Hacker News Reader

Comment by dspillett | Hacker News Reader

With energy prices shooting up, that scanned server cost is more significant now than it used to be, and per-user bandwidth costs don't drop forever due to economies of scale, so there may be more cost efficient promotion methods than 700,000 non-paying users.

There was a study a while back (can't find it ATM, I'll have another look later) that suggested those using other people's accounts were more likely to be heavy users in terms of bandwidth than average, so perhaps this sort of thing has been factored in to the plan.

Having said that, I can't imagine they expected 300k dropping their account now they can no longer share it. Perhaps there are a lot of friend groups clubbing together to share the cost of subscription services like Netflix.

View on HN

Energy prices in Europe are dropping significantly. Spain's pricing is down to 2021 levels again last month. No idea where they host their data in Europe but it's much the same in other European countries. https://www.statista.com/statistics/1267552/spain-monthly-wh...

Blimey. $39/mo to access that. Statista gives me a lot less value than a $20/mo Netflix account!

Arguably that $39 offers a lot more value by giving you access to valuable statistics than the $20 that gives you the opportunity to waste some time staring at the tele. And I'm not saying there's no value to entertainment, but there's plenty of more fulfilling things to do when it comes to entertainment.

> per-user bandwidth costs don't drop forever due to economies of scale,

Can you explain why? I'd have thought they absolutely do and on top of this their caching strategy becomes even more effective.

> Can you explain why?

Bandwidth isn't an infinite resource. It effectively is up to a certain scale and with a few caveats (the costs are not consistent globally) but I'm sure Netflix is beyond the scale where it can be assumed to be and at that point they will be bidding against other big users for peering priority.

With a dedicate server I have some stuff on I can upgrade to 10gbit unmetered for a lot less than 10x the cost the current 1gbit unmetered link (in my case neither dedicated, but 1-shared-to-10-shared will likely scale in cost similarly to 1-dedicated-to-10-dedicated). With another where I have a total bandwidth cap not a speed cap I could likely extend that cap relatively cheaply. At the scales required by a small company, even one running SaaS services, the same is usually true. But at Netflix's scale a small percent increase in any particular locality is still a huge amount of extra to ask for, and it might not really be available (it could be allowed to eat into slack for a short time, this is one of the reasons why networks are provisioned for more than 100% of expected use, but if the extra is needed permanently then infrastructure upgrades may be needed). They can get around lack of cheap bandwidth by instead temporarily increasing compression rates (this will in fact happen automatically, as dropping overall bit-rate is supported to deal with congestion at the level of a user's connection and that will kick in due to congestion elsewhere too) but this will be easily detectable by some users (particularly those with huge screens and great eyesight) who will compare the quality with other streaming content providers and talk about the difference which may affect future sales.

> on top of this their caching strategy becomes even more effective

Caching in terms of keeping data geographically spread will reduce international bandwidth needs, and ISP-level caching through their local boxes will help considerably again (and get more effective as more users join from the same ISP), but there will always be an amount of traffic that needs to travel costly routes (the ISP level caches will only be large enough to support the best caching candidates, like new high-profile shows, a small proportion of the whole catalogue, and many ISPs do not have that facility at all).

From https://www.theverge.com/22787426/netflix-cdn-open-connect

> ““Not only are we placing content on all of these servers around the world, but we are pre-placing them based on what is popular. Because we predict what is popular, we’re able to put it as close as possible under the correct server,” Haspilaire says. “This pre-placement of our films and shows allows us, based on prime time viewing hours, to store 100 percent of our catalog locally."

And they have 17,000 of these boxes. Extra users in these areas will have essentially zero marginal cost. And outside of that, I don't believe the bandwidth market we see - in terms of upgrading our small servers - has any resemblance to the deals they can do.

Reply on news.ycombinator.com