I'm wanting to use ram cache on my system drive, but I've heard reports of data corruption. Can data corruption be completely avoided by setting it to read only mode? My system drive is a 5400RPM drive and I can't afford an SSD yet. I'm on XP 32-bit.
Also, what is the difference between volume edition and disk edition? Which should I be using on a system drive?
Data corruption?
Re: Data corruption?
Data corruption can't be completely avoided in any case, even if you don't use FC. But risk would be minimal with read only caching.
Disk edition caches whole drive - all your partitions (C, D, E etc) from same drive. Partition edition caches separately each partition.
For example with DE you can use 1GB of ram to cache both C: and D:, and if D: became more actively used then most of cached data would be from D:.
If you use PE and give 512MB to C: and 512MB to D:, then no matter how you use d: and c:, you will always have 512mb cache for each partition.
So disk editions dynamically change balance between partitions. Hope it helps.
Disk edition caches whole drive - all your partitions (C, D, E etc) from same drive. Partition edition caches separately each partition.
For example with DE you can use 1GB of ram to cache both C: and D:, and if D: became more actively used then most of cached data would be from D:.
If you use PE and give 512MB to C: and 512MB to D:, then no matter how you use d: and c:, you will always have 512mb cache for each partition.
So disk editions dynamically change balance between partitions. Hope it helps.
Re: Data corruption?
Manny I'm sure everyone appreciates your input but you gotta stop passing bad information, it just pollutes the water and makes it harder for people to learn. Again, there is no such thing as data corruption on reads, the OS has numerous error checking routines that ensure data integrity. Since data in RAM is transit it will be reloaded and refreshed and any bit-rot will be taken care of.
Ased.
Many people have multiple partitions on their hard drive to separate different data types. For example your OS and your Pics/Vids/Docs etc. The reason why we do this is to isolate the data so we can manage it better. What does that mean exactly? Lets say you have your OS on Partition 1 and your Data on Partition 2 (you'll probably have a hidden partition as well which is 0). In the instance that your OS fails and you need to reinstall you can do so, without putting your data at risk (though it is very important to always have 2 backups on different media types, say DVDr and say a Web backup).
Since the data is separate we can setup specific resources for one type of data over the other. So we can cache just the data, just the OS or both can be given their own resources. So as Manny explained you can give a L1 cache of 1GB to the OS and a different L1 cache of 6GB to your Data.
If you don't follow or don't have this setup don't worry about it, full disk is probably what you want. You'll also have the bonus of caching the partition table which tells the OS where the data is physically located on the storage device.
Ased.
Many people have multiple partitions on their hard drive to separate different data types. For example your OS and your Pics/Vids/Docs etc. The reason why we do this is to isolate the data so we can manage it better. What does that mean exactly? Lets say you have your OS on Partition 1 and your Data on Partition 2 (you'll probably have a hidden partition as well which is 0). In the instance that your OS fails and you need to reinstall you can do so, without putting your data at risk (though it is very important to always have 2 backups on different media types, say DVDr and say a Web backup).
Since the data is separate we can setup specific resources for one type of data over the other. So we can cache just the data, just the OS or both can be given their own resources. So as Manny explained you can give a L1 cache of 1GB to the OS and a different L1 cache of 6GB to your Data.
If you don't follow or don't have this setup don't worry about it, full disk is probably what you want. You'll also have the bonus of caching the partition table which tells the OS where the data is physically located on the storage device.
Re: Data corruption?
Is there any benefit of this over O&O's CleverCache?
Re: Data corruption?
I never said it will be corruption on read, but you still can get data corruption with read only block level cache. For example when somehow data in cache got obsolete, and you try to find empty cluster to write your file, and cache says that it is free. But it is not. It is possible but with very tiny chance. So what everybody need to understand that you can't ever be 100% secure. In all senses. And you can get data corruption any moment even when your pc is powered off.dustyny wrote:Manny I'm sure everyone appreciates your input but you gotta stop passing bad information, it just pollutes the water and makes it harder for people to learn. Again, there is no such thing as data corruption on reads, the OS has numerous error checking routines that ensure data integrity.
