Disk Partitioning Continued
Disk device files Ways to divide up disks Traditional partitioning MBR partitioningGPT partitioning
-
MBR partitioning scheme is being superseded by GPT: GUID (Globally Unique Identifiers)
Partition Table - part of the UEFI standard ( Unified Extensible Firmware Interface - new/supported by others for Intel's EFI) - new firmware standard that replaced PC BIOS
GPT:
- one type of partition (no logical partitions inside of the extended partition)
- as many partitions as needed
- no 2TB limit - can be up to 2^64 blocks in length
- Each partition has 16-byte GUID - Globally Unique ID
- backward compatible :1st sector for MBR that fakes one large MBR partition (to prevent naïve systems from attempting to reformat the disk)
Linux partitioning
-
parted (or gparted)
- understands several label formats (Solaris, ...)
- move / resize partitions
- "lets you reallocate space dynamically from the greedy partition to the needy partition"
- improveed / supercharged / abstracted version of disk partitioning:
- individual storage devices (physical volumes in Linux LVM - may be partition or RAID array) grouped into
- volume groups - split into / blocks in a volume group can then be allocated to
- logical volumes - represented by block device files and used as the disk partitions previously - hold filesystems
- LVM allows to :
- Move logical volumes among different physical devices
- Grow and shrink logical volumes on the fly
- Take copy-on-write “snapshots” of logical volumes
- Replace on-line drives without interrupting service
- Incorporate mirroring or striping in your logical volumes
- Linux LVM:
lvm command plus many subcommands
- LVM configuration:
- Creating (defining) and initializing physical volumes
- Adding the physical volumes to a volume group
- Creating logical volumes on the volume group
- Example:
pvcreate /dev/sdc vgcreate DEMO /dev/sdc vgdisplay DEMO lvcreate -L 10G DEMO -n web1 mkfs.ext4 /dev/DEMO/web1 mkdir /extra/web1 mount /dev/DEMO/web1 /extra/web1
- FSs fill up often - perhaps more often then disks crash
- LVM doesn't know anything about the FS on its volumes => resize at both the volume and FS level
- Reduction: FS first, volume second
- Enlargement: volume first, FS second
- Example:
vgdisplay DEMO vi /extra/web1/index.html umount /extra/web1 lvdisplay DEMO/web1 lvchange -an DEMO/web1 lvdisplay DEMO/web1 lvresize -L +5G DEMO/web1 lvchange -ay DEMO/web1 lvdisplay DEMO/web1 vgdisplay DEMO resize2fs /dev/DEMO/web1 Please run 'e2fsck -f /dev/DEMO/web1' first.' e2fsck -f /dev/DEMO/web1 resize2fs /dev/DEMO/web1 mount /dev/DEMO/web1 /extra/web1 cat /extra/web1/index.html
RAID: Redundant Arrays of Inexpensive (Independent) Disks
- distributes or replicates data across multiple disks
- may avoid data loss due to disk error - mirroring and parity schemes
- may increase performance - parallel access
- software or hardware implementations
- RAID can do 2 basic things:
- improve performace by "striping" data across multiple drives
- decrease risks assoc with disk failure by replicating data across multiple drives
- mirroring - data blocks are reproduced bit-for-bit on several different drives
- parity schemes - 1+ drives with error-correcting checksum of the remaining data drives blocks
- RAID Levels - specifies array's parallelism and redundancy:
- Linear mode (not real RAID level) - just concatenates drives into large virtual drive - no data redundancy or performance benefit
- Level 0 increases performance - 2+ equal size disks
stripes data alternately among the disks
worse reliability (with 2 disks it's 2x more likely that one of them will fail)
- Level 1 / Mirroring - Writes dupplicated to 2+ disks
slightly slower writes, faster reads (close to RAID 0)
- Level 1+0 (Stripes of Mirrors) and 0+1 (Mirrors of Stripes)
Performance of RAID 0 and redundancy of RAID 1
- Level 5 stripes both data and parity info
Redundancy, read performance, disk efficiency
- Level 6 like RAID 5 but two parity disks
Withstands complete failure of two disks
- Disk failure recovery
-
Google disk failure study - 8% annual failure rate => expect 1 failure per month with just 150 disks...
RAID 1 , 10, 5, 6 - enter degraded mode , swap disk asap, and restore redundancy (takes hours)
- Drawbacks of RAID 5
- doesn't replace backups / doesn't protect from accidental deletion of files, controller failures, fires, hackers, ...
- RAID 5 write performance for random block writes => 4 operations (2xR+2xW) to maintain parity info
- vulnerable to corruption ("write hole") - incremental updating of parity data (more efficient then reading entire stripe) BUT parity data not validated or recalculated - need to do "scrubbing" (validating parity blocks one by one while the array is relatively idle)
- mdadm : Linus Software RAID