• Content Count

  • Joined

  • Last visited

Community Reputation

0 Neutral

About cracauer

  • Rank
  1. cracauer

    7200.10 upgrade to 250G/plane

    Any idea when we'll see 750s based on the platters?
  2. If you do the test, please make sure to capture the CPU load with top(1). Anything else, in particular using getrusage (inside the benchmark) doesn't correctly attribute the CPU load from delayed writes out of the filesystem buffer cache. You can re-use my scripts, they have that top output catcher.
  3. They should be in use given that he mentioned thinking about trying an online level migration (5 to 0+1, which it will probably not allow). Anyway, the biggest issue with software vs. hardware RAID isn't always speed. I used the NVIDIA fake RAID for a while and it was pretty fast in RAID 0+1. Things like processor usage, reliability, and problems with degraded arrays are the real problems. For example, a drive started acting up. Did NVRAID notify me there was a problem? Nope. Does NVRAID monitor S.M.A.R.T. values? Nope. When the drive started making weird noises and basically failing, did NVRAID notify me? Nope. Instead it crashed my entire machine. Most fake RAID implementations on motherboards are pretty poor... they're just trying to beef up the feature list. But this kind of RAID implementation really defeats the point of having RAID period. The Intel chipset on his motherboard is probably better, but some of these issues will probably be present with any software RAID solution. And even if it is just as fast with RAID-5, CPU usage might be high and the array performance would take a dive if your CPU was utilized for something else. I wasn't talking about onboard SATA "raid". I would never use that. I'm talking pure software RAID. And yes, it survived disk failures just fine for me and behaved as expected during recovery, including reboots with the degraded array (which is often a problem for onboard SATA raid). Processor usage is never over 26% for RAID-5 write and filesystem allocation overhead combined. The points you make are perfectly valid and the reason why I don't use onboard sata raid. But they do not apply to real software RAID (given a good implementation which Linux' md is). Results here:
  4. Are the drives still empty for an experiment? Can you compare to Linux software RAID ("md"), on the controller and on whatever onboard ports you have?
  5. cracauer

    Western Digital Caviar WD2500JD

    I know some IBM drives back in the UDMA-33 days had TCQ, when used with the Intel BX controller. But it seems it never went popular. I just re-read the reviews of the two drives in question and what caught me eye is that the reliability database showed the seagate to be better than 78% (97% for previous models) and the WD better than 3% of the drives (and 22% for the previous model). I tean towards the Seagate based on that, but the missing 50 GB per disk will hurt me. I think I made up my mind on PATA for flexibility.
  6. cracauer

    Western Digital Caviar WD2500JD

    Hi, folks, I am going to choose between this drive (WD2500JD) and a Barracuda 200 GB, or respective PATA versions. I have several questions for those in the know: 1) I know SATA(-1) doesn't have tagged command queueing, but do the PATA versions of this drive have it? 2) I assume the PATA-100 drives are not actually slower than the SATA versions, as long as I keep one on each channel, right? 3) from the reviews it seens the WD is a little faster for most tests, and while the seagate is quieter idle it is louder when in use. And a bit hotter which means it consume more power. So far correct? I tend to the WD drive based on this And to PATA because I often need to connect many drives to one motherboard and being able to temporarily connect multiple drives to one channel is a huge advantage. Also, I'd like to be able to use the drives on external USB/firewire enclosures and there don't seem to be SATA ones available, and even if, I already have PATA ones. %% Longer story/more info: My requirements, in order of importance: - reliability (I just had 2 out of 3 Maxtor 160 GB failing and will replace the whole array) - capacity (want to keep number of drives low), and the WD has a 250 GB version - low idle noise - low power consumption/heat - seek time I use 3-4 drives in RAID-0 and RAID-5. Since the RAID-0 part will bring sustained throughput right to the PCI bus limit I don't really care for throughput, but there's still a noticable difference in seek times, especially in the RAID-5 parts of the array. %% If you are still with me, thanks for reading. Do you think I'll be better off with the WD or the Seagate? Thanks Martin