Wednesday, August 24, 2016

Predicting Size of Computing Cloud Providers

Recently Microsoft acquired an AI startup (making virtual scheduling assistant)
Here is a new acronym for that: AIaaS "AI as a Service"
Microsoft to Acquire Genee, Adding AI Magic to Azure

Analysts claim that could help growth of Azure cloud, supported by
info about size and growth of computing cloud providers.
Numbers like this should be taken with a grain of salt

Amazon Leads; Microsoft, IBM & Google Chase; Others Trail | Synergy Research Group
CIS Q216

“In a variety of ways Amazon and the other big three players have distanced themselves from the competition in this market and continue to widen the gap,”


So Amazon is now on 32% of IaaS market growing 53% a year
and Microsoft is on 12% growing 100% a year,
and Google is on 5% now growing 162% a year.

Since Google and Microsoft are growing faster, at some point they may catch up with Amazon.
We could use regression to predict when that could be.
Question is, is the rate of growth linear, polynomial, exponential? Need more data.

If we assume it is a linear growth, formulas are:
Amazon: A(t) = 0.32 + 0.32 *  0.53 * t = 0.32 + 0.17 * t
Microsoft: M(t) = 0.12 + 0.12 * 1.00 * t = 0.12 + 0.12 * t; would never catch up.
Google: G(t) = 0.05 + 0.05 * 1.62 * t = 0.05 +  0.08 * t; would never catch up

If it is exponential growth, formulas would be:
Amazon: A(t) = 0.32 * ((1 + 0.53) ^ t)
Microsoft: M(t) = 0.12 * ((1 + 1.00) ^ t)
Google: G(t) = 0.05 * ((1 + 1.62) ^ t)

Solve on t for A(t) = M(t)
0.32 * ((1 + 0.53) ^ t) = 0.12 * ((1 + 1.00) ^ t) - Wolfram|Alpha
Microsoft would catch up in 3.7 years.
Observe nice charts and calculation on wolframalpha. "Math as as Service"

0.32 * ((1 + 0.53) ^ t) = 0.05 * ((1 + 1.62) ^ t) - Wolfram|Alpha
Google would catch up in 3.4 years.

Not realistic, but makes an interesting experiment with data.

This could be solved by using Logarithmic Differentiation

Would be a good case for The Jupyter Notebook — IPython (Project Jupyter | Home)


It is worth mentioning that better option for many users is PaaS cloud,
where Azure & Google are leaders and AWS is much less.
(Docker) Containers will likely end up being optimal way of deploying apps.

No comments: