Parallel & Distributed Software Architectures

Various hardware and software architectures exist that are usually used for distributed computing. At a lower level, it is necessary to interconnect multiple CPUs with some sort of network, regardless of that network being printed onto a circuit board or made up of several loosely-coupled devices and cables. At a higher level, it is necessary to interconnect processes running on those CPUs with some sort of communication system and execution environment. We look at some of the main approaches to the design of parallel and distributed software architectures.

Client-Server Computing

Client-Server Computing is a model of software construction and process interaction which allows for the separation, decomposition and potential distribution of system and application functionality. A server is an autonomous process which executes and fulfils requests which are generated and sent to it by client processes.

A process fills the role of a server in a system if it behaves as follows:

1. Provides services through an interface. An interface is a well defined protocol describing the mechanism by which service functions may be known and acquired and the specification of expected parameter types and returned results.

2. Responds to requests by returning expected results to the client.

3. Hides implementation and complexity of service functions.

A process fills the role of client in a system if it behaves as follows:

1. Presents a standard interface to the server and the user.

2. Forms queries by translating user input to server operations.

3. Initiates communication with the server.

4. Performs data analysis on results and displays them to the user.

With this architecture a traditional large centralised application can be downsized by dividing its functionality into separate client and server processes which can be located on different hosts. Server processes can be assigned to high performance processing hosts with large non-volatile storage availability and can offer concurrent request processing to networked clients. Client processes can run at a user's locality availing of access to local display hardware for presenting results and to reduce server load by dealing with user interactivity and input devices.

Multi-tier Architectures

A generic Client/Server architecture has two types of nodes on the network: clients and servers. As a result, these generic architectures are sometimes referred to as "two-tier" architectures.

Some networks will consist of three different kinds of nodes, clients, application servers which process data for the clients, and database servers which store data for the application servers. This is called a three-tier architecture.

In general, an n-tier or multi-tier architecture may deploy any number of distinct services, including transitive relations between application servers implementing different functions of business logic, each of which may or may not employ a distinct or shared database system.

The advantage of an n-tier architecture compared with a two-tier architecture (or a three-tier with a two-tier) is that it separates out the processing that occurs to better balance the load on the different servers; it is more scalable. The disadvantages of n-tier architectures are that it puts a greater load on the network and it is much more difficult to program and test software than in two-tier architecture because more devices have to communicate to complete a user’s transaction.

Cluster

A computer cluster is a group of loosely coupled computers that work together closely so that in many respects it can be viewed as though it were a single computer. Clusters are commonly, but not always, connected through fast local area networks. Clusters are usually deployed to improve speed and/or reliability over that provided by a single computer, while typically being much more cost-effective than single computers of comparable speed or reliability.

High-availability (HA) clusters

High-availability clusters are implemented primarily for the purpose of improving the availability of services which the cluster provides. They operate by having redundant nodes, which are then used to provide service when system components fail. The most common size for an HA cluster is two nodes, which is the minimum required to provide redundancy. HA cluster implementations attempt to manage the redundancy inherent in a cluster to eliminate single points of failure. There are many commercial implementations of High-Availability clusters for many operating systems. The Linux-HA project is one commonly used free software HA package for the Linux OS.

Load balancing clusters

Load balancing clusters operate by having all workload come through one or more load-balancing front ends, which then distribute it to a collection of back end servers. Although they are implemented primarily for improved performance, they commonly include high-availability features as well. Such a cluster of computers is sometimes referred to as a server farm. Linux Virtual Server (LVS) is an advanced load balancing solution for Linux systems.There are many commercial load balancers available including Platform LSF HPC, Moab Cluster Suite.

High-performance (HPC) clusters

High-performance clusters are implemented primarily to provide increased performance by splitting a computational task across many different nodes in the cluster, and are most commonly used in scientific computing. One of the more popular HPC implementations is a cluster with nodes running Linux as the OS and free software to implement the parallelism. This configuration is often referred to as a Beowulf cluster. Such clusters commonly run custom programs which have been designed to exploit the parallelism available on HPC clusters. Many such programs use libraries such as MPI which are specially designed for writing scientific applications for HPC computers.

HPC clusters are optimized for workloads where jobs or processes running on the cluster nodes communicate actively during the computation.

Grid Architecture

Grid computing or grid clusters are a technology closely related to cluster computing. The key differences between grids and traditional clusters are that grids connect collections of computers which do not fully trust each other, and hence operate more like a computing utility than like a single computer. In addition, grids typically support more heterogeneous collections of nodes than are commonly supported in clusters.

Grid computing is optimized for workloads which consist of many independent jobs or packets of work, 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, but intermediate results of one job do not affect other jobs in progress on other nodes of the grid.

Peer to Peer Architectures

A peer-to-peer computer network is a network that relies on the computing power and bandwidth of the participants in the network rather than concentrating it in a relatively low number of servers. Such networks are useful for many purposes. Sharing content files containing audio, video, data or anything in digital format is very common, and realtime data, such as telephony traffic. A pure peer-to-peer network does not have the notion of clients or servers, but only equal peer nodes that simultaneously function as both "clients" and "servers" to the other nodes on the network. This model of network arrangement differs from the client-server model where communication is usually to and from a central server.

Mobile Code

Mobile code includes software obtained from remote systems, transferred across a network, and then downloaded and executed on a local system without explicit installation or execution by the recipient. Examples of mobile code include scripts (JavaScript, VBScript), Java applets, ActiveX controls, and macros embedded within Office documents.

The term mobile agent refers to a process that can transport its state from one environment to another, with its data intact, and still be able to perform appropriately in the new environment. Each node must provide a mobile agent with a standard interface or operating environment through which it can carry out its work. An agent can be moved to where data is located rather than shipping the data to a remote location.