Re: Re: SSHD trouble
Title: Message whers all the funny stuff??? I loved that place.!! jag elske det --- Regards Peter Blumen
Re: Re: SSHD trouble
Title: Message whers all the funny stuff??? I loved that place.!! jag elske det --- Regards Peter Blumen
USB drive enclosure review
I've just obtained a USB device that runs a 3.5" ATA hard disk. It supports USB 1.x, USB 2.0, and firewire connections, but I only tested USB. When using USB 1.x (old Thinkpad) it was horribly slow, data transfers only barely exceeded 1M/s. I did some experiments using software RAID-1 with an internal hard disk and a USB 1.x connection and the results were pretty poor. It appears that Linux software RAID doesn't seem to notice that one drive is slower than the other and take this into account when scheduling requests. In fact it wouldn't need to take actual speed into account, if one disk has a number of writes queued and the other has none, then it would make sense to send a read request to the disk with an empty queue. Using USB 2.0 the data transfer rates I achieved were as high as 31800K/s (according to Bonnie++ and iostat) with a 3yo IBM 46G ATA disk. The disk in question can deliver as much as 33M/s if the data is contiguous and at the start of the disk, so it seems reasonably clear that at the moment the ATA disk speed not the USB 2.0 speed will be the bottleneck. I did experiments of using software RAID-1 for a backup on my Thinkpad. I converted all file systems to split arrays by using resize2fs to reduce the file system size by 500K, then making a software RAID device with an /etc/raidtab such as the one below. After making the software RAID I used resize2fs on the /dev/md device to increase the file system size to the maximum permitted by the software RAID. It seems that the amount of space used by RAID meta-data varies so I could not make the file system take the exact space needed first (but it's OK as resize2fs is fast). raiddev /dev/md2 raid-level 1 nr-raid-disks 2 persistent-superblock 1 device /dev/hda2 raid-disk 0 device /dev/sda2 failed-disk 1 After doing this I could plug in the USB hard disk and start a "backup" with the command "raidhotadd /dev/md2 /dev/sda2". Once the array was synchronised the command "raidsetfaulty /dev/md2 /dev/sda2" would stop sda2 being used in the array, and "raidhotremove /dev/md2 /dev/sda2" would make the software RAID cease it's association with the USB device (thus making it free to be unplugged). With USB 2.0 this should allow a single large hard disk to be used to backup a number of desktop workstations. The USB device should cost about $200 (cheaper than the disk you will probably put in it) and makes a very cheap portable backup device. Also with the performance of USB 2.0 the same backup technique can be used for low-end servers. For a situation where you have a Linux server with two hard disks in a RAID-1 array you can easily use the same backup method by making it a three disk RAID-1 (if using software RAID), or run software RAID-1 on top of the hardware RAID (if using hardware RAID). One of the advantages of this is that when you run the raidsetfaulty command you get an instant snap-shot of the file system state at that time, unlike backup programs which will take various files at different times and may give inconsistant (and possibly unusable) data sets. Also for situations where you have a large number of small files (EG a mail server) using this technique for backup will reduce the number of seeks and therefore increase the backup speed significantly (as long as a single disk can deliver the required performance - it will be active for the duration of the "backup"). The standard Linux USB mass storage module in 2.6.0 works well with it. The device has a solid aluminium case with a small fan to blow air through it. When operating the entire case gets warm so heat is dissipated from all sides (the fan seems to small to actually blow heat out of it, but I think it does some good in helping to move heat from the hard disk to the case). The device I tested is made by "New Motion" and appears to have model number N1106. It seems to be a new model as I couldn't find anyone selling it through a quick google search, however several older New Motion products are listed. -- http://www.coker.com.au/selinux/ My NSA Security Enhanced Linux packages http://www.coker.com.au/bonnie++/ Bonnie++ hard drive benchmark http://www.coker.com.au/postal/Postal SMTP/POP benchmark http://www.coker.com.au/~russell/ My home page
USB drive enclosure review
I've just obtained a USB device that runs a 3.5" ATA hard disk. It supports USB 1.x, USB 2.0, and firewire connections, but I only tested USB. When using USB 1.x (old Thinkpad) it was horribly slow, data transfers only barely exceeded 1M/s. I did some experiments using software RAID-1 with an internal hard disk and a USB 1.x connection and the results were pretty poor. It appears that Linux software RAID doesn't seem to notice that one drive is slower than the other and take this into account when scheduling requests. In fact it wouldn't need to take actual speed into account, if one disk has a number of writes queued and the other has none, then it would make sense to send a read request to the disk with an empty queue. Using USB 2.0 the data transfer rates I achieved were as high as 31800K/s (according to Bonnie++ and iostat) with a 3yo IBM 46G ATA disk. The disk in question can deliver as much as 33M/s if the data is contiguous and at the start of the disk, so it seems reasonably clear that at the moment the ATA disk speed not the USB 2.0 speed will be the bottleneck. I did experiments of using software RAID-1 for a backup on my Thinkpad. I converted all file systems to split arrays by using resize2fs to reduce the file system size by 500K, then making a software RAID device with an /etc/raidtab such as the one below. After making the software RAID I used resize2fs on the /dev/md device to increase the file system size to the maximum permitted by the software RAID. It seems that the amount of space used by RAID meta-data varies so I could not make the file system take the exact space needed first (but it's OK as resize2fs is fast). raiddev /dev/md2 raid-level 1 nr-raid-disks 2 persistent-superblock 1 device /dev/hda2 raid-disk 0 device /dev/sda2 failed-disk 1 After doing this I could plug in the USB hard disk and start a "backup" with the command "raidhotadd /dev/md2 /dev/sda2". Once the array was synchronised the command "raidsetfaulty /dev/md2 /dev/sda2" would stop sda2 being used in the array, and "raidhotremove /dev/md2 /dev/sda2" would make the software RAID cease it's association with the USB device (thus making it free to be unplugged). With USB 2.0 this should allow a single large hard disk to be used to backup a number of desktop workstations. The USB device should cost about $200 (cheaper than the disk you will probably put in it) and makes a very cheap portable backup device. Also with the performance of USB 2.0 the same backup technique can be used for low-end servers. For a situation where you have a Linux server with two hard disks in a RAID-1 array you can easily use the same backup method by making it a three disk RAID-1 (if using software RAID), or run software RAID-1 on top of the hardware RAID (if using hardware RAID). One of the advantages of this is that when you run the raidsetfaulty command you get an instant snap-shot of the file system state at that time, unlike backup programs which will take various files at different times and may give inconsistant (and possibly unusable) data sets. Also for situations where you have a large number of small files (EG a mail server) using this technique for backup will reduce the number of seeks and therefore increase the backup speed significantly (as long as a single disk can deliver the required performance - it will be active for the duration of the "backup"). The standard Linux USB mass storage module in 2.6.0 works well with it. The device has a solid aluminium case with a small fan to blow air through it. When operating the entire case gets warm so heat is dissipated from all sides (the fan seems to small to actually blow heat out of it, but I think it does some good in helping to move heat from the hard disk to the case). The device I tested is made by "New Motion" and appears to have model number N1106. It seems to be a new model as I couldn't find anyone selling it through a quick google search, however several older New Motion products are listed. -- http://www.coker.com.au/selinux/ My NSA Security Enhanced Linux packages http://www.coker.com.au/bonnie++/ Bonnie++ hard drive benchmark http://www.coker.com.au/postal/Postal SMTP/POP benchmark http://www.coker.com.au/~russell/ My home page -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]