On cloud technologies I

Cloud technologies are now used everywhere. They are used to connect all sorts of devices, distribute and store content, execute remote programs or databases, and organize and control access rights to scores of digitalized data. Cloud technologies enable use cases such as video streaming (e.g., Netflix, Amazon, Apple TV), which is said to consume 2/3 of the total worldwide internet bandwidth (~1.2Tbps or 1200 billion of 0 or 1 per second).

Initially, the "cloud" was focused on distributing decentralised content through what was called "CDN" for Content Distribution Networks. The technology has evolved to be much more than that with players like Amazon and their AWS division, which has democratized the technology to the point of making accessible globally distributed high-performance, safe and redundant computing, which only a few companies could afford no longer than ten years ago.

Cloud technologies accelerated thanks to the pervasiveness of internet connections. Today, Wi-Fi, cellular, and low-cost wired connections are available everywhere. This means that you can run many applications directly into the cloud, like productivity applications (Gmail, PowerPoint, etc.) and even games now. If the internet isn't suddenly available, your work isn't lost and automatically saved and organized in your personalized cloud.

It took a considerable amount of work for Microsoft to morph its Office Suite to be "cloud-friendly", and you have to picture that when you write an email from the Google Chrome browser, every character you write is instantaneously replicated in multiple Google databases (in the form of a "draft") in case your local machine would crash or one of the data centre subject to a temporary failure. Such resilience and burden put on the cloud to be a "safe" remote storage and execution engine took time to build and scale but is now vastly out there. It took time to modify how the software works on a laptop or cell phone and build enough data centres to sustain the volume of 2-way information (upload/download).

But in a world where data is king and AI a prime consumer of knowledge, all key GAFA players (Google, Apple, Facebook/Meta, Amazon) have invested billions in data centres in the US and before them, various lesser-known companies like Akamai, DigitalOcean or some historical players like Oracle, IBM, Dell, HP). China isn't at rest with companies like Alibaba, Tencent.

Datacenter = energy hog, yet...

In a world with soaring energy costs, it's important to understand that data centres consume significant energy. A study suggests that mid-size data centres represent the annual consumption of an 80,000-person town. This comes from the power consumption of processors and electronic components and the energy needed in air conditioning to cool down the dissipated heat from computer racks.

The training of the third version of the AI model used in ChatGPT is estimated to have consumed 1,287 MWh alone, corresponding to the annual consumption of 100+ US families. You may argue that for the service it brings, it's nothing compared to the thousands of paying subscribers...

Progress is constantly being made in chipset power consumption and heat dissipation, and cloud supporters are making the point that computing and storage are centralized, enabling energy saving instead of redundant, decentralised computing.

Main Players (AWS, Azure, GCP)

There's stiff competition for cloud offerings between 3 major players today: Amazon AWS, Microsoft and Google Cloud Computer (GCP). That competition is across multiple features:

• Number/capabilities of data centres
• API offerings, integration with existing software (Office, etc...)
• Cost (storage/compute//type of hardware used)
• Regional and regulatory aspects

You might ask: Why is Amazon even in the race? Well, this smart company understood that its IT infrastructure needed to accommodate peak order demand (like Christmas), and it could monetize its oversized data centres for most of the rest of the year. Smart (and the new CEO of Amazon is the gentleman who led this).

It's hard to know exactly how many computers/data centres each of these players has, but the number is certainly accelerating. In the IT world, people prefer to talk about guaranteed availability in service contracts (like response time for database queries or time to switch data centres if one fails...).

Cutting through the jungle of acronyms

As with any tech, you have to cope with many acronyms. In this article, I will not review the various APIs or software frameworks available (AWS has more than 200+) but rather focus on the leading 3 service models cloud providers offer: SaaS, PaaS, and IaaS.

• SaaS: "Software as a service" is meant for companies that don't need much control over what happens in the data centre. For example, the Google Office suite, which is available to end-users or companies, works like this.
• PaaS: "Platform as a Service" allows customers to develop, run, and manage applications without dealing with the underlying infrastructure which they are "leasing" through a subscription model.
• IaaS: "Infrastructure as a Service" is for IT departments that want to have full control over the software running on a data centre.

Each service option has an associated cost model for hardware/compute storage and software frameworks.

You might have heard the term "on-premise" cloud computing, which essentially means that a data centre is designed to be dedicated and colocated to a given company. This stems from the fact that many companies were originally concerned about outsourcing their IT needs and wanted to have tighter control of the hardware and software infrastructure.

Cloud techs: a matter of national security?

So much data from corporations or consumers is currently stored in data centres, which is causing great concern for governments. Cloud providers are becoming like banks, whose sustainability and ability to protect customer privacy are synonymous with sustainability. I often joke about Apple products because they are too expensive compared to competitors, but it's a rare company that takes privacy very seriously.

We've all heard about the TikTok issue: a Chinese company with a very large user base, notably in the US, is making the US government nervous about how US citizens' data might be used. This concern follows the Facebook crisis, which demonstrated that foreign powers could use social networks to alter the course of domestic elections.

So yes, data centres, like telecom providers, are a matter of national security.

Conclusion

Cloud computing has transformed the way software and data are produced and consumed. It is also a key element fueling the AI revolution and a matter of national interest for leading countries.

#cloud #aws #azure #googlecloud

As always, feel free to contact me @ [email protected] if you have comments or questions about this article (I am open to providing consulting services).

More at www.lohier.com and also my book . You can subscribe to this free bi-weekly newsletter here and access former editions here .