Google led the way to monetizing all the data thereby gathered by selling ads. They stored the contents of all the websites they crawled on large numbers of servers and developed quite sophisticated algorithms to characterize the web pages and sites. They then used proprietary "page rank" algorithms to decide which pages to recommend for various searches. It is estimated that google employs over a million servers for such tasks.
Prior to 2000, corporate data centers tended to be housed in the firm's basement. Dedicated "server farms" didn't emerge until the late '90s during the dotcom bubble.
Today servers are located in more than half a million server-farms
around the world (photo shows a part of Facebook's server-farm in
Lulea, Sweden). The digital world inexorably becomes more and more
complex. It records our emails, phone calls, eCommerce purchases,
searches and social media interactions. Facebook also analyzes all
this data for hints about our buying preferences, our opinions and
even for the identity of those that appear in our photos. Google's
many server farms around the world spider virtually all web-pages
cataloging their content so that it can recommend the page it
judges to be what we are looking for when we browse the Web. Other
servers at data centers owned by the likes of Amazon, Switch,
Microsoft, Twitter, and the NSA gather and store different sorts
of data for many known and unknown purposes. At least one, Cambridge
Analytica specializes in analyzing individuals' political
viewpoints to influence votes. There are millions of servers that
store, catalog, and make searchable information about people,
products, businesses, real estate, government activities,
universities, weather, crops, livestock, and almost anything else
one can imagine (see History of
Computing).Server farms handle the big-data issues in the Web. The Internet of Things (IoT) deals with the small things: smart door locks for the home, wireless cameras, smart electric plugs, smart AC and heater vents, wearable exercise monitors, smart refrigerators, baby-cams and even smart pill bottles. Yet collectively they provide more compute power and wifi capability than we could have imagined a few years ago. Most of their processing is wasted in idle loops -- so far. But various parties seek to harness the collective IoT compute power for their own use with or without permission. The number of IoT devices increased 31% to 8.4 billion in 2017, 14.2 billion in 2019 and there are expected to be 20 billion devices by 2020. They are problematic because they are very poorly protected from hackers. They are sold with simple default passwords and all too many people see no reason to change those passwords or, if they do change the defaults, the new passwords are all too often simple and easy to guess. So the hackers who seek to create large botnets out of IoT devices have found it easy. On October 12, 2016, a new botnet appeared called Mirai that nearly took down the Internet. And the code for it was put out on the net. In January 2018, a Mirai variant called iTroop or Reaper was used to target three large Financial institutions. In multicellular computing terms, botnets can best be thought of as Internet cancers.
Another facet of computation that poses problems for society is the rise of cryptocurrencies. Venezuela, among other countries, is talking about making bitcoin their national currency. Cryptocurrencies are one use of Blockchain legers. IBM, State Street Bank, Accenture, Fujitsu, Intel and other heavyweights formed The Hyperledger Fabric project around the end of 2015 to formalize and harden the notion of blockchains. They recommend isolating the ledger from the general cloud computing environment, building a security container for the ledger to prevent unauthorized access, and offering tamper-responsive hardware, that can shut itself down if it detects someone trying to hack a ledger. That is analogous to biological apoptosis (see below)
Finally, machine learning is becoming increasingly popular. Machine learning has already been adopted by many well-heeled parties. There is already specialized machine learning hardware running 7x24.
The evolution of computing is similar to the evolution of other complex systems -- biological, social, ecological, and economic systems. In each of these domains, the elements become increasingly more specialized and sophisticated, and they interact with each other in ever more complex ways. From that perspective, the similarities between biology and computing are not coincidental.
Multicellular computing already is adopting four major organizing principles of multicellular biological systems because they help tame the spiraling problems of complexity and out-of-control interactions in the Internet. They are:
- Specialization - Multicellular systems support much richer functionality than single cell and single computer systems. They do it by the collaboration between specialized cells. There are, for example, about 250 specialized types of cells in humans. Unspecialized cells such as many cancer cells are dangerous because they don't "play well with others." Specialization in computing is important for similar reasons. We are finding that unspecialized general-purpose computers, especially unprotected IoT systems with ARM processors, are increasingly dangerous to multicellular computing systems.
- Messaging - Cooperating cells or computers must communicate safely with one another. The "meaning" of cell-to-cell messages must be determined by the receiving cell, not the sender. To that end, cells communicate with each other via messenger molecules, never DNA. Similarly, communication between computers in multicellular systems relies increasingly upon message passing. Here again, the receiver not the sender determines the meaning of the messages. It is especially dangerous for computers to transfer code from one machine to another precisely because code predetermines the resulting behavior of the receiving machine, and all too often contains malware. Code endangers the health of the receiving machine and hence the larger system of which it is a part.
- Stigmergy - Messages orchestrate cooperation in real-time. Longer term or more persistent collaboration requires longer-lived messages that can be deposited in some structure where they can be later encountered by others. This sort of messaging has become known if the field of biology as stigmergy. Analogously, stigmergy in computing systems is supported by persistent external data, e.g., in databases, and network connectivity structures. Persistent data are especially important for organizing cooperative computing in the Internet.
- Apoptosis (Programmed cell death) - Despite all precautions, cells and computers do go awry. They may also simply outlive their usefulness, Every healthy cell in a multicellular organism is programmed to commit suicide if their removal is in the best interest of the organism as a whole. Cells that are infected by viruses or become cancerous usually kill themselves unless the programmed cell death mechanism itself is compromised. We are only recently learning the importance of sacrificing compromised computers for the health of the whole multicellular computing system. The recent appearance of the Internet of Things (IoT) and botnets exploiting IoT devices creates, in effect, out-of-control computing cancer cells that do not kill themselves (activate apoptosis) when compromised.
This site explores these principles in considerable detail -- more detail than most readers would want to absorb in one sitting. It presents each principle in its biological context and describes its benefits both for multicellular life and for computing.
If you are impatient, you might want to skip right to the end of the story and read the conclusions. However, as with many a mystery novel, reading the last few pages will tell you who-done-it without telling you the most interesting part...why. The conclusions may well not make much sense without seeing how we get there.
The site map can help navigate to the various pages in an order that helps make sense of the story.
Evolution of Computing -- Last revised 10/21/2019