So regular backups, raid 1, cloud storage will put you on the safe side. And you look to optimistic for me, and you try to make people more calm then they should be. If some one ask about data corruption, then probably this is the most important thing for him. So he should know and understand that everything that he installs, especially if it is something low level and affects file subsystem - increase chances to loose data. Even some performance monitors can cause BSODs and make you loose data. So if you can give guarantee that FancyCache or any other software can't cause dataloss, then great. But I can't. And had never seen such guarantees in any software.
But afterwords, as i told before, chance is small, and probably not differs much from same chance in CLEAN Windows installation.
HAHA, say it to partially corrupted RAM chips. And also please name most known OS data integrity routines, and what they cause when they finds error.dustyny wrote: Since data in RAM is transit it will be reloaded and refreshed and any bit-rot will be taken care of.
Last edited by Manny on Sat Dec 22, 2012 12:24 am, edited 1 time in total.
Re: Data corruption?
They are complete different.Ased wrote:Is there any benefit of this over O&O's CleverCache?
CleverCache should be compared to default Windows Cache, basically it just provide some settings to Windows cache, truly i did not notice any changes in work using CleverCache. So i would not recommend using CleverCache at all.
Re: Data corruption?
Okay I installed it on a test machine. XP SP3 32bit, and I tested opening Photoshop CS4 on the boot drive (7200 RPM)
I did not do any timed tests, I just made a guess.
Without any caching software, Photoshop opened slowly on the first boot. I closed it and re-opened it, and the second run opened significantly faster (about 3x faster)
With O&O CleverCache I repeated the test. The first time was just as slow as without any caching software. The second time was also roughly 3x faster - so no improvement. I opened it a 3rd and 4th time and still no improvement.
I uninstalled O&O CleverCache and installed the latest FancyCache beta. I ran it with the default settings, initialized caching, and then repeated the Photoshop test. Long story short, I saw absolutely no improvement over having no caching software.
What gives?
I did not do any timed tests, I just made a guess.
Without any caching software, Photoshop opened slowly on the first boot. I closed it and re-opened it, and the second run opened significantly faster (about 3x faster)
With O&O CleverCache I repeated the test. The first time was just as slow as without any caching software. The second time was also roughly 3x faster - so no improvement. I opened it a 3rd and 4th time and still no improvement.
I uninstalled O&O CleverCache and installed the latest FancyCache beta. I ran it with the default settings, initialized caching, and then repeated the Photoshop test. Long story short, I saw absolutely no improvement over having no caching software.
What gives?
Re: Data corruption?
Ased,
First thing I'd point out is your testing isn't going to give you a good idea of what's going on. The easiest way to test is by using benchmarking tools, they aren't always accurate but you can use them to measure. I'd suggest using ATTO, Anvil Pro, they'll show very different results (Anvil will be much lower). Turn Fancycache off, run the benchmarks, record the results that is your baseline. Then turn Fancycache on and run them again that should give you a rough idea of what Fancycache is doing .
I'd also point out that if you are on an older PC (I'm saying this based on your OS), your L1 might not be all that much faster then your drive, 3x might be about right. There are overhead costs to running this sort of software, so your L1 cache won't run at the speed of your H/W. My ram tops out at 16GBs but I only see around 3GBs on my L1, that's on the current top of the line Intel CPU architecture.
Also keep in mind Photoshop is never going to just pop open. It's a monster application and it does a lot of processing when it starts up (scanning & loading plugins, fonts, different graphics engines, etc). I run Photoshop off a bunch (8) of really fast SSDs and it still takes a few seconds before it opens. Each program is different and the performance increase you see will vary wildly based on how the program was written.
Since you've mentioned Photoshop. If you are trying to tune your system to get Photoshop running as fast as possible I'd also recommend running a Photoshop benchmark. I haven't needed to do this in a while but it's pretty straight forward, here's a one I found can't say it's good or not but I'm sure you can find others if you search around a little.
http://www.hardwareheaven.com/photoshop.php
Manny
First thing I'd point out is your testing isn't going to give you a good idea of what's going on. The easiest way to test is by using benchmarking tools, they aren't always accurate but you can use them to measure. I'd suggest using ATTO, Anvil Pro, they'll show very different results (Anvil will be much lower). Turn Fancycache off, run the benchmarks, record the results that is your baseline. Then turn Fancycache on and run them again that should give you a rough idea of what Fancycache is doing .
