Concepts with Symbian OS (779878), страница 8
Текст из файла (страница 8)
DMA is a form of asynchronous I/O, but differsfrom the generic form. Asynchronous I/O is fine-grained: it signals theCPU whenever there is even a small amount of data to transfer. DMAis very coarse-grained and assigns all data operations to the device. Theoperating system starts the I/O operation and is only notified when it iscomplete.There are, then, three modes of device communication: synchronous,asynchronous and DMA.• A handheld Linux device that plays video is likely to use synchronouscommunication between the video driver and the operating system.
Display of video is a real-time application and most real-timeapplications require synchronous I/O.• Computers with windowing systems use asynchronous I/O to monitorGUI devices such as a mouse. When a mouse moves, it generatesinterrupts that cause the operating system to read the mouse events.When the mouse does not move, the operating system can safelyignore it and move on to other duties.• Computers use DMA for larger I/O tasks. Consider reading from adisk drive. It is enough that an operating system would send a diskdrive a command to read a block of data, along with the parametersCOMPUTER STRUCTURES29needed to complete the transfer.
Reading program code from a diskto execute, for example, is usually a task that is executed using DMA.Each I/O method carries with it implications for system performance.With synchronous I/O, the operating system spends all its time monitoringand servicing devices. This means that performance and response to usersand other services is slower than with other methods. Asynchronous I/Orelieves the operating system from constant monitoring and, therefore,performance and system response increases. DMA frees the operatingsystem from almost all device I/O responsibilities and therefore producesthe fastest system service and response time. Most operating systems usea combination of methods to gain an efficient design.Storage StructuresAlong with central computer operation and device I/O, storage makesa third essential component of a computer system.
The ability to recordinformation and refer to it again is foundational to the way moderncomputer systems work. A system without storage would not even beable to run a program, since modern systems are based on storedprograms.2 Even if it was able to run instructions (perhaps asking the userfor each instruction), input could not be stored and output could only begenerated one byte at a time.The core computing cycle is very dependent on storage. This corecomputing cycle, often referred to as the ‘fetch–execute’ cycle, fetches aninstruction from memory (storage), puts the instruction in a register (morestorage), executes that instruction by possibly fetching more information(more storage), and storing the results of the execution in memory(even more storage).
This basic computing cycle is part of a design firstdeveloped by John von Neumann, who built it into a larger computingsystem based on sequential computer memory and external storagedevices.The many storage mechanisms of a computer system can be viewedas a hierarchy, as shown in Figure 2.7. Viewing these systems togetherallows us to consider their relationships with one another.2Certainly, computers without disk storage are used every day. But note that even thesecomputers have memory for storage – sometimes large amounts of it. Any computer systemhas storage at least in the form of memory or registers accessible by the CPU. Most systemsbuild their storage requirements from there.30THE CHARACTER OF OPERATING SYSTEMSRegistersCacheMain memoryDisk spaceOptical storageArchival storageFigure 2.7 Storage hierarchy• Registers are at the top of the hierarchy.
This collection representsthe fastest memory available to a computer system and the mostexpensive. Depending on how a processor is constructed, there maybe a small or large set of these memory cells. They are typically usedby the hardware only, although an operating system must have accessto (and therefore knowledge of) a certain set of them. Registers arevolatile and therefore represent temporary storage.• Storage caches represent a buffer of sorts between fast register storageand slower main memory. As such, caches are faster and moreexpensive than main memory, but slower and cheaper than registermemory. On a typical computer system, the caching subsystem isusually broken into sublevels, labeled ‘L1’, ‘L2’ and so forth.
Thehierarchy continues to apply to these sublevels; for example, L1 cachesCOMPUTER STRUCTURES31are faster and more expensive than L2 caches. Caches represent amethod to free up the hardware from waiting for reads or writes tomain memory. If an item being read exists in cache, then the cachedversion is used. If data needs to be written, then the cache controllertakes care of the writing and frees up the CPU for more programexecution. Caches are volatile and therefore also represent temporarystorage.• Main memory represents the general-purpose temporary storage structure for a computer system.
Program code is stored there while theprogram is executing on the CPU. Data is stored in the main memory temporarily while a program is executing. The I/O structures,discussed in the previous section, use main memory as temporarystorage for data. This type of memory is usually external to the CPUand is sometimes physically accessible by the user (for example, ondesktop systems, users can add to main memory or replace it).• Secondary storage is a slower extension of main memory that holdslarge quantities of data permanently.
Secondary storage is used to storeboth programs and data. The first – and still most common – form ofsecondary storage is magnetic disks. These store bits as small chunksof a magnetic medium, using the polarity of magnetic fields to indicatea 1 or a 0. Faster storage has evolved more recently in the form ofelectronic disks, large collections of memory cells that act as a disk.Formats such as compact-flash cards, secure-digital cards and miniSD cards all provide permanent storage that can be accessed in arandom fashion.
These are used in the same way as magnetic mediato manipulate file systems.• Tertiary, or archival, storage is meant to be written once for archivalpurposes and stored for a long period of time. It is not intended tobe accessed often, if ever. Therefore, it can have slow access timesand slow data-retrieval rates. Examples here are magnetic tape andoptical storage such as compact discs (CD-ROMs).
CD-ROMs can bethought of as lying between secondary and tertiary storage, becauseaccess time on CDs is quite good.There are several concepts built into this storage hierarchy that affecthow an operating system treats each storage medium. The first, and mostbasic, is the model used to access storage. The idea of a file as a group32THE CHARACTER OF OPERATING SYSTEMSof data having a specific purpose has been the model of access usedsince almost the invention of permanent storage. If many files can bestored on a medium, there is also the need for organization of thosefiles. Ideas such as directories and folders have been developed for thisorganization.
The way that these concepts have been implemented iscalled a file system. The design and appearance of file systems differsacross operating systems while the concepts of files and the structure ofdirectories remain constant.The concept of access rights has proven useful in implementing securestorage. In some systems, access to storage is granted to any processrequesting that access.
In other systems, processes requesting access tostorage must present identification along with the request and are onlygranted the access that the identification gives them. In these types ofsystems, there is typically an owner of a unit of storage and the ownersets up how others may access that unit. Note that this requires that thesystem using these access rights establish a method of user- or processidentification. For example, Symbian OS establishes identification basedon a process’s function within the operating system. There are systemprocesses and non-system (user) processes; in addition, there are otherprocesses that have more privileges than users but not the completeprivileges of the system.
Access to storage on Symbian OS is grantedbased on these classifications of processes.Another concept that has evolved from the hierarchy of storage iscaching. As shown in Figure 2.7, speed of storage access decreases asyou work down the hierarchy. Caches were developed as a way toshield devices from slower storage. Cache management has become animportant issue. For example, if a cache is full and the CPU needs to writemore data to it, some data already in the cache is overwritten. If the cacheis managed carefully, the ‘relevant’ data is kept in the cache and the rest iswritten to the next level. However, the meaning of ‘careful management’is different depending on the design of the operating system.The idea of virtual storage is a concept that works across the storagehierarchy.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
