Comment by skywhopper | Hacker News Reader

Comment by skywhopper | Hacker News Reader

Netflix has been "testing the waters" at GCP for years. And as their infrastructure grows, relying on a single cloud provider would be a silly limitation to put on themselves. Disaster-recovery/failover is a very natural use case for multi-cloud deployments. Not to mention that for certain workloads and patterns, GCP is cheaper and better. Netflix operates at a scale well beyond what most AWS managed services can really address, and broad cloud patterns like object storage, function-as-a-service, VMs, and networks, can be abstracted away given enough resources, which Netflix surely has. Not to mention that at Netflix's scale they have real leverage over cloud providers and what services they offer. They would be stupid _not_ to be playing AWS and GCP against each other for pricing (on which they already get a huge discount) and features.

I'm guessing there's some project getting underway to move/duplicate some subset of resources hosted in AWS on GCP. It's likely a huge project that dwarfs most corporations entire cloud presence, but I'm guessing for Netflix it's a small piece of a giant puzzle, and that the vast majority of their operations are still centered at AWS and will be for years and years to come.

You don't have to be Netflix to get AWS, GCP and MS to play against each on pricing. I've seen a number of my clients build a prototype on each platforms, and then use it to make final decision. Part of that final decision is having the cloud vendors compete on price.

I personally haven't seen any massive price breaks, but there definitely is room the margins.

For years Dell would let it "leak" out to the press that they were looking at supporting AMD chips. It would happen every year like clockwork. Then after a month you'd hear that Dell and Intel had come up with a new deal to remain Intel exclusive.

Hasn't Dell offered AMD chips for a long time? Dell currently sells AMD-based laptops, and was doing so back in 2008 as well.

For example, this Ryzen laptop: http://www.dell.com/en-us/shop/dell-laptops/new-inspiron-15-...

These are probably to prevent action from the FTC.

I think they are talking about when Dell made serious profits in PCs.

Before then – in the 90s both Intel and Microsoft used a lot of backroom deals to keep big vendors exclusive. If memory serves it took the Opteron’s crushing performance advantage before we started seeing that break for Intel servers; Linux users for years were stuck paying for OEM Windows installs which were never used.

Yes, but I've been around the industry a while. I specifically remember this happening during the late 90s and just in to the 2000s.

It was occurring at a time when AMD was fairly firmly beating Intel on a straight $/performance basis (~K6 era). If Dell had started shipping AMD based systems back then it could have been a big game changer.

The general narrative around the tech press at the time was that Intel knew this and were desperate to keep Dell as Intel exclusive.

Really playing these companies against each other would mean dynamically balancing your services across the providers based on pricing

I've love to see some open source tools that manage this automagically for me. :-)

It would also require an architecture that supports multiple interfaces. Totally doable (the facade pattern comes to mind) but you'd want to calculate the TCO to see it works out financially.

Shameless plug: we have an open source tool you can build stuff like this on. It'll run any logic you can implement in JavaScript, we have people doing things like "send traffic to backend with lowest latency" already, it'd be pretty easy to pull in some API data to pick based on cost.

https://github.com/superfly/fly

I work for a large CDN. Lots of big customers do this, they shift their traffic around between different CDNs based on a lot of metrics, including price and performance.

I also work for a CDN.

Some of our stuff dynamically utilizes other CDNs. A customer might go try and test deploy something and we might use (AWS or whatever) to do it.

It's CDNs all the way down.

I work for a company people use to build CDNs. Just another turtle. :D

Adding my turtle as a CDN employee ;-)

We license software[1] to ISPs/MSPs who then use it to "OEM" Kubernetes clusters to CDN providers.

We should rename our product "Infinite Regress" ;-)

[1] https://gravitational.com/telekube/

I wonder how many of the people in this chain are sitting next to me.....

like physical infrastructure or something else? a bit curious (having many neteng friends at many, many CDNs)

I have similar friends!

We the whole stack for most people, but also run our software on "private cdns" for some. We primarily target devs: https://fly.io

Did this a lot with payment processors. Literally had application router settings that could route based on % or switch off once a certain monthly or yearly amount was hit. Configs could be updated in real time.

If it's a final decision, then most of the cost is in the future, far beyond the reach of the initial quotes.

I've seen too many companies turn extractive (including, of course, that same Microsoft) to ever use the words "final" next to "platform decision".

I agree with most of what you said, however, Netflix most definitely does not operate at a scale well beyond what most AWS managed services can address.

AWS managed services literally sit behind billions of devices, a scale that makes Netflix look small in comparison: https://techcrunch.com/2018/02/27/apple-now-relies-on-google...

Seriously? or are you a member of AWFUL - Americans Who Figuratively Use Literally ;)

Literally has been used as an intensifier for a figurative expression since the 1800s and has been defined in the dictionary that way for quite a while.

It was used in this manner by Charles Dickens, James Joyce and American authors as well.

Although, it pisses me off when I hear it used that way.

Unfortunately that usage makes the word completely useless for its actual meaning. It could have been a very useful word.

And I suppose no one will ever be able to describe the act of walking quick in work out clothes again because of all the people “running” to the store.

> (on which they already get a huge discount)

just wondering, how huge (as a percentage) that discount is?

They'll play a few games with contracts to spread the costs around (e.g. giving n% discounts for m years of reserved capacity). You probably won't get much just by saying "I can just go to GCP" and asking for the world. The more you spend, the more distinct services you buy, and the longer you want to commit the steeper the discount. I've seen up to 10% but I'm sure they do more for bigger customers.

I've seen upwards of 50% for some particularly large customers. To the point of "if we lose our discount, our company goes under".

Makes you wonder if the vendor is still making money on the deal.

Reserved instances always make money because you make profit without depreciation kicking in.

Margins are pretty high, in part due to scale. If AWS were to hypothetically sell 80% of capacity at cost, that makes the last 20% more profitable than it would be to just sell the 20% alone.

They get CAF. When CAF runs out, pricing shoots up.

What's CAF? Googling wasn't very helpful :)

Customer Appreciation Funds/Customer Acquisition Funds/etc.

"We have $500M available this year that we can spend to lock in $1B in revenue. If you give us $1B in revenue, we would spend $500M on helping you to achieve your objectives. If you give us $100M in revenue, we give spend $50M in helping you to achieve your objectives"

So say I need to spend $10M on some project and I am looking at $20M of spend on hosting. You are offering me 10M CAF. I spend $20M on hosting, and 0 on some other project - instead you spend your 10M CAF on helping me with my project, which of course looks really good on my books.

Reply on news.ycombinator.com