Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

This website has 81% battery power remaining

solar.lowtechmagazine.com

Post reply on HN
Print view View original

31–40 of 365 posts

Re: This website has 81% battery power remaining

#31

Post by jolux »

ralph84 wrote: post #25
malwarebytess wrote: post #10

> This was caused by a software upgrade of the Linux kernel, which increased the average power use of the server from 1.19 to 1.49 watts I wonder what change in the linux kernel caused the increased load. Someone out there is responsible for this crime!

This is an area where Apple is way ahead of everyone else. A code change that increases power usage 25% wouldn’t make it past CI at Apple.

That's incredible. How do they measure that? Does their CI somehow measure power impact on a set of real devices?

Re: This website has 81% battery power remaining

#32

Post by marricks »

widdakay wrote: post #11

I think this also shows how inefficient modern website hosting is. The fact that this person was able to get a raspberry pi to host the #1 website on HN powered by a small 50 watt solar panel is very cool (meaning maybe 10w average power budget), but also shouldn't be as uncommon as it is today. To put this in perspective, a modern server uses 50-100 watts idle doing nothing, and many more under load. To handle the t…

I hate to be that old foggy but, aren’t websites just getting worse and bloated with JS crap?

I’ve had a couple websites I use daily for work just get flashy new interfaces which causes 1/3 to 1/2 second delays in the interface which used to not exist, previously they just had normal page load delays.

For example, SalesForce Lightning, their UI overhaul. Old UI is mainly just flat HTML with some loading on fields. New UI doesn’t have as many page loads it seems but wherever you navigate to takes far longer to load because of api calls or just baaad JS.

Slow for the user, slow for the server. It's almost like the people who push website technology are the same one selling you servers. Hate it and want to go back.

Re: This website has 81% battery power remaining

#33

Post by sz4kerto »

Waterluvian wrote: post #28

I love this experiment. The one comment I have is that when you provide “CPU%” but you’re using the special ”per core” unit that can go above 100%, you must list the number of cores. 128% of what?

CPU load is not CPU usage. Load is basically “number of processes waiting to be scheduled to run”. If load=cpu count then you can’t schedule more processes, your system runs at maximum capacity (although it might mean that the cores wait on some IO).

Eg.: https://estl.tech/cpu-usage-vs-load-ecca22287b21

Re: This website has 81% battery power remaining

#34

Post by detaro »

tentacleuno wrote: post #12
widdakay wrote: post #11

I think this also shows how inefficient modern website hosting is. The fact that this person was able to get a raspberry pi to host the #1 website on HN powered by a small 50 watt solar panel is very cool (meaning maybe 10w average power budget), but also shouldn't be as uncommon as it is today. To put this in perspective, a modern server uses 50-100 watts idle doing nothing, and many more under load. To handle the t…

Hey @dang, you wouldn't happen to know anything about how Hacker News is hosted, would you? Reading this has piqued my curiosity. If time / your position permits, of course.

From what I remember from previous posts by dang: It's living on a single dedicated server with some hosting company (I think you can look up the IP to figure out which) - at least a while ago the code also was single-core, not sure if that is still the case. (in the past it used cloudflare for caching, but hasn't in a few years)

Re: This website has 81% battery power remaining

#35

Post by amelius »

widdakay wrote: post #11

I think this also shows how inefficient modern website hosting is. The fact that this person was able to get a raspberry pi to host the #1 website on HN powered by a small 50 watt solar panel is very cool (meaning maybe 10w average power budget), but also shouldn't be as uncommon as it is today. To put this in perspective, a modern server uses 50-100 watts idle doing nothing, and many more under load. To handle the t…

Wouldn't it be way more efficient to run it on some (virtualized) node in a datacenter that is optimized for it?

Re: This website has 81% battery power remaining

#36

Post by unclebucknasty »

brightball wrote: post #14
widdakay wrote: post #11

I think this also shows how inefficient modern website hosting is. The fact that this person was able to get a raspberry pi to host the #1 website on HN powered by a small 50 watt solar panel is very cool (meaning maybe 10w average power budget), but also shouldn't be as uncommon as it is today. To put this in perspective, a modern server uses 50-100 watts idle doing nothing, and many more under load. To handle the t…

Fwiw, the top of HN isn’t all that stress. It mostly comes down to disk I/O and efficiency of the language. I’ve been at the top of HN for extended hours a couple of times on just a Heroku hobby dyno with no caching at all, but I had Cloudflare out front absorbing all the traffic that would have come from serving static assets.

Not to be contrary, but if your site is largely static and you're fronting it with Cloudfare, then you're essentially saying Cloudfare can handle load.

Not much revelation there, right?

Re: This website has 81% battery power remaining

#37

Post by DietPi »

widdakay wrote: post #11

I think this also shows how inefficient modern website hosting is. The fact that this person was able to get a raspberry pi to host the #1 website on HN powered by a small 50 watt solar panel is very cool (meaning maybe 10w average power budget), but also shouldn't be as uncommon as it is today. To put this in perspective, a modern server uses 50-100 watts idle doing nothing, and many more under load. To handle the t…

It is not a Raspberry Pi, but an Olimex Olinuxino A20 Lime 2 and a 30W solar panel: https://solar.lowtechmagazine.com/about.html#hardware

But yes, web hosting, especially for small/mid traffic websites has become very cheap (in power consumption and in money), especially for static websites where CDNs can be used to serve assets and static content from edge caches. A full x86 server or PC is often total overkill and a little SBC sufficient instead.

It is a dual core CPU btw, to bring the average CPU load into perspective. A very interesting project as a prove of concept, also for others to adopt in countries with unstable electricity supply and/or in relation high electricity costs :).

Re: This website has 81% battery power remaining

#38

Post by ruffrey »

jolux wrote: post #31
ralph84 wrote: post #25

Earlier quoted context omitted.

This is an area where Apple is way ahead of everyone else. A code change that increases power usage 25% wouldn’t make it past CI at Apple.

That's incredible. How do they measure that? Does their CI somehow measure power impact on a set of real devices?

Apple is designing much of the silicon. Measuring power usage is a core competency.

Re: This website has 81% battery power remaining

#40

Post by CraigJPerry »

ralph84 wrote: post #25
malwarebytess wrote: post #10

> This was caused by a software upgrade of the Linux kernel, which increased the average power use of the server from 1.19 to 1.49 watts I wonder what change in the linux kernel caused the increased load. Someone out there is responsible for this crime!

This is an area where Apple is way ahead of everyone else. A code change that increases power usage 25% wouldn’t make it past CI at Apple.

How many more hardware configurations is Linux deployed on than MacOS? Would it be in the region of 3-4 orders of magnitude more hardware configurations?

How long has there been a power consumption focus in the linux kernel? The efforts to reduce power usage by the linux kernel pre-date PowerTOP’s first release and wiki tells me that was 2006.

Is Apple way ahead or is it just less popular?

Post reply on HN