Forum Replies Created
-
Posts
-
WTH halladayrules? Did you even read my post? I don’t know how makes it clearer I both know how to clear the event log, how to use wevtutil’s syntax as my script for clearing works… The thread is not about clearing the event log, it’s about HALTING event logging…
Your script doesn’t even work as there are over
Please!?
😕
Anyone here have a idea or ‘go to‘ source like an active MSDN, TechNet, or Educational account to find out what’s going on here? I’ve exhausted mine…
😕 ❓
Worked… It appears to be the “System” (NT Kernel & System) process doing all the disk writing… The questions remain: what exactly is it writing? Why? And can any parameters be passed to the “System” via registry or command line to stop, or at least reduce all these disk writes?
I can’t feature the Windows Embedded OS products built on the same kernel could not offer at least some control over this level of i/o traffic, due to the issues this would present in many embedded scenarios — so it’s not difficult to imagine there is some level of control available.
❓ ❓ ❓
Thanks for the tip, will do next benchmark run. I’m farily convinced this is “normal” as I’ve observed the same behavior on doezens of systems but never bothered to run down what’s doing it. The actual disk write activity varies across different disk/controller systems due to how writes get cached, but the drives are still getting written to constantly and regularly.
My interest is in discovering if there is something that can be done to reduce or even stop it alltogether as these are very small frequent disk writes, which is sort of a worst case scenario of what you don’t want on an SSD due to write amplification, something the industry probably doesn’t give a hoot about, as they’d love to sell more new SSDs.
:geek:
Than you again halladayrules, what a great resource! Would you have any idea why Microsoft made installing the WS/IIS and it’s .NET componets part and parcel of installing the latest version of .NET on the Server OS platform? Ergo, why isn’t it the other way around where those .NET components needed for a particular role/feature are installed as there are a lot of .NET dependencies that don’t involve running a Web Server of some flavor…
❓
I agree with ExpertUser, the Microsoft Server operating systems have not been under the microsope of system internals hack & tweak Fans as much or for as long as the consumer Microsoft operating systems.
nuhi spent nearly a decade taking Windows XP apart, and Vista since before it was in MSDN beta (he had a good hook up). While there’s a lot of of simularities between the Microsoft operating systems, there are also unforgiving differences.
If you’re an enthusiast, that doesn’t mind trying to trade stability for smaller form factors and performance; go for it, all these project need people to help test, tweak and explore, but if you’re looking for the equivelent of an embedded Server OS; not yet…
:geek:
I really like the clean, rational, well organized interface that follows Microsoft’s design conventions! Does your SuperFetch feature patch the Kernel or something to get it working?
You guys have so much talent an dknowledge of Windows System Internals at work, have you considered taking this tool in the other direction as well toward removing/uninstalling features, bloat etc. from Server 2008 R2?
Impressive work, don’t hurry the versions! I’m sure you’re excited to get people to try it, but you’ll have more attention then you know what to do with when you release!
😉
Hi sviru, I think there are some motivated individuals on the MSFN Forums that have managed to get these to fly with various degrees of success on Server 2008 R2. It will probably be pretty hit and miss though as I think a lot of people kind of lost interest when nuhi stopped supporting vLite, and Rockers 7Lite while a nice effort is still new, and doesn’t have nuhi behind it…
😉
@halladayrules wrote:
From what I have been able to gather from reliable sources:
A hard link points to the actual file (it is cloned)
Yes but in the case of a directory, it appears to clone just the directory itself, not its contents, which works like a UNIX link on the FS level…
@halladayrules wrote:
A soft link points to just the pathname (aka a shortcut)
Not quite from Microsoft’s and Mark Russinovich’s documentation, a shortcut is an actual file that includes direct path information, where as a soft link is still a low level file system junction that makes the ‘soft link‘ appear to the operating system as if the file or directory is on a local FS path…
@halladayrules wrote:
The difference between the two is..
