puzzle

63 ms·

511.
by puzzle 9y ago
There's an example of the Music pipeline going wrong: https://landing.google.com/sre/book/chapters/data-integrity.... Another reference for Fi: https://www.linkedin.com/in/pliu1&#x2F
512.
by puzzle 9y ago
It was a decade or more ago; I'm not sure there are links out there. Although legendary internally, there's at least one more recent and more epic MapReduce-based story. In order to fully understand it, though, you'd need to
513.
by puzzle 9y ago
I think the Google engineer that, years ago, missed a flight connection to write a MapReduce from the airport that restored corrupted data is perhaps a step or two beyond...
514.
by puzzle 9y ago
They don't need to search backups manually. All products are required to have automated pipelines (along with monitoring) to wipe out user data.
515.
by puzzle 9y ago
It should also have the same sensor as the Pixel 2, although not the same software, OIS, etc.
516.
by puzzle 9y ago
Depending on how much extra logic you add around the model serving library, then you are exposing your source code, too. Maybe you care, maybe not. Re: Windows, there's no TF serving for that, so TF is no better than other Python frame
517.
by puzzle 9y ago
It sounds more like Matt Cutt's external role. If I were still at Google, I'd send you a welcome email and point you to the few things that I thought you got wrong in the past (mainly due to lacking the same context as those on th
518.
by puzzle 9y ago
Google TTS supports both. Maps prefers the cloud one so that you're not at the mercy of whatever TTS your manufacturer has installed (Samsung uses their own, right?), but will fall back to the local one it there's no connectivity.
519.
by puzzle 9y ago
Back when I first (and last) had to write my own text editor, 30 years ago, memory was at a premium, so I did what everybody else in the old Commodore days did: use the screen buffer, a paltry 40*25=1000 bytes, as additional memory. When yo
520.
by puzzle 9y ago
Bummer: I was expecting some reverse engineering of the specific algorithm used by the board.
521.
by puzzle 9y ago
I know how Google infrastructure worked up to a couple of years ago, including user-specific storage. What you describe would be a very large undertaking across a large number of projects. Due to legal restrictions, Maps has something ad-ho
522.
by puzzle 9y ago
How did Google comply? There are no datacenters in Russia and it shut down the only engineering office there. https://www.theguardian.com/world/2014/dec/12/google-closes-...
523.
by puzzle 9y ago
Most of the rewrites are behind the scenes. You'd only notice if you looked e.g. at latency, reliability and trigger rates. I think Scholar is even older than Borg.
524.
by puzzle 9y ago
Wow, both Alex and Anurag are still working on Scholar. They started the project almost fifteen years ago. Google infrastructure all around it changed entirely over that time; I'm wondering if this is another major rewrite or "jus
525.
by puzzle 9y ago
Perhaps they jumped the gun? It's obvious that work has been ongoing to e.g. automate GPU provisioning on Google's container OS image: https://github.com/kubernetes/kubernetes/pull/45136
526.
by puzzle 9y ago
Those are third-party numbers. The "trillions" statement came from Google, where nobody conflates searches on Gmail with those on Web search. Your original message implied that Google inflates search numbers; I'll only say th
527.
by puzzle 9y ago
Since when search metrics include YT, Maps and... Gmail?!? Unless you're talking about universal search results, I've never seen the numbers for all those products conflated.
528.
by puzzle 9y ago
What's the Enterprise pricing?
529.
by puzzle 9y ago
Three years ago or more, Bigtable served that much traffic in less than fifteen minutes: https://twitter.com/akpurtell/status/463747917782589441
530.
by puzzle 9y ago
Speaking of unconventional system stuff in Go, Google has a Linux boot FS, which, instead of featuring the usual BusyBox, ships with an init, the Go toolchain and a source tree. The /bin replacements are symlinks as expected, but the t
531.
by puzzle 9y ago
The DoD black budget looks even less like chump change, since it's at least two orders of magnitude larger than Kaspersky's worldwide revenues, yet it's always been top secret: http://www.washingtonpost.com/wp
532.
by puzzle 9y ago
More likely that not, the sources and methods behind the evidence are classified. See all the Russians that started dying under mysterious circumstances or got arrested, nominally for treason, after the US election.
533.
by puzzle 9y ago
I'm not sure if it's still the case, but there used be ample variability of compliance in Google's transparency report across the globe. Outsiders assumed that in the US the percentage was higher because Google was in bed wit
534.
by puzzle 9y ago
It's not the same at all, but yesterday my flight had 150mph head winds (compare to Irma's 185) and you couldn't even notice. I only did because of the news coverage and wanted to compare.
535.
by puzzle 9y ago
Pseudonym?
536.
by puzzle 9y ago
The real reasons for both have probably more to do with flash writes being slow and increasing the chances of running low on space. Probably more chances for the CPU to go into thermal throttling, too. Not everyone has the latest hardware.
537.
by puzzle 9y ago
Why would parallel downloads be a good thing? I can think of so many drawbacks.
538.
by puzzle 9y ago
When asked about releasing a paper a few months ago at Next, it didn't sound like a high priority, maybe because it wasn't too exciting or not too many people would be interested. I disagree, but perhaps, if enough people ask Niel
539.
by puzzle 9y ago
Does LOAS use Titan, then? https://twitter.com/jbeda/status/715373975182807040 Will you convince Niels to publish a paper?
540.
by puzzle 9y ago
I'd be concerned about code, too. It's not unheard of for a single bit flip to make petabytes go away (true story).

More ›