Memory Hierarchy in Computer Systems

The memory unit that directly communicates with the CPU, holding currently used programs and data.

Backup storage devices like magnetic disks and tapes, holding information not currently used by the CPU.

It optimizes storage by using different types of memory based on speed and cost, from fast but expensive cache to slow but cheap auxiliary.

A very fast memory that stores frequently used programs and data for faster processing, bridging the speed gap between CPU and main memory.

It facilitates communication between main memory and auxiliary memory devices, transferring information as needed.

It limits processing speed because the CPU has to wait for information to be loaded from main memory.

By providing faster access to frequently used data, it reduces the time the CPU spends waiting for information from main memory.

Main memory stores actively used data, while auxiliary memory holds less frequently used data, providing cost-effective storage.

A type of neural network where the output is fed back into the input, letting it learn and remember patterns. It's useful for tasks like speech and image recognition with incomplete or noisy data.

A neural network designed to associate one set of patterns with another. This allows for learning and remembering relationships between different types of data.

A memory system that stores data with additional tags or metadata describing its content. Retrieval happens by matching search queries with these tags.

In contrast to addressing data locations, associative memory compares search queries with stored tags to retrieve matching data.

Associative memory quickly finds data, even with incomplete or inaccurate searches, by using parallel processing and single-step retrieval.

Associative memory is used in specific memory allocation formats to manage and access data efficiently.

Associative memory is key in database management systems. It helps find relevant data quickly, improving database performance.

Associative memory is used in routing tables to quickly find paths for network traffic based on destination addresses.

A cache organization where each memory address maps to a specific location in the cache. It uses an n-bit address to access main memory and a k-bit index to access the cache.

A unique identifier associated with a data word in the cache. It helps determine if the requested data is present in the cache.

Occurs when the requested data is found in the cache. It's a fast and efficient access.

Occurs when the requested data is not present in the cache. It requires accessing the slower main memory.

Number of consecutive words from main memory stored together as a unit in the cache.

An improvement over direct mapping, where each cache location can store multiple memory words under the same index address, forming sets.

Frequent cache misses due to repeatedly accessing data with the same index but different tags in a direct mapping cache.

Addresses the problem of cache thrashing found in direct mapping by allowing multiple memory words under the same index, improving performance.

RAM is assigned addresses 0 to 511, while ROM uses addresses 512 to 1023. This distinction allows the CPU to easily access these memory types.

ROM's data bus only allows data transfer out, while RAM's data bus allows data transfer in both directions. This enables RAM to read and write data, while ROM only reads.

Auxiliary memory provides additional storage for data and programs that are not actively used by the CPU. Examples include disks and tapes.

Access time measures the time it takes to retrieve data from a memory location. It consists of seek time (finding the location) and transfer time (moving data).

Magnetic disks are circular plates coated with magnetized material used for data storage. They are typically stacked together on a spindle, with read/write heads accessing each surface.

Magnetic disks are organized into concentric circles called tracks, which are further divided into sections called sectors. Sectors are the smallest units of data that can be transferred.

Disks rotate at high speed, data is stored in magnetic spots along tracks, read/write heads move across the surface to access data.

Auxiliary memory is not directly accessed by the CPU. Instead, it is accessed through input/output (I/O) channels.

The memory that the computer uses to store data and instructions that are currently being used. It's fast but limited in size and loses data when power is off.

External storage, like a hard drive, where data is stored permanently even when the computer is off. It's slower than main memory but has a much larger capacity.

A small, super-fast memory inside the CPU that stores frequently used data and instructions to speed up processing. It has multiple levels, L1, L2, and L3, with decreasing speed and increasing size.

The fastest cache level, located inside the CPU itself. It's the smallest and has separate sections for data and instructions.

A larger and slightly slower cache than L1, located inside or outside the CPU. It can be shared by multiple cores.

The largest and slowest cache level, located outside the CPU and shared by all cores. It's only present in high-end processors.

The page table is a crucial element of virtual memory management. It translates virtual addresses used by programs into physical addresses that the memory hardware understands. Think of it as a directory that maps virtual memory locations to their corresponding physical memory locations.

Using dedicated registers for the page table would be incredibly fast but comes with significant drawbacks. It's simply not feasible to allocate a dedicated register for every virtual address in a large memory system. The cost and complexity of this approach would be prohibitive.

While storing the page table in main memory is more feasible than dedicated registers, it introduces performance overhead. Every memory access requiring a page translation results in an extra memory access to retrieve the page table entry. This can significantly slow down memory operations.

To achieve efficient virtual memory management, we use specialized high-speed memory called associative memory or translation lookaside buffers (TLBs). These provide fast lookups for frequently accessed page table entries, reducing the overhead of accessing the main memory page table.

A random-access memory page table uses a fixed-size location for each virtual address. When a system has many virtual pages but fewer physical frames, a large portion of the page table remains unused, wasting memory space. In essence, it's like having a large book with many blank pages, only utilizing a small fraction.

Associative memory tackles the space inefficiency of random-access page tables by storing only active page mappings. It uses a fixed number of entries, each capable of storing a page number and its corresponding physical frame. This approach helps to allocate memory more efficiently.

Each entry in associative memory is designed to store a page number and its corresponding physical frame. This structure allows for fast comparisons between the virtual address being accessed and the page numbers within the associative memory. If a match is found, the corresponding physical frame is immediately retrieved.

When a virtual address is supplied, associative memory compares its page number against entries in the memory array. If a match is found, the corresponding physical frame is extracted from the matched entry. This frame is then used to construct the physical address, allowing the memory hardware to access the requested data.