Chapter 6 - External Memory

Luis Tarrataca
[email protected]

CEFET-RJ

L. Tarrataca Chapter 6 - External Memory 1 / 66

Table of Contents I

1 Motivation

2 Magnetic Disks

Write Mechanism

Read Mechanism

Contemporary Read Mechanism

Data Organization and Formatting

Components of a Disk Drive

Disk Performance Parameters

L. Tarrataca Chapter 6 - External Memory 2 / 66

Table of Contents I
3 Redundant Array of Independent Disks (RAID)

RAID 0

RAID 1

RAID 2

RAID 3

RAID 4

RAID 5

RAID 6

4 Solid State Drives (SSD)

SSD compared to HDD

Practical Issues
L. Tarrataca Chapter 6 - External Memory 3 / 66
 Motivation

Motivation

Computers have a set of internal memory mechanisms

• registers, RAM, etc

Computers also need to interact with peripherals

• Some of these are storage devices;

• A.k.a external memory.

L. Tarrataca Chapter 6 - External Memory 4 / 66

 Motivation

How does such an external memory work?

How is such an external memory organized?

L. Tarrataca Chapter 6 - External Memory 5 / 66

 Motivation

Some examples of external memory include:

• Magnetic disks;

• Redundant Array of Independent Disks (RAID);

• Solid State Drives (SSD);

• Optical Memory;

• Magnetic tapes.

Lets take a closer look at some of these.

L. Tarrataca Chapter 6 - External Memory 6 / 66

 Magnetic Disks

Magnetic Disks

Circular platter coated with a magnetizable material.

Figure: Interior of a magnetic hard drive

Data are recorded on and later retrieved from the disk via a head:

• Most common design: a read head and a write head;

• During a read or write operation:

• Head is stationary while the platter rotates beneath it.
L. Tarrataca Chapter 6 - External Memory 7 / 66
 Magnetic Disks Write Mechanism

Write Mechanism

Figure: Inductive Write / Magnetoresistive Read Head. (Source: [Stallings, 2015])

Write Mechanism:

• Electric pulses are sent to the write head;

• Resulting magnetic patterns are recorded on the surface below:

L. Tarrataca Chapter 6 - External Memory 8 / 66
 Magnetic Disks Read Mechanism

Read Mechanism

Figure: Inductive Write / Magnetoresistive Read Head. (Source: [Stallings, 2015])

Read Mechanism:

• Disk surface passes under the read head;

• Generating a current of the same polarity as the one recorded.

L. Tarrataca Chapter 6 - External Memory 9 / 66
 Magnetic Disks Contemporary Read Mechanism

Contemporary Read Mechanism

Figure: Inductive Write / Magnetoresistive Read Head. (Source: [Stallings, 2015])

Contemporary disk systems use a separate read head:

• Head consists of a magnetoresistive (MR) sensor;

• Resistance of MR material depends on the direction of the magnetization;
• Sensor detects resistance changes as voltage signals.

L. Tarrataca Chapter 6 - External Memory 10 / 66

 Magnetic Disks Data Organization and Formatting

Data Organization and Formatting

Based on the read / write head mechanism how are disks organized?
Any ideas?

L. Tarrataca Chapter 6 - External Memory 11 / 66

 Magnetic Disks Data Organization and Formatting

Data Organization and Formatting

Based on the read / write head mechanism how are disks organized?
Any ideas?

Impractical to have a large tape moving under the head:

• Old magnetic tapes;

• Substantial seek time times;

L. Tarrataca Chapter 6 - External Memory 12 / 66

 Magnetic Disks Data Organization and Formatting

Circular Information Storing:

Figure: Disk Data Layout. (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 13 / 66

 Magnetic Disks Data Organization and Formatting

Each platter has a concentric set of rings, called tracks:

• Each track is the same width as the head.

• There are thousands of tracks per surface.

• Adjacent tracks are separated by gaps in order to prevent;

• misalignment of the head;

• magnetic field interference.

L. Tarrataca Chapter 6 - External Memory 14 / 66

 Magnetic Disks Data Organization and Formatting

Data are transferred to and from the disk in sectors:

• There are typically hundreds of sectors per track;

• These may be of either fixed or variable length

• Nowadays 512 bytes is the universal sector size.

• Adjacent sectors are separated by intersector gaps.

L. Tarrataca Chapter 6 - External Memory 15 / 66

 Magnetic Disks Data Organization and Formatting

Now that we know about the existence of tracks and sectors:

How can the head find these elements within the disk?

