Nullack

Member
  • Content count

    480
  • Joined

  • Last visited

About Nullack

  • Rank
    AV Forum Member

Contact Methods

  • ICQ
    0
  1. Sorry to hear that mate but atleast when you get it back it will be better
  2. Thank you Treblid too, and everyone else who has provided genuine help.
  3. Thanks Byron and Jenx Youve both explained the system has methods to protect file corruption from happening. AFAIK, the file systems used in NAS's like EXT3, ZFS etcetc feature: * Journalling * Atomic transaction control etcetc What I understand these features do is to provide levels of robustness to ensure the contents of RAM are in fact written to disk correctly. Guys I do not believe that these fatures will prevent corrupted RAM from not being written to disk. All the file system knows is that it will correctly write the contents of RAM to disk, but it doesnt know if the contents of RAM have been soft errored or not. Its the only the memory controller with ECC RAM or something like CHIPKILL that would know this, not the file system.
  4. Hey evil_josh I want to thankyou for your informed post I like the idea of buying a cheap UPS, it resolves the issue I identified earlier with not having a battery backup for the raid array to write cache to disk and as you wisely point out, overcomes the brown out issue. I respect the detail you've got into with the soft error rate. I think its not accurate and I hope you will please allow me to show why? NASA in particular has done many tests over the decades about radiation and soft errors. I have some of their research papers and also a IBM paper written for the semiconductor industry. What they all conclude on is: 1. Less voltage means a higher SER 2. Smaller process nodes means smaller transistor gate sizes and that means a higher SER and higher electron leakage 3. As the process sizes go from 120nanometers to 45 to 32 nm the higher goes the SER, but I dont have any data from the papers on how much it goes up. 4. All of the papers talk about the lack of standard ways to test SER and mention a high variance in the actual SER rate being reported between tests 5. NASA and IBM both agree that there is 2 soft errors per week in 1gig DRAM 120nm, and in 1TB of ram, one occurs every five seconds using DRAMs on a 120nm process node. 6. The higher in altitude the device is these numbers go up by heaps, and even more into space There are ofcourse benefits to smaller process nodes like the benefits treblid said - less heat, lower cost to produce per 300mm wafer etcetc What I particularly like about your post is a very good point you made - being what is the chance of a soft error bit flip corrupting memory being written to disk instead of something else. Ive tried to get answers from QNAP and Thecus on this but nothing concrete from them yet. I can live with a single file being corrupted - my worst nightmare is having a soft error in the region of memory controlling the array and that corrupting the whole array - I'd really like to know the chances of that cos that the key. Mate I have an offsite backup scheme so thats fine - just dont like the time it takes to restore from tape so I try not to have to use it. We know these types of NAS's have 1gig of SO-DIMM memory so I think what would be super would be having some really good and honest numbers of how much of a problem it is.
  5. No, less voltage and smaller gate sizes in the transistors actually increases the soft error rate. As for the rest of your verbage - you think Im paranoid??!! haha
  6. Thanks Treblid. I dont need 600mb/s DAS like storage or iSCSI or whatever What I need is *nearline* storage that is going to remain accurate and is reliable. I have offsite backups so thats ok. That is, I need my source footage and production footage to not get strange pixels in it. And I want to avoid the time it will cost me to restore from tape. I was reading a research paper that estimates soft error rates are about 2 per week for 1gig DRAM, going up to 1 every five seconds for a system with 1TB of ram. Whats relevant here is that the study is old and since then the production of DRAMs has gotten smaller. 45nanometre process nodes are in production and these are far more susceptible to bit flips from environmental radiation than the 120nanometre process node on which the study was based. I dont know how much going down in the process node size effects how much more the ram will bit flip - but I do know it will do it more. As for being paranoid Im not sure I agree There is way to prevent this from happening, it is a real event that does happen, and the added cost to me with a diy nas using an AMD cheapie chip isnt really there. Unbuffered ecc ram is about the same as unbuffered non-cec ram. My existing storage with the DAS has ECC ram.
  7. Recommend you try a process of elimination. Dont have XBMC running in RAM and see if you can replicate the fault. If you start getting into the systems level and not an application level where the fault still occurs let me know and I can help you debug it further if needed within Vista.
  8. Im self employed and work at home. I do video content creation for advertisements, film clips and so on. I'm a mcrosoft partner and also a free open source software developer. Purchases I can claim on tax, but I try to not spend money on things I dont need. According to mr Bitey this makes me a filthy tight arsed troll, but for the more reasonable individuals on this forum I can only say that dont we all not want to spend money needlessly? My inquiry is a genuine inquiry, Im not trying to waste anyones time. I would be delighted to be shown why I can use a cheap NAS and why Im wrong about ECC ram. I cant afford to have my content not be data accurate or be reliable. I could spend up to $4000, but I'd prefer to spend $40 and a case of beer if that makes sense mate Some years ago I purchased a DAS with fibre channel and a host bus adapter off ebay, but it uses old SCSI drives and they are expensive to replace plus the storage capacity is low. I have offsite tape backup - grandfather, father son revolving backups. What I need is some sort of nearline storage that is atleast 6TB, data accurate and reliable.
  9. Mr Bitey, the issue I was communicating was about using media foundation over direct show as its more or less unsupported for any serious bug fixing by MS now. Its not as you like to twist it, a lecture. You seek every opportunity you can Mr Bitey to harrass my posts - even to the point of picking at straws to do so. You will resort to personal insults and any other tactic you can, for in your mind anyone who isnt your best mate is a troll and you will hunt down trolls and harrass them off the forum. An objective yiouve said openly. It wasnt about name dropping, the fact is the XBMC guys are a good bunch of blokes that genuinely care about their software. If Byron isnt comfortable installing debugging tools then as I said there is support avenues with reporting bugs to the XBMC team.
  10. Gee if thats happening they should be reported to crimestoppers and arrested
  11. ajm - I have an existing DAS I bought off ebay. As I said in my first post "I need to upgrade my storage to atleast 6TB." If your trying to help great, otherwise if youve joined the ranks of Mr Bitey with his trolling then thats unfortunate.
  12. Thanks mate I'll check em out and look at staticice.
  13. Marshall is a NZ guy who loves helping Aussies on his software, same with Pike whos Euopean but dont hold that againt him heh - really if your tests are showing its a XBMC problem start a XBMC bug report going on sourceforge. If your dead keen on using ffmpeg libraries and XBMC, they recently added VDPAU support to their stack so maybe a Linux distro would suit what you want better? Otherwise, it is possible to diagnose whats going wrong on your vista install to know for sure. MS give away their SDK for driver development which has many debugging tools in it.
  14. Mr Bitey your completely unable to have a rational discussion on what the real causes and hence fixes are for these problems. You make sweeping assumptions that make no sense and youve proven by your actions you are simply not interested in an honest talk - your motivation is one of trolling and harassment.
  15. It has allot to do with the video driver and the refresh rate of the display beyond the decoder.