Cover image
Brian T. Liao
Jan 17, 2019 • 7 min read

The Future: The Cloud


I believe that the cloud is going to be a massive part of computing in the future. No longer are applications based locally, it is easier to have it as a web service which can be updated and improved quickly and effectively. There is no need for users to update their software. Companies are now moving to subscription services where users pay and get the newest most updated service.

What is the cloud? The cloud is a collection of services that are run in a datacenter somewhere. They provide services like drive space so your files are connected on all your devices. Your email is run at the datacenter, not your computer but the information is passed to your computer. Netflix is a streaming service that uses cloud infrastructure to run their streaming service. Amazon Web Services provides tons of infrastructure and platforms allowing websites and businesses to be run on the cloud. And they make a ton of money from it.

Why will the cloud be the future? It truly embraces economies of scale, is easily scaled, simple, reliable, and fast, and is agile in being updated and improved. This is where computing can still be improved as Moore’s Law dies for the personal computer. The upcoming of the Internet of Things and Smart Homes and appliances as well as machine and deep learning also require the cloud to run these devices and tasks.

To run your website, you would have to get servers and scale it for your need. When things break, it’s frustrating and you lose money. The cloud takes away this frustration, it adds a layer of abstraction to focus on your business. It takes away your worries with the infrastructure and if your business grows, you won’t have to worry about scaling up your infrastructure.

Cloud services can do this do to the economies of scale. It is cheaper to run a hundred servers than one and a million servers than a hundred. For one, you have figure out how to run it, how to replace parts, how to debug problems. And every business would have to do that on their own. At a hundred, you can bulk discounts on equipment. You can have teams that have solved the problems you may encounter with servers. You can buy redundant parts for fast repair and high uptime. At a million, you can build infrastructure for your datacenter, predict usage, improve efficiency and cooling of datacenters and continue to reduce prices, while making profit from giving a simple service people can use.

Amazon Data Center

The cloud is also reliable and fast. Say a power outage knocked out your servers, then your online business is down and you lose money. Even worse, you lose your data due to accidents, natural disasters, a whole list of possibilities. But on a cloud service, even if one datacenter goes down, there are redundant copies of the information and redundant servers that can be used to keep your business and many others up and running.

Individual consumers also benefit from this reliability. People take hundreds of photos they don’t want to be loss. Saving it a backup drive, it can still be lost or corrupted. Cloud storage gives an extra sense of redundancy again. That important document you were working on and got corrupted? Luckily it was backed up on a cloud storage.

This brings us to speed. Processing data on your computer is faster in that there is less latency to go from processing data to showing up on your screen. In the cloud, you have to send requests to the datacenter and the datacenter has to send it’s processed info back. Their distributed computers might be faster, but going through the internet networks still can be slow. This isn’t a problem for something like an image or file store, where you are not constantly sending information back and forth or an expensive computing job like a computer animation rendering or a machine learning training job.

It will be important to update and improve our networking infrastructure, mobile, telecom, and internet as the cloud-centric world will involve sending much data around. Doing so will allow us to reap the benefits of the cloud while reducing latency making work faster and more fluid.

As for the role of the personal computer and the cloud datacenters, both will still exist. Some tasks are better suited for personal computers. You might not always have internet, but it’s possible in the future, you will be connected everywhere. If so, cloud computers will probably be faster and most workloads will be done on the cloud. Personal computers have come to a point with Moore’s Law that they aren’t getting much faster. On the other hand, cloud computers can be replaced and improved quickly.

Matrix Multiplication Optimization. It is hard to make massive optimization on general person computers but can be done on specialized domain-specific cloud computers.

We see with GPUs, the newest ones are being put in but for an average consumer to do so, he would be paying thousands every year. Instead he can pay hundreds in service fees to getting the cutting edge hardware. Google also has stated that domain specific hardware is the future, thus creating the TPU tensor processing unit for their AI and machine learning operations. Personal computers can’t run these tasks, it would take a week to do what one of these can do in 5 minutes. For Google to train it’s most complicated models, it can take multiple weeks on these, equivalent to centuries on a personal computer.

Google’s Tensor Processing Unit, specialized hardware for fast and efficient machine learning.

As with AI and machine learning being powered with the cloud, the emergence of smart homes will too. The idea is having all your appliances and components of your house connected to a master entity, in the cloud, which will control them for your conveniences and pleasure.

A smartwatch can track your temperature and reduce the thermostat if you are too warm. The refrigerator will know when you are low on food and automatically order for some to be delivered. You will be transported in your self-driving car while a digital personal assistant handles all of your tasks. Your personal assistant will remind you of bed and dim the lights, while placing you in your automatically adjusted temperature and firmness bed. All data about you is collected to make your life easier and more pleasant. And all of this will be processed by the cloud, tons of data processed and organized by cloud computers.

Personal assistant today already do that, when asking Alexa for something, it will be sent to the cloud, where it will be parsed by a voice recognition system, processed by a natural language system, and have the action processed and returned to you in seconds. Voice recognition and natural language processing are neural network processed tasks which require GPUs on the cloud to process quickly. This isn’t possible for the Echo to do on it’s own but can be done in seconds querying to the cloud.

Conclusion

The cloud is the underlining of future computing. Nowhere else has the scale or ability to process the computing required in the future, while it holds buisness and personal saving applications today, it’s scale is what will enable it in the future. Finally, since the cloud is centralized, it is easy to improve and adapt. Software patches can be applied quickly. Hardware can be replaced quickly. Development will run at high rates while improving to the future powered by the cloud.

Post by: Brian T. Liao