.. except that a large part of the kernel networking stuff (among other things) are also FreeBSD derived.
Can you please email me? (email@example.com) - we have a vested interest in getting the r-pi USB bugs ironed out on FreeBSD and would likely really benefit from the work you've done.
... do you have any other references besides the above? A FOIA document or two would be nice.
.. and we have almost stopped using Akamai too for video. We serve it ourselves now.
The video content isn't coming off of AWS. It's coming from the Open Connect Appliance Platform.
(I'm on the OCA team at Netflix.)
It doesn't have to be faster or cheaper than a commuter air flight. It just has to scale better.
I don't know how often you fly SF LA, but there are a lot of flights going on there and frequently I've been bumped or heavily delayed because of something that happened far before I even turned up to the airport.
Anything that can change how the transport system works and scales over changing loads will be welcome.
No, it's not the same.
The idea is that you use the 32 bit pointer model, with 32 bit indirect instructions, but you're doing it all using the x86-64 instruction set. Ie, the task is in 64 bit mode. The 64 bit mode includes primarily more registers, so you can write / compile to tighter code.
The stuff you described is for running 32 bit binaries that use the i386/i485/i586 instruction set, complete with the limited set of temporary registers. x86-64 has many more registers to use.
It's not just about cache lines.
Hi. I'm Adrian. I work on wireless. It won't do what you say it will do. Don't cheap out on infrastructure.
Go get Windows 3.1 and Works. Stick it in a vmware VM. Cry at how fast the VM is.
Works fine for me on chips supported by dri. The dri2 support is being nailed down now and once that's in it'll work fine on the same bleeding edge Intel hardware Linux does.
I'm the wifi guy. The WiFi is now up to date on Intel and Atheros 11n. I'd like some help with broadcom. I'll do the Intel and Atheros 11ac stuff early next year.
I'm currently evaluating power management. FreeBSD and xorg on my ivybridge lenovo x230 draw 9w when idle. We are ok at using the deep sleep states per core and package but there's room for improvement.
I'm making the turbo boost stuff work out of the box. Powerd is
I'm using an x230 in vesa mode but it works fine if you use the new DRI and xorg code. I do day to day hacking on the lenovo t400, mostly due to the cardbus slot I still use.
The only thing missing is hotplug express card.
So.. It's not perfect. 10.0 will not be laptop great. I expect 10.1 with updated dri2 and xorg along with Intel WiFi fixes and my power management stuff to be great.
Dude. OC48's are small. ISPs connect in multiples of 10GE these days.
... it's called fuzzing.
You spend a bit more time writing some randomisation into your clients so they go off and do completely ridiculous stuff. stuff you can't comprehend. That's why it's random (ie, fuzzing.)
again, this isn't new.
And yes, if you write your client simulation object(s) in something not stupid, you can scale it up to 100,000 active user simulation instances on a single server. Computers are fast.
Why do people keep saying that over and over again?
It's easy. You write a test suite that pretends to be a real user. You script it so there's some actions that aren't just "do A do B do C." You make them make errors. You have them put in garbage details. You have them fill out the forms incorrectly or incompletely. You have them skip pages or press "back".
Then you add a "pretend I'm the internet!" layer in between that simulates latency, so you make sure that your servers can handle the number of concurrent requests going on. A lot of not-so-seasoned web developers still fall for the "it worked on the LAN to 100,000 users, why not on the internet?" latency fallacy. Increased latency (due to RTT, packet drops, TCP retransmits, etc) leads to having more and more sessions going concurrently. That ties up resources at the server end.
Then you add a "pretend shit breaks!" layer. Ie, the user internet connection breaks. They forget and come back after a while, and hit the restart page. The connection dies half way during the transaction.
Then, once you've written that, you create 5 million instances of that. 100,000 per box sounds about right.
This isn't 1995. Computers are really god damned fast.
... replying from my real account.
Yeah, FreeBSD? dev boards? Any interest?
11n will work on Atheros hardware when they either/or:
* update pfsense to work against FreeBSD-10;
* start releasing snapshots of pfsense that work against FreeBSD-HEAD;
* backport the net80211, driver and userland tools from -HEAD to -8 (which I've done a few times, I've just not committed it to FreeBSD.)
11ac is a different story. I'm going to let the Linux side shake out before I start work on the FreeBSD 802.11ac support.
(FreeBSD wireless maintainer.)