I’m aware of how the other differences are supposed to operate, what I want to understand is what’s going on at the FS level, and which perfoms better when used for local FS linking…
:geek:
Well halladay, I’ve been doing more in-depth reading, and while still far from certain, I think the your first and intuitive response might be closer to the actual truth. Both of these are abstrats in the file system and do behave differently and probably have different i/o overhead, which is my biggest concern, though I haven’t been able to glean which way that goes.
I think part of the problem is a lot of the documentation appears to be written by people that don’t really understand the vageries of Windows UNIX deravitives of symbolic links, junctions, and shortcuts — and they all are quite different at the low level. Still looking for some definitive documentation or word from ‘Thee Final Anuthority‘, will let ya know if I learn anything useful… Hopefully anyone else that beats me too it (which seems likely) will let me know what they discover as well…
😐
@halladayrules wrote:
Either one will create a symbolic link between local directories. The difference between them is the mklink /d uses a soft link while mklink /j uses a hard link. A few restrictions for hard links is that both links must be on the same file system and remote sharing is not supported. Soft links allow access across remote storage shares but there is increased disk I/O overhead as a consequence. If you plan to use symbolic links to offer access to an application, the mklink /J link is preferable in your situation.
Hey halladayrules, thanks for responding, I appreciate it and you do ‘rule‘!
😀
I do understand these difference between Directory Symbolic Link and Junction as far and as in depth as your recitation goes, it’s pretty much the same info I was able to glean from Microsoft’s documentation on TechNet. What I don’t understand is your reasoning for saying a Junction would be better in the application described, where none of the differences I’ve been able to glean between a Directory Symbolic Link and Junction really matter in this instance? Please say more!
😉
Just reread your post; is the increse in disk i/o why you’d recommend going with Junctions in this example? I’d think the disk i/o difference would be neglagable, but don’t have any experince here…
:geek:
3D game render performance is going to be hardware, game engine, and DirectX iteration specific; and some of my tests show that changing any one can ‘flip‘ results to favor a different Microsoft OS… Early this summer I ran Futuremark’s 3D benchmarks on x64 editions of: Windows 7, Server 2003, and Server 2008 R2. Previously I’d done this with the 32-Bit verison of Server 2003 Web and XP and found Server 2003 to turn substantially faster 3d performance then Windows XP. This time the results were odd and mixed generally giving Windows 7 small synthetic DirectX 9 & 10 performance leads over the other two Operating Systems but typically results were within the margin of error for the benchmarks.
Running real games however told another story, yielding more dramatic differences… I used three engines to try to get an idea what was going: Frostbite 1.5, id Tech 4, and Unreal Warfare 2.5… Sever 2003 was able to consistently run DirectX 9 games faster then Windows 7 or Server 2008 R2, and running DirectX 10 games Windows 7 was faster then Server 2008 R2… While Server 2008 R2 was slightly slower running Frostbite 1.5 (I didn’t have a formal benchmark to run at the time) the frame rates it turned were more consistent and less ragged then Windows 7.
I was quite formal in my testing paradigm, ran carefully ordered installations of each OS from Boot VHDs with identical service complements running, the game engines and benchmarks run from a separate drive so the tests would be consistent across all OS tests. Server 2003 was the exception as it couldn’t be run from a VHD (at that time — but now there’s a kludge that will allow this using GRUB), and Windows 7 and Server 2008 R2 get a slight disk read performance boost when run from a Boot VHD.
As my tests were only to satisfy my own curiosity, and were very tedious to setup and execute, I was pretty bored with the whole procedure, anxious to reformat, and get a working installation and the tools I use for work installed before the weekend was over so I didn’t save anything or make any make pretty graphs. It might be fun to do this all again, and format the results for presentation with newer hardware, and with a someone partner testing a parallel generation and price point platform so both ATI and Nvidia hardware/driver 3D performance can be compared…
:geek:
