Mass-Storage Systems

Mass-Storage Systems
 Overview of Mass Storage Structure
 HDD Scheduling
 NVM Scheduling
 Error Detection and Correction
 Storage Device Management
 Swap-Space Management
 Storage Attachment
 RAID Structure
Objectives
 Describe the physical structure of secondary storage
devices and the effect of a device’s structure on its uses
 Explain the performance characteristics of mass-storage
devices
 Evaluate I/O scheduling algorithms
 Discuss operating-system services provided for mass
storage, including RAID
Overview of Mass Storage Structure
 Bulk of secondary storage for modern computers is hard
disk drives (HDDs) and nonvolatile memory (NVM)
devices
 HDDs spin platters of magnetically-coated material under
moving read-write heads
 Drives rotate at 60 to 250 times per second
 Transfer rate is rate at which data flow between drive and
computer
 Positioning time (random-access time) is time to move disk
arm to desired cylinder (seek time) and time for desired sector
to rotate under the disk head (rotational latency)
 Head crash results from disk head making contact with the disk
surface -- That’s bad
 Disks can be removable
Moving-head Disk Mechanism
Hard Disk Drives
 Platters range from .85” to 14”
(historically)
 Commonly 3.5”, 2.5”, and 1.8”
 Range from 30GB to 3TB per drive
 Performance
 Transfer Rate – theoretical – 6
Gb/sec
 Effective Transfer Rate – real –
1Gb/sec
 Seek time from 3ms to 12ms –
9ms common for desktop drives
 Average seek time measured or
calculated based on 1/3 of tracks
 Latency based on spindle speed
 1 / (RPM / 60) = 60 / RPM
 Average latency = ½ latency
Hard Disk Performance
 Access Latency = Average access time = average seek
time + average latency
 For fastest disk 3ms + 2ms = 5ms
 For slow disk 9ms + 5.56ms = 14.56ms
 Average I/O time = average access time + (amount to
transfer / transfer rate) + controller overhead
 For example to transfer a 4KB block on a 7200 RPM disk with
a 5ms average seek time, 1Gb/sec transfer rate with a .1ms
controller overhead =
 5ms + 4.17ms + 0.1ms + transfer time =
 Transfer time = 4KB / 1Gb/s * 8Gb / GB * 1GB / 1024 2KB = 32 /
(10242) = 0.031 ms
 Average I/O time for 4KB block = 9.27ms + .031ms = 9.301ms
The First Commercial Disk Drive

1956
IBM RAMDAC computer
included the IBM Model 350
disk storage system

5M (7 bit) characters
50 x 24” platters
Access time = < 1 second
Nonvolatile Memory Devices
 If disk-drive like, then called solid-state disks (SSDs)
 Other forms include USB drives (thumb drive, flash drive),
DRAM disk replacements, surface-mounted on motherboards,
and main storage in devices like smartphones
 Can be more reliable than HDDs
 More expensive per MB
 Maybe have shorter life span – need careful management
 Less capacity
 But much faster
 Busses can be too slow -> connect directly to PCI for example
 No moving parts, so no seek time or rotational latency
Nonvolatile Memory Devices
 Have characteristics that present
challenges
 Read and written in “page”
increments (think sector) but
can’t overwrite in place
 Must first be erased, and erases
happen in larger ”block”
increments
 Can only be erased a limited
number of times before worn out –
~ 100,000
 Life span measured in drive
writes per day (DWPD)
 A 1TB NAND drive with rating of
5DWPD is expected to have 5TB
per day written within warrantee
period without failing
NAND Flash Controller Algorithms
 With no overwrite, pages end up with mix of valid and invalid
data
 To track which logical blocks are valid, controller maintains
flash translation layer (FTL) table
 Also implements garbage collection to free invalid page
space
 Allocates overprovisioning to provide working space for GC
 Each cell has lifespan, so wear leveling needed to write
equally to all cells

