littledanehren

6 ms·

1.
by littledanehren 15y ago
Note: This is my own opinion, and not that of my employer (Google) or based on trade secrets or other IP from my employer. Keeping pre-erased blocks is useful, but it can only reduce the write latency to what the chip gives you. And the chi
2.
by littledanehren 15y ago
Note: This is my own opinion, and not that of my employer (Google) or based on trade secrets or other IP from my employer. I don't know if this is actually true in SSDs that are actually used, but there is a real cost/write-performance trad
3.
by littledanehren 15y ago
Lwn and linuxnewbies aren't perfect, but they're some of the best resources we have. There are also a few books out there that are pretty useful, but they are always out of date. No need to start from scratch and replace these things;it's e
4.
by littledanehren 15y ago
Can we really credit this entirely to the increased efficiency of iPads? There are lots of usages of RAM--now people carry around cell phones with more RAM than before. And data centers can make use of tons of RAM--the main constraint is it
5.
by littledanehren 15y ago
Tell your Google recruiter about your competing offers, especially if you're considering choosing another job on this basis.
6.
by littledanehren 15y ago
Why do you want to do this, anyway? Linux already isolates processes' memory from each other. With cgroups you can ensure that resources are allocated fairly, and with chroot and namespaces you can ensure that they're securely isolated from
7.
by littledanehren 15y ago
Good point about C. You misunderstand the intent of my post though. First, as a small point, I was not seriously suggesting using thread pools to implement mlock_eventfd. That's just a way that we could play around with it short of a kernel
8.
by littledanehren 16y ago
Yeah, and there's HTML too! Complete bullshit!
9.
by littledanehren 16y ago
I don't get it. He was spending $40,000/month on his living expenses while being afraid that he wouldn't be able to afford diapers and food for his baby? How was this guy running his life so wrong? Did it somehow cost him that much to maint
10.
by littledanehren 16y ago
Yep, that's basically what we do. User-level threads are called 'coroutines' and they're scheduled cooperatively, so they don't incur the overhead of user-level preemption. Cooperative scheduling is fine for us because we can ensure that a
11.
by littledanehren 16y ago
This is something which will be good to look into in the future. It may be that, in certain system configurations, that's fine. But if there are 10,000 coroutines each with 10 MB stacks, and memory overcommitting is turned off or configured
12.
by littledanehren 16y ago
Thanks for the suggestion! I'm actually looking into doing something like that right now.
13.
by littledanehren 16y ago
I guess telling Google what they should be doing differently might work, but I wasn't really prepared in my interview to give them advice. Still got accepted, though.
14.
by littledanehren 16y ago
A problem with this is, most people don't have the money to just start investing in startups. This is one reason why startups tend to need VCs in the first place. It's not like the only thing holding people back from investing in startups i