4 ms·

Citation needed. According to https://www.nature.com/articles/d41586-018-06610-y https://www.nature.com/articles/d41586-018-06610-y datacenter uses 200TWh/yr w

Citation needed. According to https://www.nature.com/articles/d41586-018-06610-y https://www.nature.com/articles/d41586-018-06610-y datacenter uses 200TWh/yr which ~= 22GW. Nuclear plants are ~1GW, so you're saying TLS is responsible for ~50% of all IT power usage?

bullen 5y ago
Internet uses 84 to 143 gigawatts of electricity every year, HTTPS uses the majority of that.
- francislavoie 5y ago
  You really think that the simple act of encrypting and decrypting the traffic of web requests is the heavy part?
  - bullen 5y ago
    I know it is, because I made my own HTTP server from scratch.
    - francislavoie 5y ago
      If your site is pure static files, sure, it's added overhead. But if you're doing anything dynamic on the server, connecting to a database etc, then HTTPS is a tiny, tiny percentage of the work being done by the hardware. CPUs are very well optimized for these tasks these days https://en.wikipedia.org/wiki/AES_instruction_set https://en.wikipedia.org/wiki/AES_instruction_set.
      - bullen 5y ago
        Imagine all the energy wasted making those instruction sets, hardcoded crypto silicon and that part of the CPU... the power that part of the CPU wastes while not in use. For eternity, or atleast until the CPU breaks. You just made the situation worse for your own arguments it seems!
        tedunangst 5y ago
        You can actually pinpoint exactly where the argument goes from not great to super dumb.
- adgjlsfhk1 5y ago
  That's just because the majority of content is served over HTTPS.
- giantrobot 5y ago
  Claims without evidence. Most processors released in the past decade have some sort of cryptographic acceleration. Even before the likes of AES-NI there's been two decades of vectorized optimizations for encryption and hash algorithms. Sun among others released SSL accelerator expansion cards 20 years ago. The processing overhead for TLS is trivial anymore. You'd be hard pressed to find devices (in any market segment) sold in the past decade without some sort of acceleration or access to vector optimized code paths. The idea that HTTPS is using tons of extra power over exabytes of HDDs spinning up, inefficient microservice architectures, or poorly optimized code interpreted code is pretty laughable. But you've made bold claims. I'm sure you've got compelling evidence backed by extensive and reproducible testing.

bullen 5y ago

Internet uses 84 to 143 gigawatts of electricity every year, HTTPS uses the majority of that.

francislavoie 5y ago
You really think that the simple act of encrypting and decrypting the traffic of web requests is the heavy part?
- bullen 5y ago
  I know it is, because I made my own HTTP server from scratch.
  - francislavoie 5y ago
    If your site is pure static files, sure, it's added overhead. But if you're doing anything dynamic on the server, connecting to a database etc, then HTTPS is a tiny, tiny percentage of the work being done by the hardware. CPUs are very well optimized for these tasks these days https://en.wikipedia.org/wiki/AES_instruction_set https://en.wikipedia.org/wiki/AES_instruction_set.
    - bullen 5y ago
      Imagine all the energy wasted making those instruction sets, hardcoded crypto silicon and that part of the CPU... the power that part of the CPU wastes while not in use. For eternity, or atleast until the CPU breaks. You just made the situation worse for your own arguments it seems!
      - tedunangst 5y ago
        You can actually pinpoint exactly where the argument goes from not great to super dumb.
adgjlsfhk1 5y ago
That's just because the majority of content is served over HTTPS.
giantrobot 5y ago
Claims without evidence. Most processors released in the past decade have some sort of cryptographic acceleration. Even before the likes of AES-NI there's been two decades of vectorized optimizations for encryption and hash algorithms. Sun among others released SSL accelerator expansion cards 20 years ago. The processing overhead for TLS is trivial anymore. You'd be hard pressed to find devices (in any market segment) sold in the past decade without some sort of acceleration or access to vector optimized code paths. The idea that HTTPS is using tons of extra power over exabytes of HDDs spinning up, inefficient microservice architectures, or poorly optimized code interpreted code is pretty laughable. But you've made bold claims. I'm sure you've got compelling evidence backed by extensive and reproducible testing.