NAND block with valid and invalid pages

Volatile Memory
 DRAM frequently used as mass-storage device
 Not technically secondary storage because volatile, but can have file
systems, be used like very fast secondary storage
 RAM drives (with many names, including RAM disks) present as
raw block devices, commonly file system formatted
 Computers have buffering, caching via RAM, so why RAM drives?
 Caches / buffers allocated / managed by programmer, operating
system, hardware
 RAM drives under user control
 Found in all major operating systems
 Linux /dev/ram, macOS diskutil to create them, Linux /tmp of file system
type tmpfs
 Used as high speed temporary storage
 Programs could share bulk date, quickly, by reading/writing to RAM
drive
Magnetic Tape
Disk Attachment
 Host-attached storage accessed through I/O ports talking to I/O
busses
 Several busses available, including advanced technology
attachment (ATA), serial ATA (SATA), eSATA, serial attached
SCSI (SAS), universal serial bus (USB), and fibre channel (FC).
 Most common is SATA
 Because NVM much faster than HDD, new fast interface for NVM
called NVM express (NVMe), connecting directly to PCI bus
 Data transfers on a bus carried out by special electronic processors
called controllers (or host-bus adapters, HBAs)
 Host controller on the computer end of the bus, device controller on
device end
 Computer places command on host controller, using memory-mapped I/O
ports
 Host controller sends messages to device controller
 Data transferred via DMA between device and computer DRAM
Address Mapping
 Disk drives are addressed as large 1-dimensional arrays of
logical blocks, where the logical block is the smallest unit of
transfer
 Low-level formatting creates logical blocks on physical media
 The 1-dimensional array of logical blocks is mapped into the
sectors of the disk sequentially
 Sector 0 is the first sector of the first track on the outermost
cylinder
 Mapping proceeds in order through that track, then the rest of the
tracks in that cylinder, and then through the rest of the cylinders
from outermost to innermost
 Logical to physical address should be easy
 Except for bad sectors
 Non-constant # of sectors per track via constant angular velocity
HDD Scheduling
 The operating system is responsible for using hardware
efficiently — for the disk drives, this means having a fast
access time and disk bandwidth
 Minimize seek time
 Seek time  seek distance
 Disk bandwidth is the total number of bytes transferred,
divided by the total time between the first request for service
and the completion of the last transfer
Disk Scheduling (Cont.)
 There are many sources of disk I/O request
 OS
 System processes
 Users processes
 I/O request includes input or output mode, disk address, memory
address, number of sectors to transfer
 OS maintains queue of requests, per disk or device
 Idle disk can immediately work on I/O request, busy disk means
work must queue
 Optimization algorithms only make sense when a queue exists
 In the past, operating system responsible for queue management,
disk drive head scheduling
 Now, built into the storage devices, controllers
 Just provide LBAs, handle sorting of requests
 Some of the algorithms they use described next
Disk Scheduling (Cont.)
 Note that drive controllers have small buffers and can
manage a queue of I/O requests (of varying “depth”)
 Several algorithms exist to schedule the servicing of disk I/O
requests
 The analysis is true for one or many platters
 We illustrate scheduling algorithms with a request queue (0-
199)

