cloud

Could you have ever imagined an executive order from the president of the United States, saying that in 45 days all firms have to stop working with a Chinese run app?

After speculation that Microsoft would get the deal, it seems like Oracle have found the formula that works for both the US and China.

Oracle was ‘OK’ with not becoming TikTok’s parent company and was content becoming its cloud services provider, keeping US data in the US.

You can’t help but feel if it goes ahead, Oracle has got a great deal.

TikTok spends almost as much with Google Cloud as Netflix do with AWS.

Netflix’s recent earnings report showed they have committed $900m “primarily related to cloud computing costs,” over the next three years.

In 2019, TikTok signed a three-year agreement with Google Cloud (GCP) to buy more than $800 million in cloud services.

A deal anything close to the Google deal could be huge for Oracle, it represents a 5% increase in their cloud services business which had only grown 1% in 2020.

However you cut it that’s a lot of cloud services.

If you asked these companies how much they spend per user on the cloud, they wouldn’t know. A crude model is taking subscribers and dividing by cloud spend.

There are much better ways to do this but even some of the biggest companies don’t even use this basic model.

So, do those numbers sound alright or like a lot of money?

On the surface they both sound like very reasonable numbers.

The reality is they are both excessively high. Nearly all organisations are spending between 50-70% more with their cloud providers than they need to. For Netflix, this could see their cloud spend per user down to 45c and TikTok to 15c.

Why should I care about the cost per user?

You could look at this two ways: 1) Spending more liberally means potentially faster products, more innovation 2) Less efficient code, less understanding of how your tech works, less innovation.

Either way, a cost per user is a good place to start. The more you understand your systems at a deep level, the fewer incidents and outages and the less they cost.

TikTok might be more worried about operating in the US or not.

What’s Netflix’s excuse?

Performance testing on its own isn’t useful, it’s a waste of time and money. Testing can be a great method to understanding at a deep level where you can make radical, innovative and disruptive business change. But few use it this way.

Almost every dev/product team out there can at some level of run automated performance testing, it might be very basic but it’s a start. They’ll typically:

1. Setup a JMeter or python script
2. Run x iterations of a user journey
3. See if it was as quick as the last time
4. Check-in code

If you are lucky you might get a meaningless report with response times and throughput, maybe even on a graph. But what does good actually look like?

Good companies know what response times ‘should’ be and the relationship to the throughput (users/traffic/something) on the system.

Great companies measure of application efficiency. Monitor and fix stability issues in dev. Consistently instrument their design, using data to influence the future. Create small, flexible, mutable services and applications that scale up, down and across. Forecast and provision the right type, level and placement of capacity, moving with the sun. Build scalable resilience into applications before it’s needed. And of course, know how fast their application should be and how many things it should do at any time.

I’ve talked previously about creating logical maps of systems, end to end. Using business>service>infrastructure data and map this to customer behaviours. Then using this to influence the new products that you bring to market.

Test for insight.

Testing for the sake of testing is wasting everyone’s time. Test for insight. The insight you gain will let you create great products, market to the right audience and stay ahead of the future.

Those who really understand their systems and how they work will have a superior understanding of their customer’s needs and wants. They will be able to make whatever changes they want without worrying about what the underlying tech stack is doing. How liberating would that be?

Testing is often used as a tick box exercise and in most cases, the results you get aren’t even good enough for that. Forget about testing, start thinking about how and where can I get the insight that will let me do something nobody has done before.

Thats where the real value is.

Thomas

Sign up if you’d like regular insights with ‘Unconsidered Needs’, a short weekly email about the trials and tribulations of tech value creation.

A common symptom of working with private equity is that the people you start the journey with are not the same as you end with. Most CTO’s will get their key leadership team post-acquisition and tell them:

1. PE partners will have ‘some’ involvement going forward
2. Expect ‘some’ changes in people, strategy and velocity.
3. Don’t worry!

With 10% of key employees leaving during the first 12 months, how do you keep teams motivated and engaged? Great CTOs will give consistently give clear strategic direction to their technology leadership teams.

10 things great CTOs need to make clear to their technology teams

1. What the vision is
2. Why we are doing it
3. Your role in all this
4. How long we have
5. How PE will be involved
6. When PE will be involved
7. How we will measure success
8. The rewards for success
9. The cascading messages
10. How often we will review progress

Great CTO’s won’t treat this as a once and done exercise, they will make a regular commitment to their teams to continually assess and reiterate this vision.

Great Leaders understand the importance of bringing their team on the journey, making sure everyone knows their place in the new world.

Thomas

Thanks to Clare Eagle for the discussion that led to this unconsidered need!

Sign up if you’d like regular insights with ‘Unconsidered Needs’, a short weekly email about the trials and tribulations of tech value creation.

The real value of ‘cost optimising’ in the cloud isn’t really about how much money you save, it’s what you can do next…

Value is made on the journey, understanding at a deep level where you can make radical, innovative and disruptive business change.

Today most companies have internal teams who can do this work with some special tools etc. Low hanging fruit is not technically challenging or particularly exciting. You will see this trend:

1. Move to the cloud
2. Adopt more cloud technologies
3. Decide its costing too much so buy RI’s (or equivalent)
4. Renegotiate
5. Turn off dev/test instances at the weekend
6. Rightsize simple workloads
7. Rest

What good companies do is all the above, earlier and quicker.

What great companies do is create logical maps of their systems, end to end. They use business>service>infrastructure data and map this their customer behaviours. They then use this to influence the new products that they bring to market.

Those who really understand their customer’s needs and understand how their system ‘really’ works can start making radical, innovative, disruptive business changes without worrying about what the underlying tech stack is doing.

I suspect we are 2 years away from cost optimisation being run like a disc cleanup tool. You’d get an optimisation score from 1-100 ‘would you like to optimise your cloud estate now?’

Cost optimisation isn’t sexy, it’s not even that interesting. Imagine how much value you can create by doing something new with your business and not having to worry about the tech, that is sexy!

Thomas

Sign up if you’d like regular insights with ‘Unconsidered Needs’, a short weekly email about the trials and tribulations of tech value creation.