Classification Based On Number Of Processors

1. Single Processor Systems 2. Multiprocessor Systems 3. Clustered Systems

Sep 2007 [email protected]

Single Processor Systems

One CPU present. Most commonly found, from mainframes to laptops. special purpose processors may be present like:Math processor Disk processor Keyboard processor Video processor Audio processor I/O processor Co-processor
Sep 2007 [email protected]

Multi Processor Systems (Parallel or Tightly Coupled)

Have 2 or more General Purpose Processors in close communication. Normally the processors share the computer bus, clock, memory and peripheral devices. Are of importance and are increasing in number. Are much costlier but have 3 major advantages :1. Increased throughput 2. Economy of scale 3. Increased reliability
Sep 2007 [email protected]

Increased throughput
Throughput is the amount of work that a

computer can do in a given time period.

However speed up ratio is not equal to N but a little lesser, where N is the no. of processors (because the bus, memory, peripheral devices are shared).

Sep 2007

[email protected]

Economy of scale
1. 2. 3. 4. Works out cheaper or there is economy with scaling of processors. Compare multiple Single Processor System with a single Multi Processor System. Which costs lesser ? And Why? They share power supplies They share peripherals They share mass storage If several programs ar operating on the same data just one copy of the data is sufficient.
[email protected]

Sep 2007

Increased reliability
Failure of one processor need not halt the system. The other processors can take up the work to completion. Such systems are termed fault tolerant or systems with graceful degradation

Sep 2007

[email protected]

Fault Tolerant
even if a processor or a component fails, the operation of the system is not hindered. faults are detected, diagonised and if possible corrected too. Eg:- The HP non-stop system or the HP tandem system it uses both hardware and software duplication to ensure continued operations. It has multiple pairs of CPUs which work in lock step.
Sep 2007 [email protected]

Lock Step
A pair of processors will execute the same instructions and will compare the results. If there is a mismatch it means one CPU is corrupt. Immediately both are halted. The executing process is carried out by another pair of processors. ( is an expensive solution)

Sep 2007

[email protected]

Graceful Degradation
The ability to provide service proportional to the level of surviving hardware is called graceful degradation.

Sep 2007

[email protected]

Types Of Multiprocessors
2 kinds namely symmetric and asymmetric multiprocessors.

Sep 2007

[email protected]

Symmetric Multiprocessors
Are more common. All CPUs are like peers. If n processors present, n processes can get executed in true parallel way.
Eg:- Sun Solaris operating system. (it can be configured to employ dozens of processors all running Solaris.) What care has to be taken in SMS?
Sep 2007 [email protected]

 Sophisticated I/O control is required to ensure that the right data reaches the right processor. Disparity among the processors to be minimised i.e. one sitting idle and the other unduly getting overloaded. Note:- All modern operating systems including Windows, Windows XP, Mac OS X, and Linux support SMP.

Sep 2007

[email protected]

Asymmetric Processors
Here each processor is assigned a distinct task. A master processor controls all others. The relationship amongst the processors is masterslave. The master processor schedules and allocates work to the slaves. Eg:- Sun OS (version 4) provided asymmetric multiprocessing.
Sep 2007 [email protected]

Recent Trends in CPU design

Dual core processors and multicore processors. They are found in fast higher ended machines. All the core cpus are kept in the same chip. They look like n standard processors to the o.s. Blade servers are a recent development. Their web definition is here:-

Sep 2007

[email protected]

 A blade server is essentially a housing for a number of individual minimally-packaged computer motherboard "blades", each including one or more processors, memory, storage, and network connections, but sharing the common power supply and air-cooling resources of the chassis.
Sep 2007 [email protected]

Definition of blade PC
A PC architecture that houses multiple PC modules ("blades") in a single chassis. It takes the machines off the users' desks and houses them in rack mounted cabinets in the datacenter similar to blade servers. The user's keyboard, monitor and mouse plug into a device at the desk that is wired via a TCP/IP or direct connection to the assigned blade in the datacenter.

Sep 2007

[email protected]

Advantages Of Blade PCs

Having the physical PCs in the datacenter reduces noise and heat at users' desks. the pcs are inherently secure in a locked room. Data cannot be downloaded or uploaded because the USB ports and hard disk drives are away from the desk. Incase of user/employer relocation the actual machine relocation is avoided. If a PC fails, a spare blade may be available in each chassis for hot swapping. Repairs are always performed in the same place, and technicians do not have to travel throughout the building.

