HNHacker News
TopNewBestAskShowJobs

orm

135 karma · joined January 17, 2014

submissionscomments
orm··on Deep Learning with Spark and TensorFlow
I think one reason would fault tolerance. Is there a fault tolerance layer in GNU parallel? last time I checked their homepage ( a few minutes ago), there was no reference to fault tolerance.

Another reason is, perhaps, scheduling.

orm··on Packet copies: Expensive or cheap?
Thinking about it again, from my calculation below, if those packets are coming from RAM then you are bound to, say, 25GB/s incoming, 25GB/s outgoing.

Since Haswell L1 cache seems to allow a peak bandwidth of 64 byte loads + 32 byte store per cycle, so this would match the CPU throughput and then this suggests, if the lifecycle is "bring from ram" "copy in and out of L1 cache", "put back in RAM", then those copies may well be free given you are choking on RAM to begin with.

orm··on Packet copies: Expensive or cheap?
New to this, but basing such calculation on purely on CPU capacity would imply a bandwidth to memory (or some cache)

64 B / inst * 2 (read and writeback) * 3 inst / cycle * 3 Gcycle/sec > 1000 gigabytes/sec, which is insane considering 50GB/s is the norm to RAM right now.

*edit: Haswell L1 cache seems to allow a peak bandwidth of 64 byte loads + 32 byte store per cycle, so like other comments point out, this would match the CPU capacity. On the other hand, there is also latency. But it is hard for me to tell how latency factors into this.

Could someone explain how they would decide whether copying packets is cheap or not?

orm··on My husband needed therapy after working for Amazon
Some things about this article bother me. It starts by describing what is objectively a problem ascribable to Amazon itself ( letting on-call rotations reach the 3 week mark is terrible ), and then descends to pure appeal to emotion.

Here's what I mean: the only thing that seemed an actual problem, to judge from the article, is that the on-call rotations became once every three weeks. Perhaps the length of the on-call, or the number of on-call people at the same time is another problem Amazon could address in this case.

On-call in general (24/7 availability for some period of time), is not exclusively an Amazon thing. I'm surprised the writer tried to portray it so.

After that, the author stops giving reasons. For example, the decision to go on a business trip after 4 weeks of their daughter being born seems to have been their own, not Amazon's. " I imagine he could have turned that down, but I was insistent that he go so". She did not explain if they felt coerced by Amazon and how.

Same thing goes for the hotel they cancelled a day early. How did that happen and why? The writer did not thing this was important enough somehow, but I think it is.

Normally I'd call this sloppiness, but considering the author is a professional and this had to get proof-read, it may just be the author is being a bit disingenuous as well.

← PreviousPage 2 of 2