L. Tarrataca Chapter 6 - External Memory 16 / 66

 Magnetic Disks Data Organization and Formatting

How can the head find these elements within the disk?

Head needs to locate sector positions within a track, requiring knowing:

• Starting point on the track;

• Start and end of each sector;

L. Tarrataca Chapter 6 - External Memory 17 / 66

 Magnetic Disks Data Organization and Formatting

How can the head find these elements within the disk?

Head needs to locate sector positions within a track, requiring knowing:

• Starting point on the track;

• Start and end of each sector;

How can this be performed? Any ideas?

L. Tarrataca Chapter 6 - External Memory 18 / 66

 Magnetic Disks Data Organization and Formatting

How can this be performed? Any ideas?

• Location data needs to be recorded on the disk:

• Disk is formatted with extra data;

• This data is used only by the drive

• Not accessible to the user.

• Reason why:
• Space seen as available by OS 6= to physical disk space

L. Tarrataca Chapter 6 - External Memory 19 / 66

 Magnetic Disks Data Organization and Formatting

Example (1/3)

Lets look at an example:

Figure: Winchester Disk Format. (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 20 / 66

 Magnetic Disks Data Organization and Formatting

Example (2/3)

Figure: Winchester Disk Format. (Source: [Stallings, 2015])

Each track contains 30 fixed-length sectors of 600 bytes each.

• Each sector holds 515 bytes of data plus other control information;

• This means that only 515/600 ≈ 85% is available for data...

L. Tarrataca Chapter 6 - External Memory 21 / 66

 Magnetic Disks Data Organization and Formatting

Example (3/3)
Lets look at a specific example:

Figure: Winchester Disk Format. (Source: [Stallings, 2015])

The ID field uniquely identifies a sector containing:

• SYNCH byte is a special bit pattern that delimits the beginning of the field;

• Track number

• Head / Surface number for disks with multiple surfaces;

• L. Tarrataca Chapter 6 - External Memory 22 / 66

 Magnetic Disks Components of a Disk Drive

Components of a Disk Drive

One way of organizing the components of a disk drive:

Figure: Interior of a magnetic hard drive

Figure: Components Of A Disk Drive. (Source:

[Stallings, 2015])
L. Tarrataca Chapter 6 - External Memory 23 / 66
 Magnetic Disks Disk Performance Parameters

Disk Performance Parameters

So, what do you think are some variables that influence the performance
of a disk?

L. Tarrataca Chapter 6 - External Memory 24 / 66

 Magnetic Disks Disk Performance Parameters

Disk Performance Parameters (1/4)

• Seek time - the time it takes to position the head at the track;

• Rotational delay - once the track is selected:

• The sector still needs to line up with the head;

• E.g.: an hard disk rotating at 20000 rpm:

• 20000 rpm → one revolution per 3 ms;
• On average we will have to wait for half the plate to spin;
• rotational delay = 1.5 ms;

• Access time = SeekTime + Rotational Delay

L. Tarrataca Chapter 6 - External Memory 25 / 66

 Magnetic Disks Disk Performance Parameters

Disk Performance Parameters (2/4)

• Transfer time - Data transfer portion of the operation;

b
T =
rN

• b - number of bytes to transfer;

• r - rotation speed per second (rps)

• N - number of bytes on a track;

L. Tarrataca Chapter 6 - External Memory 26 / 66

 Magnetic Disks Disk Performance Parameters

Disk Performance Parameters (3/4)

• Total average time:

T = Tseek + Trotational delay + Ttransfer time

1 b
T = Tseek + +
2r rN

L. Tarrataca Chapter 6 - External Memory 27 / 66

 Magnetic Disks Disk Performance Parameters

Disk Performance Parameters (4/4)

There are also delays associated with I/O operation:

Figure: Timing of a disk I/O transfer (Source: [Stallings, 2015])

1 Wait for device:

• A process must wait in a queue for a device to be available;

2 Eventually, the device is assigned to the process;

3 Wait for I/O channel:

• If the device shares the I/O channel it must wait for it to be available;

4 Only after this point do we proceed with the head seek.

L. Tarrataca Chapter 6 - External Memory 28 / 66
 Magnetic Disks Disk Performance Parameters

Figure: Typical Hard Disk Driver Parameters (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 29 / 66

 Redundant Array of Independent Disks (RAID)

Redundant Array of Independent Disks (RAID)

Now that we know about how hards disks works:

How can we deal with hard disk failure? Any ideas?

