Windows 8: Building robust USB 3.0 support | Hacker News Reader

Windows 8: Building robust USB 3.0 support

blogs.msdn.com

This article is not actually about USB. Neither about Windows 8.

What's interesting is that Microsoft uses model checking on real shipping products. In fact it has been pushing static checker (SLAM and SDV) for years.

Other operating systems including Linux have their own verification projects, but they all appear to stay academic (though look promising).

We can expect more reliable operating systems in this decade. Maybe.

Actually looking foward to USB 3, not the speed but the fact there is command queuing for usb storage, which makes ssd performant (without queuing you get far fewer iops, factor of about 10 from memory)

Can you elaborate on what command queuing is, and whether it helps sequential or parallel transfers?

Command queueing allows the device controller to handle requests out of order. Since it has intimate knowledge of the hardware it can order requests for maximum throughput.

It's worth noting this feature is used in hard drives, and was a Big Deal when it first came out (Google "NCQ")

Actually a USB bus doesnt know much about the hardware, but can forward multiple simultaneous (SCSI) requests and if the hardware is fast it can deal with them in whatever order it wants.

If you benchmark an SSD with and without queueing you get very different speeds; with hard drives there is still a difference, although not as much.

EDIT benchmarks (Intel SSD)

no queuing, 4k reads: 3965 iops depth 4 , 4k reads: 16,900 iops depth 8, 4k reads: 25,700 iops depth 16, 4k reads: 32,600 iops depth 32, 4k reads: 38,800 iops

The benefit decreases with larger reads, but still exists.

USB 3.0 is cool and all, but I'm not sure I'd start the Win8 engineering discussion there. With that said, it's probably a cool feature that doesn't require showing off any new UI.

Meanwhile Apple is building robust Thunderbolt support.

This is starting to look like a move to have engineers promoting various aspects of Windows 8. Great stuff, and I hope it keeps up.

These two videos make me feel like I want a copy of Windows around again. In particular, the showing multiple apps on screen, the way that was done just seems right, and pretty much the way I hoped the iPad would work, but sadly no.

I found the part about designing the system with flowcharts in Visio and generating code based on the flowcharts, and apparently the model is validated based on the flowcharts as well.

What did you find it?

Linux started supporting USB 3.0 in the September 2009 release of the 2.6.31 Linux kernel.

http://kernelnewbies.org/Linux_2_6_31#head-2c74ce6091fb2feb4...

And this matters in this context why?

Your parent probably glossed over "robust" and imagined that Windows 7 must not actually support USB 3.0 yet.

Actually, after reading the entire article that is exactly the impression it gives.

-edit- A quick google also leaves the impression that Windows support for USB 3 is new.

http://www.google.com/search?q=windows+usb+3+support

It's certainly not native in Windows 7, but I don't see how you get the impression it is not supported at all.

Isn't native support the issue at hand?

In this context, "native support" means built into the OS as shipped.

As with most emerging technologies, thinks tend to show up as 3rd party drivers first and are later integrated into the next shipping Windows version or major service pack.

Typically, a USB 3.0 device would ship with a driver, or the OS would pull it off of Windows Update for Windows 7.

nope. just regurgitating a fact i remembered from some time ago.

dramatic responses are entertaining... but really guys wtf?

My interpretation is: "Without the source code, there is no way of telling if it actually is robust or not, regardless of what they say"

Yes, yes you can. Robustness isn't determined by auditing -- we know this doesn't work in practice. Robustness is determined by beating on it and seeing how it breaks; this is the point of heavy testing, including fuzzing in the case of drivers. You absolutely don't need source for this.

Single person testing on a single prototype device vs a team testing on a custom-built test device that was designed after going through and capturing the behaviors of many devices.

Guess which one will work better with a wider range of devices?

Reply on news.ycombinator.com