Systems architecture

(a) ALU — performs all arithmetic operations (such as addition and subtraction) and logical operations (such as AND, OR and comparisons) (1) (b) Control unit — decodes instructions and sends control signals that coordinate and control the activity of the CPU and the flow of data (1) (c) Clock — sends out a regular electrical pulse that synchronises the components of the CPU so they work in step (1) (d) Register — a very small, very fast piece of memory inside the CPU that temporarily holds data or instructions currently being worked on (1) (e) Bus — a collection of wires through which data/signals are transmitted from one component to another (1)

(5)

Fetch — the next instruction is fetched to the CPU from main memory (1).

Decode — the instruction is decoded to work out what it is, by the control unit (1).

Execute — the instruction is executed (carried out). This may include reading from or writing to main memory (1).

The cycle then repeats continually with the next instruction, for as long as the computer is running (1).

(4)

Difference 1 — volatility. RAM is volatile, so its contents are lost when the power is turned off, whereas ROM is non-volatile and keeps its contents without power (1).

Difference 2 — read and write. RAM can be read from and written to, whereas ROM is read only in normal use and its contents cannot be changed (1).

RAM is used to hold the instructions and data of programs currently running, along with the parts of the operating system in use (1).

ROM is used to hold the boot-up instructions that start the computer and load the operating system into RAM (1).

(4)

(a) Clock speed. A higher clock speed means more clock pulses per second (1). Each pulse drives the CPU forward, so more Fetch-Execute cycles are completed per second and more instructions are executed in a given time — improving performance (1).

(b) Number of cores. Each core is a complete processing unit with its own ALU, control unit and registers, so a multi-core CPU can process several instructions at the same time, in parallel (1). However, the software must be written to make use of multiple cores — a program that is not will use only one core, so extra cores bring no benefit (1).

(c) Cache size. Cache is small, very fast memory in or near the CPU that stores frequently used instructions and data (1). A larger cache means more requests can be met from fast cache rather than the slower main memory, so the CPU spends less time waiting for data and performance improves (1).

(6)

(a) Main memory is any form of memory that is directly accessible by the CPU — excluding cache and registers — and is volatile, so its contents are lost when the power is turned off (1). Secondary storage is any non-volatile storage mechanism that is not directly accessible by the CPU, so it keeps its contents without power (1). Because the CPU cannot access it directly, programs and data must be loaded from secondary storage into main memory before they can be used (1).

(b) Two reasons secondary storage is required:

• RAM is volatile, so without secondary storage every file, program and setting would be lost the moment the computer was switched off — including the operating system itself (1)
• RAM is limited in capacity and expensive per byte, so it could not hold everything a user owns; secondary storage provides far greater capacity at a much lower cost per gigabyte (1)

(Also creditworthy: it allows data to be transported between machines, archived or backed up.)

(a) An embedded system is a computer system built into a larger device to perform a specific task or set of tasks, rather than being a general-purpose computer (1).

(b) Embedded: a washing machine controller (or microwave, traffic lights, digital camera, car engine management). Non-embedded: a desktop PC (or laptop, tablet, smartphone) (1).

(c) Difference 1 — purpose and software. An embedded system runs a single fixed program, usually stored in ROM and not changed by the user (1), whereas a non-embedded system is general purpose: the user can install, update and remove whatever software they choose (1).

Difference 2 — resources. An embedded system has limited memory and processing power, matched to the one task it performs (1), which makes it cheaper, smaller and lower-powered than a general-purpose computer, and more reliable because there is far less software to go wrong (1).

(6)

Computer A — recommend SOLID STATE (SSD).

Operation: an SSD uses electrical circuits to persistently store data, holding it as charge in millions of tiny cells, with no moving parts at all (1).

Justification:

• Staff travel constantly, and with no moving parts an SSD is far more durable — it survives knocks and drops that would damage a spinning platter and its read/write head (1)
• Fast boot times are required, and an SSD has no head to move and no platter to spin up, so data access is much faster (1)
• SSDs use less power, giving longer battery life on a laptop, and at only 512 GB the higher cost per gigabyte is affordable (1)

Computer B — recommend MAGNETIC (HDD).

Operation: an HDD stores data as magnetised patterns on rotating platters. A read/write head on a moving arm magnetises tiny areas in one of two directions to write 1s and 0s, and detects the direction of magnetisation to read them back (1).

Justification:

• The budget must be as small as possible and 40 TB is required. Magnetic storage costs far less per gigabyte, and over 40 TB that difference is very large indeed (1)
• The footage is rarely read back, so the SSD's speed advantage would deliver almost no practical benefit and would not justify the cost (1)
• The server is stationary, so the HDD's main weakness — fragility from moving parts — matters far less here than it would in a device carried around (1)

(8)

What cloud storage is. Storing data on remote servers accessed over the internet — it still uses magnetic and/or solid state storage, simply at a remote location rather than in the office (1).

Advantages for this firm

• Staff could access files from anywhere with an internet connection, and from any device, which suits a firm whose staff work away from the office (1)
• Data is backed up by the provider, so it survives fire, theft or a hardware failure at the office — a single local server offers no such protection (1)
• No server hardware to buy or maintain, and capacity can be expanded instantly as large design files accumulate (1)

Disadvantages for this firm

• Cloud storage requires an internet connection, and staff work on site with poor mobile signal — so files may be very slow to open or completely inaccessible exactly when they are needed (1)
• The plans are commercially confidential, yet cloud storage places them with a third party, where a breach at the provider could expose client work the firm has no control over (1)
• The files are large, so transferring them over a limited connection would be slow, and subscription costs recur indefinitely — over several years exceeding the cost of a server (1)

Recommendation. A hybrid approach: keep the working copies on local storage so that poor signal never prevents access and confidential plans stay under the firm's control, while using cloud storage for encrypted backups so the data survives fire or theft at the office (1). If a pure cloud move were required, encryption of files before upload would be essential.

(8)

The main bottleneck is the 4 GB of RAM. Video editing works with very large files, and when RAM fills, the operating system must swap data to and from the magnetic hard disk, which is thousands of times slower than RAM — so the whole system stalls (1). Upgrading the RAM would let far more of the video be held in main memory at once, removing most of that swapping (1).

The second-best upgrade is replacing the hard disk with an SSD. An SSD uses electrical circuits with no moving parts, so there is no read/write head to move and no platter to spin, making access much faster (1). Large video files would load and save far more quickly, and any swapping that still occurs would be far less costly (1).

Considering the alternatives fairly:

• More cores. Her software is written to use multiple cores, so extra cores would genuinely allow more instructions to be processed in parallel (1). However, with only 4 GB of RAM the extra cores would spend much of their time waiting for data, so the benefit would be limited until the memory problem is fixed (1)
• Higher clock speed would give more Fetch-Execute cycles per second, and a larger cache would mean more frequently used data was held close to the CPU rather than fetched from main memory (1). Both give real but incremental gains, neither addresses the bottleneck, and changing the CPU often requires a new motherboard too, making it poor value here

Recommendation: upgrade the RAM and fit an SSD — together these remove the swapping bottleneck and the slow storage that causes it, which will produce a far larger improvement than any CPU change within the same budget (1).

(8)