Chapter 1. And Then There Was Disk
First came to market in the late 50s and 60s, disk drives have relied on the following core components
- Read/write heads that use electrical impulses to store and retrieve magnetically recorded bits of data
- Magnetically coated disk platters that spin and house these bits
- Mechanical actuator arms that move the heads back and forth across the spinning disk platters, forming concentric ‘tracks’ of recorded data
The Past: Disk Drives Prior to 1985
The Early Days of Disk Drive Communications
- Bus-and-Tag Systems
- via two copper-wire cables: Bus and Tag, Bus for data, Tag for communication protocols
- SMD Disk Drives
- Control Data Corporation first shipped its minicomputer with storage module device (SMD) disk drives in late 1973
- Used much smaller ―A and B flat cables to transfer control instructions (from the A cable) and data (from the B cable)
- The disk controller shrank down to a single board which was inserted into the system‘s CPU card cage
Disk drives in the late 80s and 90s went through a number of significant transformations that allowed them to be widely used in the emerging open systems world of servers and personal computers. These included:
- 19” disk drives => less expensive 5.25” (and, eventually, 3.5”) drives.
- Advances associated with redundant arrays of independent disks (RAID) technology.
- The development and widespread adoption of the Small Computer Systems Interface (SCSI).
Disk drives produced today fall into four categories, depending on their cable connections
- SCSI
- Fibre Channel
- Serial ATA (SATA)
- Serial Attached SCSI (SAS)
SCSI used a single data cable to present its Common Command Set (CCS) interface with built-in intelligence.
Today: Disk Drives, Pork Bellies and Price Tags
Are Disk Drives in Our Future?
Today‘s latest battle cry is that solid state disks (SSDs) will completely replace magnetic disk storage
Research into the area of higher capacities for magnetic disk drives
- Perpendicular Magnetic Recording (PMR)
- 多層次的儲存，原本是平面，變成是 3D
- Patterned Media Recording
- Heat-Activated Magnetic Recording
- relies on first heating the media so that it can store smaller bits of data per square inch
- PMR appears to be winning the short-term race
- Nanostorage
Address protection in two separate chapters
- Protecting against disk drive failure
- users can continue to access data previously stored on failed disks
- Protecting against data loss or corruption
- moves more deeply into storage software intelligence
The Past: Protecting SLEDs
The Middle Ages: RAID in the 80s
1987 paper
- provide greater efficiency and faster I/O performance
- successfully survive a failure of any one disk drive
- describe five different RAID methods (RAID 1 through RAID 5)
|
RAID Technique
|
Description
|
|
No Parity
(RAID 0)
|
Ø increase I/O performance by striping (or logically distributing) data across
several disk drives.
Ø offered no protection against failed disk drives
|
|
Mirroring
(RAID 1)
|
Ø data is mirrored onto a second set of disks
Ø exacts a high capacity penalty
|
|
Fixed Parity
(RAID 3 / RAID 4)
|
Ø both use parity calculations (sometimes known as checksum) to
perform error-checking
Ø recovery of missing data from failed drives
|
|
Striped Parity
(RAID 5)
|
Ø parity is striped (or logically distributed) across all disks in
the RAID set in an attempt to boost RAID read/write performance
|
|
Multiple Parity
(RAID 6)
|
Ø using multiple iterations of fixed or striped parity on a group of
drives, which allows for multiple drive failures without data loss.
|
The Future: Smarter, Self-Healing Disk Drives
S.M.A.R.T. technology
Approaches to Data Loss or Corruption
|
Approach
|
Description
|
|
Data replication
|
Mirroring critical data to an alternate
location
|
|
Data backup
|
Restore data that may have been
accidentally deleted or earlier data version
|
The Past: The Tale of the Tape
Today’s Backup Tapes
Decades of “format wars” ensued amongst vendors fighting for market share. Sample formats from this era included:
- Quarter-Inch Cartridge (QIC)
- 4mm or 8mm Tape
- Digital Linear Tape (DLT)
- Advanced Intelligent Tape (AIT)
- Linear Tape Open (LTO)
LTO-4 has become the reigning tape format today.
The Middle Ages: Early Tape Backup Automation
|
Tape-Based Innovations:
|
|
|
Interleaving
|
Ø improved tape backup speeds by allowing backups to be written to
multiple tapes concurrently
Ø writing to several tapes in parallel
|
|
Synthetic Backup
|
Ø required just a single full backup and used an intermediate
database to track and map the location of the continuous incremental backups
performed to tape thereafter
|
|
Reclamation
|
Ø Also pioneered by Tivoli Storage Manager (TSM)
Ø the tape reclamation process solved a problem created by Synthetic
tape backups
|
|
Disk-Based Innovations
|
|
|
Disk Staging
|
Ø data stored on optical media that could be moved by a staging
manager to magnetic disk drives
Ø could be used to send the data to tape without affecting
production workloads
|
|
D2D2T
|
disk-to-disk-to-tape
|
|
D2D
|
disk-to-disk without tape
|
The Modern Age: Emerging D2D Efficiencies
A Snapshot is Not a Backup
NetApp explains this space-saving functionality as follows
- We are able to create a snapshot in constant time because we have a map of the blocks that are allocated on disk. A snapshot is really just a copy of the block map rather than the actual disk blocks
Use of local snapshots alone, however, still exposes the data to other corruption risks, such as
- Widespread data corruption of the primary data set
- Hardware failure impacting the data stored within
- The SnapVault “primary” system needing data protection.
- The SnapVault “secondary” system where backup data is stored.
- SnapVault initially stores one “full” backup of the primary system‟s data set on the secondary
- builds on NetApp Snapshot efficiencies by quickly transmitting only the changed blocks found in the most recent snapshot of the primary system
The Future of Data Protection
- The new gold standard: Annual off-site archival of data to tape
- Tape backup will become a service in lieu of local tape libraries.
- D2D will be managed by the storage array itself.
- Say goodbye to VTLs
沒有留言:
張貼留言