L. Tarrataca Chapter 6 - External Memory 30 / 66

 Redundant Array of Independent Disks (RAID)

Redundant Array of Independent Disks (RAID)

Now that we know about how hards disks works:

How can we deal with hard disk failure? Any ideas?

Redundancy is very important in computation:

• If a disk dies we do not want to lose everything;

• RAID is a form of redundancy used to deal with hard disk failure.

L. Tarrataca Chapter 6 - External Memory 31 / 66

 Redundant Array of Independent Disks (RAID)

RAID has seven design architecture levels sharing these characteristics:

1 RAID is a set of physical disk drives:

• viewed by the operating system as a single logical drive;

2 Data are distributed across the physical drives of an array:

• this scheme known as striping;

3 Redundant disk capacity is used to store parity information

• guaranteeing data recoverability in case of a disk failure. (RAID ≥ 2)

L. Tarrataca Chapter 6 - External Memory 32 / 66

 Redundant Array of Independent Disks (RAID)

RAID strategy employs multiple disk drives:

• Data are distributed:

• enable simultaneous access from multiple drives;

• separate I/O requests can be handled in parallel;

• thereby improving I/O performance;

• Same concept of parallelism used throughout computation

• Performance bottleneck? Parallelise the components...

L. Tarrataca Chapter 6 - External Memory 33 / 66

 Redundant Array of Independent Disks (RAID)

Lets have a look at the different RAID levels....

L. Tarrataca Chapter 6 - External Memory 34 / 66

 Redundant Array of Independent Disks (RAID) RAID 0

RAID 0

Good for high I/O request rate:

• Disk array provides high I/O execution rates:

• This is done by balancing the load across multiple disks.

RAID 0 has no redundancy:

• Therefore, not a true member of the RAID family....

L. Tarrataca Chapter 6 - External Memory 35 / 66

 Redundant Array of Independent Disks (RAID) RAID 0

Data are striped across the available disks:

Figure: Data Mapping for a RAID Level 0 Array (Nonredundant). (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 36 / 66

 Redundant Array of Independent Disks (RAID) RAID 0

Data are striped across the available disks:

Figure: Data Mapping for a RAID Level 0 Array (Nonredundant). (Source: [Stallings, 2015])

Data are viewed as being stored on a logical disk:

• The logical disk is divided into strips:

• these strips may be physical blocks, sectors, or some other unit

• Strips are mapped round robin to consecutive disks in the array.

L. Tarrataca Chapter 6 - External Memory 37 / 66
 Redundant Array of Independent Disks (RAID) RAID 0

RAID 0 advantage:

• A single I/O request consists of multiple logically contiguous strips:

• then up to n strips for that request can be handled in parallel;

• greatly reducing the I/O transfer time.

• Recall that total read / write time is a function of seek and rotational delay:
• By distributing across multiple disks we are diminishing these times;

• Instead of having multiple seek and rotational delays for a single hard disk.

L. Tarrataca Chapter 6 - External Memory 38 / 66

 Redundant Array of Independent Disks (RAID) RAID 1

RAID 1

Very simple:

• redundancy is achieved by duplicating all the data.

Figure: Data Mapping for a RAID Level 1 Array (Mirrored). (Source: [Stallings, 2015])

• Read requests can be serviced by the lowest access-time disk;

• A write request requires updating in parallel both strips:

• the write performance is dictated by the slower of the two writes

L. Tarrataca Chapter 6 - External Memory 39 / 66

 Redundant Array of Independent Disks (RAID) RAID 1

• Recovery from a failure is simple:

• When a drive fails, the data may still be accessed from the second drive.

• Disadvantage:
• Principal disadvantage of RAID 1 is the cost:
• Requires twice the disk space of the logical disk that it supports.

• Advantage:
• Achieves high I/O request rates if the bulk of the requests are reads.
• Performance of RAID 1 can approach double of that of RAID 0.

L. Tarrataca Chapter 6 - External Memory 40 / 66

 Redundant Array of Independent Disks (RAID) RAID 2

RAID 2

• Makes use of a parallel access technique:

• All member disks participate in the execution of every I/O request;

• Spindles of the individual drives are synchronized:

• Each disk head is in the same position on each disk at any given time.

Figure: Data Mapping for a RAID Level 2 Array (Hamming code). (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 41 / 66

 Redundant Array of Independent Disks (RAID) RAID 2

Figure: Data Mapping for a RAID Level 2 Array (Hamming code). (Source: [Stallings, 2015])

• Typically: Hamming code error-correction is used:

• Stripping occurs at the bit-level;

• Code bits are stored in the corresponding bit positions on multiple parity
disks;

• Rarely used in practice!