98, 183, 37, 122, 14, 124, 65, 67

Head pointer 53
FCFS
Illustration shows total head movement of 640 cylinders
SCAN
 The disk arm starts at one end of the disk, and moves toward
the other end, servicing requests until it gets to the other end
of the disk, where the head movement is reversed and
servicing continues.
 SCAN algorithm Sometimes called the elevator algorithm
 Illustration shows total head movement of 208 cylinders
 But note that if requests are uniformly dense, largest density
at other end of disk and those wait the longest
SCAN (Cont.)
C-SCAN
 Provides a more uniform wait time than SCAN
 The head moves from one end of the disk to the other,
servicing requests as it goes
 When it reaches the other end, however, it immediately returns to
the beginning of the disk, without servicing any requests on the
return trip
 Treats the cylinders as a circular list that wraps around from
the last cylinder to the first one
 Total number of cylinders?
C-SCAN (Cont.)
Selecting a Disk-Scheduling Algorithm`
 SSTF is common and has a natural appeal
 SCAN and C-SCAN perform better for systems that place a heavy load on the disk
 Less starvation, but still possible
 To avoid starvation Linux implements deadline scheduler
 Maintains separate read and write queues, gives read priority
 Because processes more likely to block on read than write
 Implements four queues: 2 x read and 2 x write
 1 read and 1 write queue sorted in LBA order, essentially implementing C-
SCAN
 1 read and 1 write queue sorted in FCFS order
 All I/O requests sent in batch sorted in that queue’s order
 After each batch, checks if any requests in FCFS older than configured age
(default 500ms)
 If so, LBA queue containing that request is selected for next batch of I/O
 In RHEL 7 also NOOP and completely fair queueing scheduler (CFQ) also
available, defaults vary by storage device
NVM Scheduling
 No disk heads or rotational latency but still room for
optimization
 In RHEL 7 NOOP (no scheduling) is used but adjacent LBA
requests are combined
 NVM best at random I/O, HDD at sequential
 Throughput can be similar
 Input/Output operations per second (IOPS) much higher with
NVM (hundreds of thousands vs hundreds)
 But write amplification (one write, causing garbage collection
and many read/writes) can decrease the performance advantage
 Error Detection and Correction
 Fundamental aspect of many parts of computing (memory,
networking, storage)
 Error detection determines if there a problem has occurred
(for example a bit flipping)
 If detected, can halt the operation
 Detection frequently done via parity bit
 Parity one form of checksum – uses modular arithmetic to
compute, store, compare values of fixed-length words
 Another error-detection method common in networking is cyclic
redundancy check (CRC) which uses hash function to detect
multiple-bit errors
 Error-correction code (ECC) not only detects, but can
correct some errors
 Soft errors correctable, hard errors detected but not corrected
Storage Device Management
 Low-level formatting, or physical formatting — Dividing
a disk into sectors that the disk controller can read and write
 Each sector can hold header information, plus data, plus error
correction code (ECC)
 Usually 512 bytes of data but can be selectable
 To use a disk to hold files, the operating system still needs to
record its own data structures on the disk
 Partition the disk into one or more groups of cylinders, each
treated as a logical disk
 Logical formatting or “making a file system”
 To increase efficiency most file systems group blocks into
clusters
 Disk I/O done in blocks
 File I/O done in clusters
Storage Device Management (cont.)
 Root partition contains the OS, other partitions can hold
other Oses, other file systems, or be raw
 Mounted at boot time
 Other partitions can mount automatically or manually
 At mount time, file system consistency checked
 Is all metadata correct?
 If not, fix it, try again
 If yes, add to mount table, allow access
 Boot block can point to boot volume or boot loader set of
blocks that contain enough code to know how to load the
kernel from the file system
 Or a boot management program for multi-os booting
Device Storage Management (Cont.)