Sep 2007

[email protected]

A Blade PC

Sep 2007

[email protected]

 ClearCube Technologies was the first to make blade PCs, which were introduced in 2000. This ClearCube cage contains eight fully functional PCs, each with its own storage (hard disk). A one-foot rack can hold nearly 20 PCs. So there is good space management. ClearCube management software backs up the data to spare network drives, periodically. if a blade fails, the mgmt software is used, to switch the user to a spare blade and restore the data.

Sep 2007

[email protected]

A Blade Server

Sep 2007

[email protected]

 Blade servers such as this unit from HP are widely used in datacenters to save space and ease systems management. Managed by HP's Insight Manager, this unit has redundant power supplies . The exposed blade (left) is a complete server with hard disk. ---------------Sep 2007 [email protected]

Clustered Systems
A computer cluster is a group of tightly coupled computers which work together closely, so that in many respects they can be viewed as though they are a single computer. The components of a cluster are in most cases connected to each other through fast LANs and they also share storage. Clusters are usually deployed to improve performance and availability Also typically they are much more cost-effective than single computers of comparable speed.
Sep 2007 [email protected]

Sep 2007

[email protected]

 High Availability is possible by adding a layer of redundancy. Imagine a layer of cluster software which runs on the nodes. Each node is made to monitor 1 or 2 other nodes. If by chance a monitored machine fails, the monitoring machine can take ownership and can restart the application. (the users cannot really notice this. There will be a small interruption of service, thats all.)
Sep 2007 [email protected]

Types Of Clustering
There are different kinds of clusters namely namely:(i) Asymmetric (ii) Symmetric (iii) Parallel clusters (iv) WAN clusters

Sep 2007

[email protected]

Assymetric Clustering
Means the machines are not alike in their working philosophy. While one machine runs an application the other is just a standby, also termed hot standby. The standby does nothing but monitors the active server. And if the server fails the standby will take over and will be the active server!
Sep 2007 [email protected]

Symmetric Clustering
Here, 2 or more hosts are running applications and are monitoring one another. If one fails the other will take over. This is more efficient because the machines act alike and resources on both machines get used fully.

Sep 2007

[email protected]

Parallel Clustering
Parallel clusters allow multiple hosts to access the same data on the shared storage. (cost effective & more reliable) To provide multiple access simultaneously the operating system too should provide this facility and locking to ensure that no conflicting operations occur. This is termed distributed lock manager.
Sep 2007 [email protected]

 However, since most operating systems lack support for simultaneous data access by multiple hosts, special versions of the software/applications are being used. Eg:- Oracle parallel server is a version of Oracles data base server designed to run on a parallel cluster.
Sep 2007 [email protected]

WAN Clusters
Custering over a wide area network means the nodes are separated by high distances. Eg:- a city area or so.

Sep 2007

[email protected]

SAN and clusters

Storage Area Networks allow many systems to attach to a pool of storage. If the applications and their data are stored on the SAN, the cluster software can assign the application to run on any host. If the host fails any other host can take over.
Sep 2007 [email protected]

Grid computing
grid clusters are a technology closely related to cluster computing. The key differences (by definitions which distinguish the two at all) between grids and traditional clusters are that grids connect collections of computers which do not fully trust each other, or which are geographically dispersed. Grids are thus more like a computing utility than like a single computer. In addition, grids typically support more heterogeneous collections than are commonly supported in clusters.

Sep 2007

[email protected]

 Grid computing is optimized for workloads which consist of many independent jobs which do not have to share data between the jobs during the computation process. Grids serve to manage the allocation of jobs to computers which will perform the work independently of the rest of the grid cluster. Resources such as storage may be shared by all the nodes..
Sep 2007 [email protected]

 An example of a very large cluster is the folding@home project. It is analyzing data that is used by researchers to find cures for diseases such as Alzheimer's and cancer. Another large project is the seti@home project, which may be the largest distributed cluster in existence. It uses approximately three million home computers all over the world to analyze data from the Arecibo Observatory radiotelescope, searching for evidence of extraterrestrial intelligence.
Sep 2007 [email protected]

Computing Environments
Traditional computing Client-server computing Peer to peer computing Web based computing

Sep 2007

[email protected]