• Currently all hard disk drives implement internal error correction:

• Complexity of an external Hamming code offers little advantage...

L. Tarrataca Chapter 6 - External Memory 42 / 66

 Redundant Array of Independent Disks (RAID) RAID 3

RAID 3

Similar fashion to RAID 2:

• Difference:
• requires only a single redundant disk, no matter how large the disk array.

Figure: Data Mapping for a RAID Level 3 Array (Bit-interleaved parity). (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 43 / 66

 Redundant Array of Independent Disks (RAID) RAID 3

• Parity bit is computed for the set of individual bits in the same position on
all of the data disks.

X4 (i ) = X0 (i ) ⊕ X1 (i ) ⊕ X2 (i ) ⊕ X3 (i )

• X0 , X1 , X2 , X3 represent the data hard disks;

• X4 is the redundancy hard disk

• i represents the i-th bit.

L. Tarrataca Chapter 6 - External Memory 44 / 66

 Redundant Array of Independent Disks (RAID) RAID 3

• Suppose that drive X1 has failed. If we add X4 (i ) ⊕ X1 (i ) to both sides of

the preceding equation, we get

X1 (i ) = X0 (i ) ⊕ X2 (i ) ⊕ X3 (i ) ⊕ X4 (i )

• Now we just need to process each individual bit i.

L. Tarrataca Chapter 6 - External Memory 45 / 66

 Redundant Array of Independent Disks (RAID) RAID 4

RAID 4

• Makes use of an independent access technique:

• each member disk operates independently;

• separate I/O requests can be satisfied in parallel.

• More suitable for applications that have high I/O request rates;

• Less suitable for applications that require high transfer rates.

Figure: Data Mapping for a RAID Level 4 Array (Block-level parity). (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 46 / 66

 Redundant Array of Independent Disks (RAID) RAID 4

Bit-by-bit parity calculated across corresponding strips on each disk:

• Parity bits are stored in the corresponding strip on the parity disk:

X4 (i ) = X0 (i ) ⊕ X1 (i ) ⊕ X2 (i ) ⊕ X3 (i )

• X0 , X1 , X2 , X3 represent the data hard disks;

• X4 is the redundancy hard disk

• i represents the i-th bit.

L. Tarrataca Chapter 6 - External Memory 47 / 66

 Redundant Array of Independent Disks (RAID) RAID 5

RAID 5

• Similar to RAID 4
• Difference: RAID 5 distributes the parity strips across all disks.

Figure: Data Mapping for a RAID Level 5 Array. (Source: [Stallings, 2015])

• Distribution of parity across all drives avoids potential I/O bottle-neck:

• Instead of having a single hard disk with the parity information...
• Concurrent access to the parity bits are multiplexed;

L. Tarrataca Chapter 6 - External Memory 48 / 66

 Redundant Array of Independent Disks (RAID) RAID 6

RAID 6

• Two different parity calculations are carried out and stored in separate
blocks on different disks.
• Idea: Regenerate data even if two disks containing user data fail.

Figure: Data Mapping for a RAID Level 6 Array (Dual redundancy). (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 49 / 66

 Redundant Array of Independent Disks (RAID) RAID 6

A brief summary of all the info regarding RAID levels:

Figure: RAID levels (Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 50 / 66

 Solid State Drives (SSD)

Solid State Drives (SSD)

• SSD is made with solid state components:

• Use of flash memory (instead of magnetic disk option)

• In recent years the cost and performance of flash memory has evolved.

• Increased use of solid state drives (SSDs) to complement or even replace

hard disk drives (HDDs);
• Dedicated SSD;

• SSD + HD;

• Hybrid HD ( SSD + HD ).

• Significant development in external memory;

L. Tarrataca Chapter 6 - External Memory 51 / 66

 Solid State Drives (SSD) SSD compared to HDD

What do you think are the advantages of using an SSD instead of a HDD?

L. Tarrataca Chapter 6 - External Memory 52 / 66

 Solid State Drives (SSD) SSD compared to HDD

SSD compared to HDD

SSDs have the following advantages over HDDs (1/2):

• Lower access times and latency rates. Why?

• High-performance input/output operations per second. Why?

• Durability. Why?

L. Tarrataca Chapter 6 - External Memory 53 / 66

 Solid State Drives (SSD) SSD compared to HDD

SSD compared to HDD

SSDs have the following advantages over HDDs (1/2):

• Lower access times and latency rates. Why?

• Due to having no mechanical parts;

• High-performance input/output operations per second. Why?

