Post a Comment
Pretty sad for 5 years of dev. - lqsh
More like 3 years. MS tossed out what the had and restarted the Vista development some time ago. - DevL
So the time they spent on the parts that were tossed out just disappears? Tossing out two years of development isn't a time travel trip to two years in the past, it's still two years of development.
They didn't start over after those two years; lessons learned (some about how not to do things) were used in the subsequent three years. Vista's current state can perhaps be said to have been in development for three years, but they still spent five years to get there.
Starting over after two years doesn't mean the two years never happened, it means they squandered two years, and those years can't be ignored when talking about how long it took them to produce Vista.
Edited 2007-02-12 22:41
This entire site is dedicated to operating systems. The topic is "ReadyBoost," so the discussion should be related to memory management and its effects on Vista performance. This doesn't include comments of the type: "I tried Vista, and I like XP better."
Everybody here knows that Vista has some issues that prevent it from being a smart upgrade decision for many users at this point. There are enough stories linked from OSNews that discuss these issues in general, so they don't have to be rehashed in response to a story about a specific Vista feature.
I respect your willingness to help people with their problems. But the he didn't ask for help, he simply said he didn't like Vista. That earned him (at the moment) a mod score of 5. Are you folks happy? Because I'd like to think we can do better than this.
If you don't like my comment, then move along, simple as that. I honestly couldn't care less what you think is proper or improper.
I didn't ask him the question in order to help him (though if I could be of assistance, I would help). I asked him because I happen to be a beta tester for many products, including Vista and now Longhorn server.
My issues with RTM were:
Increased system instability due to extra/continuous resource usage
File operation (including: zip operations, copy, move, rename!) performance was terrible (zip operations: >10x slower than 7-zip)
UAC ('nuff said) - specifically, the black screen flickers when it activated.
Driver support - I doubt wether my DTV tuner will ever get updated drivers even tho it was released just over a year ago. (There is native BDA support for this in Vista, but this only works with the ridiculously expensive Vista Ultimate/Premium).
I only lasted a week with Vista too.
Yeah, Windows' zip app is just awful. Don't know why it is so slow.
UAC really isn't all that bad, especially after you get your apps set up. And then once developers start writing the apps correctly (using the Users folder instead of Program Files, or other misc./stupid places for their files) you will see UAC even less.
Black flicker is definately interesting though (and definately irritating, I'm sure).
"Most people think a hard disk is a replacement for RAM. I'm afraid to ask what sort of medical procedure you recommend for this..."
I recommend mindlock and arrest. :-)
A harddisk is a replacement for RAM as well as a replacement for backups on external media. An USB memory stick is a replacement for a hard disk and a CD/DVD drive, and Bluetooth is a replacement for network, while a website is a replacement for a harddisk, and a dual core processor is a replacement for RAM, while processor cores are a replacement for clock cycles... :-)
What? this is terrible. For $15CAD more than a 1gb key, you could get an 80gb hard drive and setup a huge cache on that. It will be faster and last longer. Or partition it, and have a few extra gb's and a big fat cache.
We've hashed this out on OSNews before.. it's a gimmick that doesn't work.
I pulled up NCIX, had 80gb HD's @ $65 and 1gb USB flash @ $30.
Faster random access, yea, but it still doesn't make up for overall slower bandwidth. Besides, caching is typically sequential, no?
I once attempted to use a 2gb flash-key as a swap drive on linux during a rather trying compilation of MythTV (I think). Stealing a bit off the end of the windows HDD yielded not just better results, but substantially better results.
Random access latency is the name of the game for desktop workloads. The problem is that flash is not significantly faster for sequential access than for random access, which is a weird characteristic for a disk cache. How do you manage a cache that doesn't always outperform its backing store? Unless you do some clever tricks to bypass the flash cache for sufficiently sequential requests, you end up hurting performance.
Now, DRAM is about 10x the price of flash, so the solution isn't to just tell people to buy more memory. A good hard disk such as a WD Raptor will provide excellent paging performance with plenty of capacity left over for the OS, applications, and other small files. Use larger, cheaper drives like Seagate 7200.10 for large files such as media. Consider RAID1 for just about anything, since you get nearly double the read throughput and complete redundancy with very little write throughput penalty (those 16MB disk buffers hide this really well). Yeah, you waste capacity, but disks are cheap.
It won't be writing that often this is an app cache not a swap file. That said I still don't quite get how this is an improvement over the prefetch folder other than possibly shaving a few seconds off boot time. Maybe if I ever develop the desire to use Vista all will become clear.
It's not specifically for the superblock.
Flash ROMs have a limited number (usually 10000 to 1M) of write cycles before failure.
Current devices include logic to remap sectors so spread the changes around to avoid having the superblock (that's updated more often) to always trash the first sector. But still it's not meant as a short-time storage. Using a flash key for paging is like using an expensive book as post-its.
If I understand it corrently, Readyboost isn't exacly like traditional paging. R'Boost stores frequently used pages semi-permenantly on the flash drive. So that when App X regularly uses a particular page, it is stored on the flash drive and pre-loaded into memory next time it starts. Therefore the data stored on the memory stick is semi-persistent.
Feel free to correct me if I'm wrong, (I have a feeling I might be
)
This page might answer some of your questions...
http://blogs.msdn.com/tomarcher/archive/2006/06/02/615199.aspx
Q: Isn't this just putting the paging file onto a flash disk?
A: Not really - the file is still backed on disk. This is a cache - if the data is not found in the ReadyBoost cache, we fall back to the HDD.
Q: Aren't Hard Disks faster than flash? My HDD has 80MB/sec throughput.
A: Hard drives are great for large sequential I/O. For those situations, ReadyBoost gets out of the way. We concentrate on improving the performance of small, random I/Os, like paging to and from disk.
Q: How much of a speed increase are we talking about?
A: Well, that depends. On average, a RANDOM 4K read from flash is about 10x faster than from HDD. Now, how does that translate to end-user perf? Under memory pressure and heavy disk activity, the system is much more responsive; on a 4GB machine with few applications running, the ReadyBoost effect is much less noticable.
http://www.pcmag.com/article2/0,1895,1973121,00.asp
"Calling your right hand a girlfriend doesn't count as real girlfriend, says experts". Horrible title and article. Firstly running only 1 program at time to determine performance boost??? Nobody does so! TGpublishing is worst example of generic, far from real life situation, testing sites. Atleast sites like Anandtech try to use more realistic view and make test using multiple software same time.
http://www.anandtech.com/systems/showdoc.aspx?i=2917&p=6
Edited 2007-02-13 08:56