I'd also point out that if you are on an older PC (I'm saying this based on your OS), your L1 might not be all that much faster then your drive, 3x might be about right. There are overhead costs to running this sort of software, so your L1 cache won't run at the speed of your H/W. My ram tops out at 16GBs but I only see around 3GBs on my L1, that's on the current top of the line Intel CPU architecture.
Also keep in mind Photoshop is never going to just pop open. It's a monster application and it does a lot of processing when it starts up (scanning & loading plugins, fonts, different graphics engines, etc). I run Photoshop off a bunch (8) of really fast SSDs and it still takes a few seconds before it opens. Each program is different and the performance increase you see will vary wildly based on how the program was written.
Since you've mentioned Photoshop. If you are trying to tune your system to get Photoshop running as fast as possible I'd also recommend running a Photoshop benchmark. I haven't needed to do this in a while but it's pretty straight forward, here's a one I found can't say it's good or not but I'm sure you can find others if you search around a little.
http://www.hardwareheaven.com/photoshop.php
Manny
Yes you did and you just said it again in this sentence. Reed–Solomon & LDPC error correction, will handles this on the H/W level.I never said it will be corruption on read, but you still can get data corruption with read only block level cache.
See the problem here is you've jumped from what isn't dangerous read caching to what actually is. Here's my frustration with you, you make very obvious mistakes and for some reason you insist that you are correct when at best you're partially right if not completely wrong. Cache data is going to become obsolete, any information being held in ram will be corrupted over time the H\W and the OS will handle of most of this but with software like Fancycache the programmers SHOULD (I have no idea if Romex does or doesn't) have their own routines to detect and correct this as well.For example when somehow data in cache got obsolete, and you try to find empty cluster to write your file, and cache says that it is free. But it is not. It is possible but with very tiny chance.
Obviously but if you being an alarmist, read operations are for all effective purposes safe. They can and do fail and that is unavoidable.So what everybody need to understand that you can't ever be 100% secure.
Yes it's called bit rot and it's a result of density of the platters.In all senses. And you can get data corruption any moment even when your pc is powered off.
You mean say it to the instability that you're whole system will experience if you have a bad chip.?? Data in RAM is constantly refreshed, that's how RAM works. It's a physical layer issue, electrical signals degrade over time (Ohm's law) and they need to be regularly refreshed..HAHA, say it to partially corrupted RAM chips.
Re: Data corruption?
So if for you CleverCache works better, then use it. For me it was just opposite to your results. First of all i'm using Windows 7, it has pretty good default caching. I think CleverCache gives to your XP same cache features that already in Win7. Basically on Win7 when you close app, and then open it again, it is always much faster. For example 4GB of ram and Tauron-64 X2 1.8 Ghz. Second launch of Visual Studio is 1-2 seconds. First one 30 seconds. So this point is clear for me.Ased wrote:Okay I installed it on a test machine. XP SP3 32bit, and I tested opening Photoshop CS4 on the boot drive (7200 RPM)
I did not do any timed tests, I just made a guess.
.....
What gives?
What is still not clear - that why FancyCache did not give you an advance. First of all I need to know how much free memory you have before PS launched, and after. And how many RAM you have dedicated to FC. I would say that less then 1GB of Fancy Cache don't give it work correctly, it will be overwritten all the time. So it looks like an issue for me. Please provide more details about your pc and experiment and then we can think together and post a bug report in case. Also please open performance monitor and keep it opened while launching PS, and please post screen shot of it, after PS is launched.
Last edited by Manny on Sun Dec 23, 2012 3:06 am, edited 1 time in total.
Re: Data corruption?
Dustyny.
There are a lot of error correction codes, and some are even used, but there is no any ECC that can find and fix ANY bit change. Even with hashing you can get such set of errors that hash remains the same. Just with low chance. But the more errors ECC can detect or correct - the more overhead to size of data it gives.
Also i'm not sure that you completely understand how general fault chance is calculated for serial connected elements. The simple rule - general fault chance is higher then the biggest chance among the elements.