• Significantly increases performance I/O subsystems.

• Again due to having no mechanical parts;

• Durability. Why?
• Less susceptible to physical shock and vibration;

• Again due to having no mechanical parts;

L. Tarrataca Chapter 6 - External Memory 54 / 66

 Solid State Drives (SSD) SSD compared to HDD

SSD compared to HDD

SSDs have the following advantages over HDDs (2/2):

• Longer lifespan. Why?

• Lower power consumption. Why?

• Quieter and cooler running capabilities. Why?

L. Tarrataca Chapter 6 - External Memory 55 / 66

 Solid State Drives (SSD) SSD compared to HDD

SSD compared to HDD

SSDs have the following advantages over HDDs (2/2):

• Longer lifespan. Why?

• SSDs are not susceptible to mechanical wear.

• Lower power consumption. Why?

• SSDs use as little as 2.1 watts of power per drive.

• Quieter and cooler running capabilities. Why?

• Less floor space required, lower energy costs, and a greener enterprise.

L. Tarrataca Chapter 6 - External Memory 56 / 66

 Solid State Drives (SSD) SSD compared to HDD

What do you think are the disadvantages of using an SSD?

L. Tarrataca Chapter 6 - External Memory 57 / 66

 Solid State Drives (SSD) SSD compared to HDD

What do you think are the disadvantages of using an SSD?

• cost per bit is high;

• capacity is low;

• Not for long ;)

L. Tarrataca Chapter 6 - External Memory 58 / 66

 Solid State Drives (SSD) SSD compared to HDD

Figure: Comparison of Solid State Drives and Disk Drives. (Source: [Stallings, 2015])

Disadvantage of mechanical vs. digital in terms of speed.

L. Tarrataca Chapter 6 - External Memory 59 / 66

 Solid State Drives (SSD) SSD compared to HDD

SSD contains the following components:

• Controller: Device level interfacing and
firmware execution;
• Addressing: Logic to select the flash
memory components;
• Data buffer/cache: To further increase
performance;
• Error correction: Logic for error detection
and correction;
• Flash memory components: Individual
NAND flash chips.

Figure: Solid State Drive Architecture.

(Source: [Stallings, 2015])

L. Tarrataca Chapter 6 - External Memory 60 / 66

 Solid State Drives (SSD) Practical Issues

SSD Practical Issues

Performance has a tendency to slow down as the device is used (1/2):

• Flash memory is accessed in blocks containing memory pages;

• Consider what must be done to write a page onto a flash memory:

1 Block must be read from flash memory and placed in a RAM buffer;

2 Appropriate page in the RAM buffer is updated;

3 Before block can be written back:

• Entire block of flash memory must be erased;
• Not possible to erase just one page of the flash memory.

4 Entire block from the buffer is now written back to the flash memory.

L. Tarrataca Chapter 6 - External Memory 61 / 66

 Solid State Drives (SSD) Practical Issues

SSD Practical Issues

Performance has a tendency to slow down as the device is used (2/2):

• Over time files become fragmented:

• File pages are scattered over multiple blocks;

• File pages not contiguously in memory → loss of efficiency;

• Several techniques exist to compensate for this property of flash memory;

L. Tarrataca Chapter 6 - External Memory 62 / 66

 Solid State Drives (SSD) Practical Issues

• Worse than fragmentation issues:

• Flash memory becomes unusable after a certain number of writes.

• Depends on the quality of the flash memory (good estimate 105 writes)

• There are also a variety of techniques to handle the life of flash memory:
• A cache to delay and group write operations;
• Wear-leveling algorithms: evenly distribute writes across block of cells;
• Bad-block management techniques: when eventually blocks do kaputz!

L. Tarrataca Chapter 6 - External Memory 63 / 66

 Where to focus your study

Where to focus your study

After this class you should be able to:

• Understand the key properties of magnetic disks.

• Understand the performance issues involved in magnetic disk access.

• Explain the concept of RAID, importance of redundancy and mechanisms

for redundancy;

• Compare and contrast hard disk drives and solid disk drives.

• Describe in general terms the operation of flash memory.

L. Tarrataca Chapter 6 - External Memory 64 / 66

 Where to focus your study

Less important to know:

• details of all RAID levels;

• details of specific magnetic disk formats;

Your focus should always be on the building blocks for developing a solution
=)

L. Tarrataca Chapter 6 - External Memory 65 / 66

 References

References I

Stallings, W. (2015).

Computer Organization and Architecture.

Pearson Education.

L. Tarrataca Chapter 6 - External Memory 66 / 66