> Random IO is processed first through the SSDs (the thing that they are really good at) while sequential IO short-cuts to the hard drives - which is pretty slick.
Any idea why the sequential benchmark numbers improved 4-5x when it is still "short-cutting" to the HDs?
> I was still working full time at Powerset on July 1, 2008
I'm curious how this works normally. He was employed full time at PowerSet nine months after he started working "full time" on Github. What did Powerset think of his devoting his time and energy to Github? What is the industry norm? What about the IP issues that complicate the matter? Couldn't Powerset have claimed ownership over Github since he was a salaried employee?
California has fairly strong protection for employee's intellectual property in stuff they do on their own time that is not relevant to their employer's business. This protection is, admittedly, more effective if you're employed by a small company with a focused business than if you're employed by a major corporation that does a bit of everything. But github doesn't sound much like a natural language search engine, so I doubt there were any IP issues to worry about.
Like a butt-load of positive PR? I think they calculated that the positive PR value (or maybe mitigating negative PR) was worth the incremental cost of capital from the normal markets vs. from the DOE
This is interesting and expected for evenly distributed request patterns. How about for more typical request patterns that follow power-law distributions? I would guess that it'd lead to much fewer page faults. I could write some math but does the benchmark tool let you choose a distribution of keys, which would help check that type of pattern?
He can feel confident because Facebook has a history of misleading users or oversharing data and this isn't a court room, rather the internet, and there is little consequence if he's wrong.
I'm not 100% confident that Facebook's intention was to mislead but the simple fact that these posts aren't being shown to the user(s) that they are being linked to is pretty damning.
I can be confident because Facebook spends more time and money studying user interaction with their advertisements than any other single aspect of their business. They have focus groups, they study eye movement with iris tracking, and they are leading authorities in online marketing. They don't make a change to their online ads without understanding 100% of the variables.
And, for the record, this analysis of Facebook's investment in user study isn't mere conjecture. I've been lucky enough to attend a talk by Facebook employees on their efforts.
Web.py is arguably an example of shoddy engineering in explosive software startups that did not become a medical experiment.
"The results show that foremost among the causes of growth in U.S., German and Japanese manufacturing value added is electric power consumption."[1] Now I wouldn't pay too much attention to the word "causes" in there, but...
Define your terms too precisely and there would be nothing to think about.