 Raw disk access for apps that

want to do their own block
management, keep OS out of
the way (databases for
example)
 Boot block initializes system
 The bootstrap is stored in
ROM, firmware
 Bootstrap loader program
stored in boot blocks of boot
partition
Booting from secondary
 Methods such as sector storage in Windows
sparing used to handle bad
blocks
Swap-Space Management
 Used for moving entire processes (swapping), or pages
(paging), from DRAM to secondary storage when DRAM not
large enough for all processes
 Operating system provides swap space management
 Secondary storage slower than DRAM, so important to optimize
performance
 Usually multiple swap spaces possible – decreasing I/O load on
any given device
 Best to have dedicated devices
 Can be in raw partition or a file within a file system (for
convenience of adding)
 Data structures for swapping on Linux systems:
Storage Attachment
 Computers access storage in three ways
 host-attached
 network-attached
 cloud
 Host attached access through local I/O ports, using one of
several technologies
 To attach many devices, use storage busses such as USB, firewire,
thunderbolt
 High-end systems use fibre channel (FC)
 High-speed serial architecture using fibre or copper cables
 Multiple hosts and storage devices can connect to the FC fabric
Network-Attached Storage
 Network-attached storage (NAS) is storage made available
over a network rather than over a local connection (such as a
bus)
 Remotely attaching to file systems
 NFS and CIFS are common protocols
 Implemented via remote procedure calls (RPCs) between host
and storage over typically TCP or UDP on IP network
 iSCSI protocol uses IP network to carry the SCSI protocol
 Remotely attaching to devices (blocks)
Cloud Storage
 Similar to NAS, provides access to storage across a
network
 Unlike NAS, accessed over the Internet or a WAN to remote
data center
 NAS presented as just another file system, while cloud
storage is API based, with programs using the APIs to
provide access
 Examples include Dropbox, Amazon S3, Microsoft OneDrive,
Apple iCloud
 Use APIs because of latency and failure scenarios (NAS
protocols wouldn’t work well)
Storage Array
 Can just attach disks, or arrays of disks
 Avoids the NAS drawback of using network bandwidth
 Storage Array has controller(s), provides features to attached
host(s)
 Ports to connect hosts to array
 Memory, controlling software (sometimes NVRAM, etc)
 A few to thousands of disks
 RAID, hot spares, hot swap (discussed later)
 Shared storage -> more efficiency
 Features found in some file systems
 Snaphots, clones, thin provisioning, replication, deduplication, etc
Storage Area Network
 Common in large storage environments
 Multiple hosts attached to multiple storage arrays – flexible
Storage Area Network (Cont.)
 SAN is one or more storage
arrays
 Connected to one or more
Fibre Channel switches or
InfiniBand (IB) network
 Hosts also attach to the
switches
 Storage made available via
LUN Masking from specific
arrays to specific servers
 Easy to add or remove
storage, add new host and
allocate it storage A Storage Array
 Why have separate storage
networks and communications
networks?
 Consider iSCSI, FCOE
RAID Structure
 RAID – redundant array of inexpensive disks
 multiple disk drives provides reliability via redundancy
 Increases the mean time to failure
 Mean time to repair – exposure time when another failure
could cause data loss
 Mean time to data loss based on above factors
 If mirrored disks fail independently, consider disk with 1300,000
mean time to failure and 10 hour mean time to repair
 Mean time to data loss is 100, 000 2 / (2 ∗ 10) = 500 ∗ 106 hours, or
57,000 years!
 Frequently combined with NVRAM to improve write
performance
 Several improvements in disk-use techniques involve the use of
multiple disks working cooperatively
RAID (Cont.)
 Disk striping uses a group of disks as one storage unit
 RAID is arranged into six different levels
 RAID schemes improve performance and improve the
reliability of the storage system by storing redundant
data
 Mirroring or shadowing (RAID 1) keeps duplicate of each
disk
 Striped mirrors (RAID 1+0) or mirrored stripes (RAID 0+1)
provides high performance and high reliability
 Block interleaved parity (RAID 4, 5, 6) uses much less
redundancy
 RAID within a storage array can still fail if the array fails,
so automatic replication of the data between arrays is
common
 Frequently, a small number of hot-spare disks are left
unallocated, automatically replacing a failed disk and
having data rebuilt onto them
RAID Levels
RAID (0 + 1) and (1 + 0)
Other Features
 Regardless of where RAID implemented, other useful features
can be added
 Snapshot is a view of file system before a set of changes
take place (i.e. at a point in time)
 More in Ch 12
 Replication is automatic duplication of writes between
separate sites
 For redundancy and disaster recovery
 Can be synchronous or asynchronous
 Hot spare disk is unused, automatically used by RAID
production if a disk fails to replace the failed disk and rebuild
the RAID set if possible
 Decreases mean time to repair
Extensions
 RAID alone does not prevent or
detect data corruption or other
errors, just disk failures
 Solaris ZFS adds checksums of all
data and metadata
 Checksums kept with pointer to
object, to detect if object is the right
one and whether it changed
 Can detect and correct data and
metadata corruption
 ZFS also removes volumes,
partitions
 Disks allocated in pools ZFS checksums all

metadata and data
Filesystems with a pool share that
pool, use and release space like
malloc() and free() memory
allocate / release calls
Traditional and Pooled Storage
Object Storage
 General-purpose computing, file systems not sufficient
for very large scale
 Another approach – start with a storage pool and place
objects in it
 Object just a container of data
 No way to navigate the pool to find objects (no directory
structures, few services
 Computer-oriented, not user-oriented
 Typical sequence
 Create an object within the pool, receive an object ID
 Access object via that ID
 Delete object via that ID
Object Storage (Cont.)
 Object storage management software like Hadoop file
system (HDFS) and Ceph determine where to store objects,
manages protection
 Typically by storing N copies, across N systems, in the object
storage cluster
 Horizontally scalable
 Content addressable, unstructured