Inside the Windows Vista Kernel, Part 2

BuR4N writes "Mark Russinovich takes a look at the Windows Kernel and the changes made in Vista. In this second part he describes the workings of the features SuperFetch, ReadyBoost, ReadyBoot, and ReadyDrive and how they improve system performance."

Comments

Re:Is this secure

[atarione]

yeah it is secure

http://blogs.msdn.com/tomarcher/archive/2006/06/02 /615199.aspx

 

Q: Isn't user data on a removable device a security risk?
A: This was one of our first concerns and to mitigate this risk, we use AES-128 to encrypt everything that we write to the device.

Improve?

[Anonymous Daredevil]

This benchmark article shows that SuperFetch and ReadyBoost can help improve app launch times a bit, but mostly only if you have woefully tiny amounts of RAM in your computer.

However, this slew of benchmarks shows Vista to be slower across the board then XP.

Re:Where's the Beef?

[Rycross]

Yes, Vista is noticably more performant on my system. However it is a higher end system, and the increase in speed is due to the fact that Vista makes better use of the resources. Its an Athlon X2 system with 2 gigs of ram and an nVidia 7800 GT. Offloading rending to the card, better use of the second processor core, and using more of the RAM to cache applications, I did notice an increase in performance.

Re:Where's the Beef?

[PhrostyMcByte]

I've been using Vista for quite a while now for primarily programming and gaming. "faster" has two areas for me:

1. When using desktop applications- Vista does feel more responsive. This is probably a combination of the I/O optimizations they have done (actual speedup) and the 3d desktop keeping window movements smooth and removing that ugly redraw affect XP has (percieved speedup). Vista also seems to go from cold boot to functional desktop faster. The only OS component which is slower is explorer, because it tries to preview everything (this can be turned off).
2. When gaming, however- Vista is slower. Not by a huge amount, but it is noticable. This is probably because of the 3D drivers using a new API that doesn't seem to give games exclusive access to the card anymore.

I think Microsoft may have unknowingly shot themselves in the foot by making some of the betas public. This made a lot of the "almost-enthusiast, but not really knowledgable" people decide that because the beta had some performance quirks, the RTM must too. And they've been surprisingly loud with it.

Other than some old hardware not having drivers yet, every person I've talked to who has actually ran Vista for a week agreed it is an improvement.

Re:Why 'Ready'?

[Molt]

Hard disks are faster than Flash RAM for raw transfer speed, but the idea here is to use the Flash to cache small frequently-read files where the hard disk's latency and seek time would be the limiting factor.

Re:Where's the Beef?

[Rycross]

You need a high end system that isn't being totally utilized. I imagine that if I had a single core system with a lesser video card, it wouldn't be as apparent, if at all. Vista only operates smoother on my system because there was a lot of potential there that wasn't being utilized by XP.

I imagine that if I ran solid benchmarks for a single type of task that it would come out less than for XP, but when I multitask my perception is definately that Vista runs smoother.

Re:Is this REALLY secure?

[mpapet]

Instead of just assuming that the AES-128 is the golden key that locks cached data, please consider that their implementation may be lacking.

For example, where are they storing the encryption key? It's certainly on the PC somewhere accessible to all for now.

Security programming is hard, really hard. I don't doubt that Microsoft has very gifted security programmers, but I very much doubt that they were given free reign. Most likely they were forced to implement managerial compromises that, well, compromise the system security.

Also consider the CPU cycles required to do the encrypting/decrypting and that this is just one of MANY tasks the OS is doing with encryption-bound services. Those are just two factors that hardly constitutes speedy/secure anything.

Re:Why 'Ready'?

[|Cozmo|]

Actually flash can be a lot faster than 10mb/sec. This one I'm using is nothing special and I just benchmarked it at 21mbytes/sec doing sequential reads and 16.3mb/sec doing sequential writes. While it is still slower than modern SATA disks, that is almost the same speed my old ATAPI drive would get. Running the benchmark only uses about 7% of my cpu. There were some big changes in USB in vista in terms of CPU use.

Re:Vista seems quite slow to me

[RKThoadan]

You haven't purchased a Dell laptop recently have you. For your average business class Latitude a Core 2 Duo at 1.66 Ghz is the default processor... I just confirmed it. A Latitude D620 (a very mainstream model) has that as it's default CPU. Ghz just doesn't mean much anymore.

Re:I think I will be ReadyNever

[rev063]

Yes, you can simply put your swap partition on a flash device as opposed to the hard disk. An experienced Linux user can do this in a few seconds at the command line, and it wouldn't be hard to automate it with a HAL script if anyone thinks it's worthwhile.

That's hardly the same thing. Simply putting the swap file on a flash device is unlikely to be beneficial, and probably harmful. From what I read in the article, the Windows feature selectively uses flash when it would be faster (random access reads) but defers to the disk for block-sequential reads, which would be faster than flash memory. Seems smart to me, and according to them gets you a 20% performance boost, which is nice. It also encrypts the data on the flash drive, so you don't have sensitive info unencrypted on an easily lost or stolen thumbdrive.

Doesn't seem like Linux can do that. I'm not making any judgement on the relative benefits of Windows and Linux here, but the kneejerk fanboy "linux can do that too!" response needed to be addressed.