4 ms·

Why can’t it be parallelized?

nivertech 5y ago
Because each computation needs an output of the previous iteration, so the only way to beat other miners/validators/block providers/etc. is to have a better single thread performance, basically you need to build a faster CPU. PoET (Proof of Elapsed Time) works differently, there you basically running a sleep or a timer inside an Intel SGX secure enclave, so you have to trust Intel, and also need to hope that there are no security vulnerabilities there.
- keymone 5y ago
  Ok, I have best single thread performance on my validator machine. What if I have two of those machines?
  - nivertech 5y ago
    One of them will finish calculating VDF earlier, and that result will be verified by other nodes. When used for leader selection, VDF’s offer a substantial improvement over verifiable random functions. Instead of requiring a non-colluding honest majority, VDF-based leader selection only requires the presence of any honest participant. This added robustness is due to the fact that no amount of parallelism will speed up the VDF, and any non-malicious actor can easily verify anyone else’s claimed VDF output is accurate. [1] [1] https://blog.trailofbits.com/2018/10/12/introduction-to-verifiable-delay-functions-vdfs/ https://blog.trailofbits.com/2018/10/12/introduction-to-veri...
    - keymone 5y ago
      > One of them will finish calculating VDF earlier how is that different from one miner finding a valid signature faster than the other? and how is that not parallelization?
      - phoyd 5y ago
        Point is, that you can not finish VDF twice as fast (and probably not even by any other factor >1) with two CPUs, because they can't share work.
        keymone 5y ago
        That is not an answer to my question. So far it’s just PoW. Where’s the difference?
      - lupire 5y ago
        There's nothing to parallelize. There is one computation that takes a long time. If two computers do the same work, only the faster one matters, not the luckier one. Bitcoin miners can parallelize because it's a search problem, so they can probe multiple different candidates simultaneously. That said, VDF isn't a drop in replacement for Bitcoin style POW. VDF just gives you the ability to say (provably, so no one can cheat) "everyone has to do a 10 minute serialized computation before launching a (parallelizable) POW hash search" so we don't use all the world's available electricity just for mining.
        keymone 5y ago
        If there is PoW component then it’ll take just as much energy as Bitcoin does to provide the same security, because energy spent = mining reward. Except it’s going to be worse because VDF is the centralizing factor, it’s like having bitcoin where constructing a block takes really long time. The actor that finishes first has the advantage to start PoW and will win disproportionate amount of blocks by revealing their data at the last possible moment, preferably when or just before another actor publishes their “worse” PoW.

Because each computation needs an output of the previous iteration, so the only way to beat other miners/validators/block providers/etc. is to have a better single thread performance, basically you need to build a faster CPU. PoET (Proof of Elapsed Time) works differently, there you basically running a sleep or a timer inside an Intel SGX secure enclave, so you have to trust Intel, and also need to hope that there are no security vulnerabilities there.

keymone 5y ago
Ok, I have best single thread performance on my validator machine. What if I have two of those machines?
- nivertech 5y ago
  One of them will finish calculating VDF earlier, and that result will be verified by other nodes. When used for leader selection, VDF’s offer a substantial improvement over verifiable random functions. Instead of requiring a non-colluding honest majority, VDF-based leader selection only requires the presence of any honest participant. This added robustness is due to the fact that no amount of parallelism will speed up the VDF, and any non-malicious actor can easily verify anyone else’s claimed VDF output is accurate. [1] [1] https://blog.trailofbits.com/2018/10/12/introduction-to-verifiable-delay-functions-vdfs/ https://blog.trailofbits.com/2018/10/12/introduction-to-veri...
  - keymone 5y ago
    > One of them will finish calculating VDF earlier how is that different from one miner finding a valid signature faster than the other? and how is that not parallelization?
    - phoyd 5y ago
      Point is, that you can not finish VDF twice as fast (and probably not even by any other factor >1) with two CPUs, because they can't share work.
      - keymone 5y ago
        That is not an answer to my question. So far it’s just PoW. Where’s the difference?
    - lupire 5y ago
      There's nothing to parallelize. There is one computation that takes a long time. If two computers do the same work, only the faster one matters, not the luckier one. Bitcoin miners can parallelize because it's a search problem, so they can probe multiple different candidates simultaneously. That said, VDF isn't a drop in replacement for Bitcoin style POW. VDF just gives you the ability to say (provably, so no one can cheat) "everyone has to do a 10 minute serialized computation before launching a (parallelizable) POW hash search" so we don't use all the world's available electricity just for mining.
      - keymone 5y ago
        If there is PoW component then it’ll take just as much energy as Bitcoin does to provide the same security, because energy spent = mining reward. Except it’s going to be worse because VDF is the centralizing factor, it’s like having bitcoin where constructing a block takes really long time. The actor that finishes first has the advantage to start PoW and will win disproportionate amount of blocks by revealing their data at the last possible moment, preferably when or just before another actor publishes their “worse” PoW.