5 ms·

> The electricity gains to be had are likely very small. Minuscule, even. That's just wrong. Electricity used by data centers today add up to the whole consump

by brabel 3y ago

> The electricity gains to be had are likely very small. Minuscule, even. That's just wrong. Electricity used by data centers today add up to the whole consumption of a largish country. If you can save even just 5% of that, it's an enormous difference. Companies like Google and Netflix spend millions to save much less than that in computing.

packetlost 3y ago
If only it were that simple! If such a gain could be made by simply switching your programming language to one that is "more efficient" (in power consumption terms) we would see Google writing everything in C++, instead of the mix of Go, Python, Java, PHP, etc. that comprises most of their software stacks! But no, that's not how it works at all. Nearly half (about 40%)[^1] of datacenter's total power draw on average is from cooling, which is usually shared across every server in the datacenter. Next, about 30%[^2] of the average server's power draw comes from its CPU (this obviously varies, even at different times on the same server), most servers sit at about 50% utilization on average. So already we're looking at 50% of 30% of 60% = 9% total of energy consumption going to the CPU with very rough, over-estimations. Now Clojure makes up about 1.25% of code, ish[^3]. Let's round that up to 1.5% to be generous. So a reeeeally rough estimate has Clojure code taking up 0.135% of total datacenter power usage globally assuming averages for pretty much everything. I had trouble finding benchmarks for Clojure vs. Common Lisp, but Clojure is basically just Java at runtime, so this[^3] will do. Benchmarks are kinda silly, but we're talking about really average performance. The benchmarks I could find had Java beating SBCL on most benchmarks, but let's say you stand to gain a generous 20% efficiency by using SBCL over Java. That leaves you at a 0.027% gain in efficiency with a lot of generous rounding and assumptions. Now, of course most code is not CPU constrained... you get the idea. The efficiency gains are not much at all. [^1]: https://davidmytton.blog/how-much-energy-do-data-centers-use/ https://davidmytton.blog/how-much-energy-do-data-centers-use... [^2]: https://www.researchgate.net/publication/355862079_A_Review_of_Data_Centers_Energy_Consumption_And_Reliability_Modeling https://www.researchgate.net/publication/355862079_A_Review_... [^3]: https://benchmarksgame-team.pages.debian.net/benchmarksgame/fastest/lisp.html https://benchmarksgame-team.pages.debian.net/benchmarksgame/...
- refset 3y ago
  Assuming current multi-tenant infrastructure trends continue, over-provisioning of RAM and slow start up times are likely the much bigger efficiency issues.
- brabel 3y ago
  Everything you said is a waste of your time. I am not talking about replacing Clojure with Common Lisp. Why did you even come up with that strawman?? The point is that if everyone used more efficient languages, a very large reduction in energy costs could be expected (not just in data centers by the way, every device takes power, and the ridiculously inefficient apps people are running probably cost a lot of energy). It's the same argument as to why you should have a ecofriendly house or whatever: you try to do your part and hope the others will do too. You may be cynic enough to think that makes no difference, but you're just blind to your own unjustified prejudices.
  - medo-bear 3y ago
    Precisely. If you were to divide languages as high level vs low level, or garabge collected vs manual mem management, I feel that SBCL would be on top in terms of energy efficiency in both the formers.
    - packetlost 3y ago
      I think there's some Schemes that could give SBCL a run for it's money, but I have no doubt it would be a very strong contender. I hope this doesn't come across as bashing SBCL, that is not my intent at all.
  - packetlost 3y ago
    > I am not talking about replacing Clojure with Common Lisp. Why did you even come up with that strawman?? Because that's what I was talking about in the first place? I was responding directly to the text under the linked video making an ecological appeal to using SBCL over Clojure. Both are very efficient languages, so that particular appeal seems silly. > The point is that if everyone used more efficient languages, a very large reduction in energy costs could be expected (not just in data centers by the way, every device takes power, and the ridiculously inefficient apps people are running probably cost a lot of energy). Sure, I agree with that. But that's not what was being talked about. If you have data or sources you can share that show studies that model how much energy could be saved, I'd love to see them. > It's the same argument as to why you should have a ecofriendly house or whatever: you try to do your part and hope the others will do too. So don't write load bearing applications in Python. But most code is not load bearing or even executed that often. The energy usage of SBCL vs Clojure is like making the decision to unplug your toaster when you're not using it, and it's not even clear which is more efficient. > You may be cynic enough to think that makes no difference, but you're just blind to your own unjustified prejudices. I looked at the data and came to a conclusion. I was talking about choosing one pretty efficient language over another, you were talking about something else. > Everything you said is a waste of your time. You have no right to tell me what is a waste of my time. I found it quite interesting, I just wish we could've had a conversation made in good faith instead of whatever this was.