So just decompose into components the full process from hdd to cpu and back. And try to write down for each block ECC algorithm. And find the weakest part.
So ECCs that you have mentioned that used for storing data on hdd, works well, and some time even can repair bed block with remapping and without data corruption. But not always. Other links in the chain are much less protected. For example RAM for servers has ECC and that cause one extra chip on the module. Home users don't want to pay more for ECC and loose ram size. So it not protected much.
Also ram check software has a lot of different patterns, to find the error for sure. Some issues can be found with very specific scenarios. Yes mainly memory cell just works or not. But some time it looses charge non linear and depending of near cells. And when it is time to recharge, controller can't understand for sure is it HIGH or LOW, and changes one bit value. And when you are lucky you can have windows and drivers never get important data in that cell. So System will perform fine, but apps may work not correct. Same as caches.
So in general caching with non ECC RAM - is not professional idea. That is why most of enterprise RAID controllers has it's own EEC cache.
The other thing. That you know something, and i know something, may be you know something better. But i know that no matter how cool you are, there is always something that you don't know. So you can very long say to me how solution safe is solution, but i will always say: "There is a way how it will go wrong! And it will!" So for me this is GREY-BOX testing, i don't know how exactly work all PC subsystems (i don't need to), to find issues in it, and write them down. So that is called experience, and my experience says - the more software you install - the more chance of error. Especially when software written in C++ and install drivers and hooks.
And yes I prefer to be alarmist with people, but i still uses FancyCache with DefferWrites enabled, but i can control my environment. And i have no idea what the other guy can do to his pc, then came here and say "Hey man! You said it is 100% safe, and i have placed my cat into microwave, and now you will pay for that! " If some one is smart enough to maintain caching and all possible issues - then he would not post such questions, and if he does, probably it is not good idea to advice him to use caching at all. General rule - "Afraid? - Don't use!" especially when it is in BETA.
There are a lot of error correction codes, and some are even used, but there is no any ECC that can find and fix ANY bit change. Even with hashing you can get such set of errors that hash remains the same. Just with low chance. But the more errors ECC can detect or correct - the more overhead to size of data it gives.
Also i'm not sure that you completely understand how general fault chance is calculated for serial connected elements. The simple rule - general fault chance is higher then the biggest chance among the elements.
So just decompose into components the full process from hdd to cpu and back. And try to write down for each block ECC algorithm. And find the weakest part.
So ECCs that you have mentioned that used for storing data on hdd, works well, and some time even can repair bed block with remapping and without data corruption. But not always. Other links in the chain are much less protected. For example RAM for servers has ECC and that cause one extra chip on the module. Home users don't want to pay more for ECC and loose ram size. So it not protected much.
Also ram check software has a lot of different patterns, to find the error for sure. Some issues can be found with very specific scenarios. Yes mainly memory cell just works or not. But some time it looses charge non linear and depending of near cells. And when it is time to recharge, controller can't understand for sure is it HIGH or LOW, and changes one bit value. And when you are lucky you can have windows and drivers never get important data in that cell. So System will perform fine, but apps may work not correct. Same as caches.
So in general caching with non ECC RAM - is not professional idea. That is why most of enterprise RAID controllers has it's own EEC cache.
The other thing. That you know something, and i know something, may be you know something better. But i know that no matter how cool you are, there is always something that you don't know. So you can very long say to me how solution safe is solution, but i will always say: "There is a way how it will go wrong! And it will!" So for me this is GREY-BOX testing, i don't know how exactly work all PC subsystems (i don't need to), to find issues in it, and write them down. So that is called experience, and my experience says - the more software you install - the more chance of error. Especially when software written in C++ and install drivers and hooks.
And yes I prefer to be alarmist with people, but i still uses FancyCache with DefferWrites enabled, but i can control my environment. And i have no idea what the other guy can do to his pc, then came here and say "Hey man! You said it is 100% safe, and i have placed my cat into microwave, and now you will pay for that! " If some one is smart enough to maintain caching and all possible issues - then he would not post such questions, and if he does, probably it is not good idea to advice him to use caching at all. General rule - "Afraid? - Don't use!" especially when it is in BETA.