Linux - HardwareThis forum is for Hardware issues.
Having trouble installing a piece of hardware? Want to know if that peripheral is compatible with Linux?
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
The fastest filesystem is XFS. You can improve on its read and write throughput by adjusting certain format parameters. There are two definitions for speed for a hard drive. One is throughput or input/output. This the amount of megabytes per second it can read and write. Another term for speed is accessing time. The lower the accessing time. The least time you have to wait for the hard drive to read a file. Samsung hard drives maybe fast on throughput, but they are horrible on accessing time.
My intention is to speed up the transfer of files from one directory/drive to another. I am using SuSE 10, it has ReiserFS set as the default filesystem. To be honest, I have noticed that NTFS under windows is faster (but faster does not justify the risks associated with NTFS). I constantly work with files larger than 4Gb (movies). Is it SuSE, the hard drive, the filesystem or a combination of all?
I am frankly suprised you have a 5400 RPM drive above 40 Gig, very few were made, and they are not common. Unless you have a laptop, in which case 5400 rpm is still more common than 7200... I would assume there should be a way to defrag, just like in Windows, but in SuSE, I am not sure what the tool is, or how well it works. The other thing to take into account is how you are transferring... http, ftp, wireless lan, wired lan, direct drive to drive on same machine.(Those are in order of speed.) You may be able to increase speed depending on how you are transferring too. You could, for instance, maybe optimize your network connection to improve things vastly.
I am sure that it is a 5400. In fact the drive model was discontinued when HP put the machine I am using on the market.
It is a matter of drive to drive transfers. No network problems. Those would be slow anyway. I think it is a SuSE thing. I have tried other distros, and they were faster. Probably because they used ext3. A trade off using SuSE?
Since you are doing drive to drive transfers One thing you'll want to check is that they are on seperate IDE channels (IE one is primary master and the other Secondary master in the IDE controler.) if they are on the same channel they will not be able to transfer as fast. An easy way to tell is watch how they are declared when the computer starts up (right at the initial boot up) or if you are confortable doing so.. just pop open the case for a sec (assuming it isn't a laptop) and see if the drives are on seperate cables or the same one.
As for the hdparm command many Linuxes tend to default to PIO mode which is much slower then DMA mode. The -d option sets DMA off and on (-d 1 for on). You might want to test the, DMA On, setting for a bit before doing serious work with it on. I have seen some flavours of Linux hang when DMA was turned on. This can sometimes be solved by turning down the transfer speed (the -X option in hdparm) However if you find you are needing to turn it down a lot it might come to a question of stability vs speed.. I think most newer Linuxes are less fussy on the DMA.. but I have no experience with Suse so you might want to see what the experiences of others are.
If you are REALLY nuts for speed you could try the unmask IRQ settings in hdparm but it has been my experience that unless you system is JUST doing the file transfer this would lead to undesired results. So definitely read up on that one first as it is less safe to play with then the DMA/transfer speed settings. (I always leave at the default. just to Flaky when not.)
7200 is vastly faster. Also 5400 cannot use 133 transfer speed, only 100. So if you get a UDMA 133mhz 7200rpm drive, that will defintely help. If your board supports SATA, unlikely in an HP, you can get 150Mhz.
Also, it is likely a 2MB cache drive now, try and get an 8MB or 16MB cahce on the drive, makes large transfers much faster.
Tom's Hardware guide will have spefics on how much improvement you can get in percentages and exact rates, but I will leave it at noticably faster.
The UDMA 100, UDMA 133, ATA-100, or ATA-133 does not mean it goes at 100 MHz or 133 MHz. The 100 and 133 is the burst transfer rate in megabytes per second. There is no difference between in throughput of ATA-100 and ATA-133 because hard drives rarely transfer that kind of data in one second. In raw throughput hard drives transfer between 20 MB per second to 60 per second. When adding filesystems to the equation, hard drives transfer between 5 MB per second to 30 MB per second.
SATA 150 means 150 MB not 150 MHz. I do not know how controller manufactures get 150 MB because 133 MB is the max bandwidth that PCI can handle.
Not all 7200 RPM drives perform better than 5400 RPM drives. 5400 RPM drives can actually be faster because of less ECC they have to do and they can squeeze more data on one platter than 7200 RPM drives. Usually but not always a high density platter increases data throughput of the drive. 5400 RPM drives produces less heat than 7200 RPM drives because it consumes less energy.
The cache in hard drives is rarely used in Linux so do not spend your money on a hard drive that has 8 MB or 16 MB. Linux uses its own cache which is very efficient for its filesystems.
ReiserFS in my honest opinion is not fast as they say. Also I had problems with it. If you are handling files as big as 4 GB or more and you have at least 256 MB of memory, use XFS instead. I have only but good experiences with XFS in any Linux distribution and with kernels as old as 2.4.19. The recovering utilities are fast and efficient. XFS is much faster than ReiserFS because XFS handles all requests in memory which is a lot faster and more efficient.
Hitachi's T7K250 (PATA or IDE version) is great for an OS drive. The accessing time is near Western Digital Raptor hard drives and it connects to your PATA controller. Unfortunately, it has 8 MB of cache.
To convert from ReiserFS to XFS, I suggest using a Live Linux boot CD and another hard drive. First check if XFS is set as modules. If it is, you have to create a ramdisk or initrd file. After that, boot up with a Live Linux boot CD. Next mount the root partition and one of the partitions on the extra hard drive. Copy the files from the root to another partition on the new hard drive using cp -av source destination. After it is done copying, unmount the root partition and format it. If you want XFS for the root partition, set the blocksize to 1024 (default is 4096 for 32-bit processors), set the agcount to 32 (default is 8), and set the real-time size to 4096 or 8192 (default is 65536) during formating. Then mount the new XFS root partition. Finally copy the files from the partion of the new hard drive to the root partition using the same copy command.
Few Big files = HIGH performance
Many Small files = LOW performance
Many Small files = HIGH performance
Few Big files = LOW performance
I do not believe in filesystem benchmarks from some person that does not do any research to correctly make a comparison. I only believe in real world benchmarks which the benchmarks are not. XFS can write multiple files at once. When I copied files from my USB flash drive which only has a throughput of 300 KB per second to 600 KB per second, it takes only a second to write on my XFS partition. Another example of XFS speed is copying multiple files from one drive to another drive. XFS is more reliable and stable than ReiserFS. I do think Reiser4 has even worst reliability and stability than ReiserFS. XFS was designed for servers and ReiserFS is not. XFS is most reliable and stable filesystem that you can ever use for a desktop. For those people that are impatient, XFS restores and mount itself in an instant. Companies rely on XFS for a fileserver.
But Journalling to journalling is a bit difficult task, I feel.
That is why people, including me, should use two hard drives and multi-processor systems. One hard drive for data and the other for journal. The mulit-processor system will cut down on the load of filesystem operations while you are using a program.