by brabel 3y ago

packetlost 3y ago

If only it were that simple! If such a gain could be made by simply switching your programming language to one that is "more efficient" (in power consumption terms) we would see Google writing everything in C++, instead of the mix of Go, Python, Java, PHP, etc. that comprises most of their software stacks! But no, that's not how it works at all. Nearly half (about 40%)[^1] of datacenter's total power draw on average is from cooling, which is usually shared across every server in the datacenter. Next, about 30%[^2] of the average server's power draw comes from its CPU (this obviously varies, even at different times on the same server), most servers sit at about 50% utilization on average. So already we're looking at 50% of 30% of 60% = 9% total of energy consumption going to the CPU with very rough, over-estimations. Now Clojure makes up about 1.25% of code, ish[^3]. Let's round that up to 1.5% to be generous. So a reeeeally rough estimate has Clojure code taking up 0.135% of total datacenter power usage globally assuming averages for pretty much everything. I had trouble finding benchmarks for Clojure vs. Common Lisp, but Clojure is basically just Java at runtime, so this[^3] will do. Benchmarks are kinda silly, but we're talking about really average performance. The benchmarks I could find had Java beating SBCL on most benchmarks, but let's say you stand to gain a generous 20% efficiency by using SBCL over Java. That leaves you at a 0.027% gain in efficiency with a lot of generous rounding and assumptions. Now, of course most code is not CPU constrained... you get the idea. The efficiency gains are not much at all. [^1]: https://davidmytton.blog/how-much-energy-do-data-centers-use/ https://davidmytton.blog/how-much-energy-do-data-centers-use... [^2]: https://www.researchgate.net/publication/355862079_A_Review_of_Data_Centers_Energy_Consumption_And_Reliability_Modeling https://www.researchgate.net/publication/355862079_A_Review_... [^3]: https://benchmarksgame-team.pages.debian.net/benchmarksgame/fastest/lisp.html https://benchmarksgame-team.pages.debian.net/benchmarksgame/...

refset 3y ago
Assuming current multi-tenant infrastructure trends continue, over-provisioning of RAM and slow start up times are likely the much bigger efficiency issues.
brabel 3y ago
Everything you said is a waste of your time. I am not talking about replacing Clojure with Common Lisp. Why did you even come up with that strawman?? The point is that if everyone used more efficient languages, a very large reduction in energy costs could be expected (not just in data centers by the way, every device takes power, and the ridiculously inefficient apps people are running probably cost a lot of energy). It's the same argument as to why you should have a ecofriendly house or whatever: you try to do your part and hope the others will do too. You may be cynic enough to think that makes no difference, but you're just blind to your own unjustified prejudices.
- medo-bear 3y ago
  Precisely. If you were to divide languages as high level vs low level, or garabge collected vs manual mem management, I feel that SBCL would be on top in terms of energy efficiency in both the formers.
  - packetlost 3y ago
    I think there's some Schemes that could give SBCL a run for it's money, but I have no doubt it would be a very strong contender. I hope this doesn't come across as bashing SBCL, that is not my intent at all.
- packetlost 3y ago
  > I am not talking about replacing Clojure with Common Lisp. Why did you even come up with that strawman?? Because that's what I was talking about in the first place? I was responding directly to the text under the linked video making an ecological appeal to using SBCL over Clojure. Both are very efficient languages, so that particular appeal seems silly. > The point is that if everyone used more efficient languages, a very large reduction in energy costs could be expected (not just in data centers by the way, every device takes power, and the ridiculously inefficient apps people are running probably cost a lot of energy). Sure, I agree with that. But that's not what was being talked about. If you have data or sources you can share that show studies that model how much energy could be saved, I'd love to see them. > It's the same argument as to why you should have a ecofriendly house or whatever: you try to do your part and hope the others will do too. So don't write load bearing applications in Python. But most code is not load bearing or even executed that often. The energy usage of SBCL vs Clojure is like making the decision to unplug your toaster when you're not using it, and it's not even clear which is more efficient. > You may be cynic enough to think that makes no difference, but you're just blind to your own unjustified prejudices. I looked at the data and came to a conclusion. I was talking about choosing one pretty efficient language over another, you were talking about something else. > Everything you said is a waste of your time. You have no right to tell me what is a waste of my time. I found it quite interesting, I just wish we could've had a conversation made in good faith instead of whatever this was.