Untitled

Distributed Systems Introduction

• A distributed system is where software/hardware components at networked computers communicate and coordinate actions by passing messages.
• Another definition is that it's autonomous computers linked by networks with distributed systems software.
• A distributed system appears as a single computer to its users.
• It involves several computers doing something together, characterized by multiple computers, interconnections, and shared states.
• Examples include the Internet, WWW, web search, online gaming, financial trading systems, email, social networks, and eCommerce.
• Constructing distributed systems involves sharing resources that can be managed by servers and accessed by clients or encapsulated as objects for access by other client objects.

Challenges in Distributed Systems

• Include the heterogeneity of components, openness for adding or replacing components, security, and scalability.
• Scalability is the ability to work well under increased load or user numbers
• Failure handling, component concurrency, transparency, and quality of service are also key challenges.
• Heterogeneity involves diverse networks, computer hardware, OS, PL, and implementations by different developers.
• Heterogeneity applies to networks, computer hardware, operating systems, programming languages, and different developer implementations.
• Openness determines whether a system can be extended and reimplemented, which is achieved by adding resource-sharing services.
• Openness includes providing specifications, documentation of key interfaces, and a uniform communication mechanism

Security Components

• Confidentiality, integrity, and availability are the three security components for information resources
• Denial of service attacks and security of mobile codes can further exacerbate security issues in distributed systems
• A system described as scalable remains effective with increased resources and use
• Scalability involves controlling the cost of physical resources and performance loss and preventing the exhaustion of software resources like IP addresses.

Failure Handling

• Computer systems sometimes fail due to hardware/software faults which causes incorrect results pre-completion.
• Detecting failures uses checksums to detect corrupted data.
• Masking failures involves retransmitting messages or writing file data to multiple disks.
• Recovery from failures involves designing software to recover or roll back permanent data after a server crash.
• Redundancy involves using redundant components to tolerate failures, such as multiple routes in the Internet or replicating name tables in DNS.

Redundancy and Resiliency

• Relying on a singular DNS server creates a single point of failure
• Secondary servers can provide redundancy to reduce the likelihood of service disruptions
• Secondary DNS servers share the load to prevent primary server overload using round-robin DNS as a load balancing technique.

Concurrency

• Applications and services in a distributed system provide resources that can be shared by clients and multiple client requests are generally completed concurrently
• Therefore several clients are likely to attempt to access shared resources at the same time
• Objects in concurrent environments must synchronize their operations to maintain data consistency, a challenge that can benefit multicore systems.
• Concurrency can increase performance, reliability that provides fault tolerance, and specialized processors; some applications are inherently distributed.

Transparency

• Transparency in a distributed system hides its distributed nature, functioning as a normal centralized system to users
• Transparency is defined as hiding the separation of components from users and programmers, so the system is seen as a whole
• Access and location transparency are most important, impacting the utilization of distributed resources.
• Access transparency enables local and remote resources to be accessed using identical operations.
• Location transparency enables resource access without knowledge of physical or network location.
• Concurrency transparency allows processes to operate concurrently without interference.
• Replication transparency enables using multiple resources to increase reliability without knowledge of replicas.
• Failure transparency conceals faults, allowing users to complete tasks despite hardware or software component failures.
• Mobility transparency allows resource and client movement without affecting operation.
• Performance transparency allows system reconfiguration to improve performance with varied loads.
• Scaling transparency allows scaling of system/applications without structural or algorithmic changes.

Services and Servers

• A service manages related resources and presents functionality to users and applications (e.g., file, printing, electronic payment)
• Accessing the service is done through the set of operations it exports
• A server is a running program on a networked computer that accepts requests and responds appropriately
• Requesting processes are clients, and the approach is client-server computing.
• Requests and replies are sent in messages between clients and servers.
• A client invokes an operation on the server, a complete interaction is a remote invocation.

No Global Clock

• Programs coordinate by exchanging messages, depending on a shared sense of the time in which actions occur
• There is no single global notion of the correct time to which computers in a network can synchronize with
• Consequence of only communicating by sending messages.

Distributed Systems Examples

Examples of distributed systems include

• Web Search
• Pervasive networking and the modern Internet
• Mobile and ubiquitous computing
• Massively multiplayer online games (MMOGs)
• Financial Trading

Financial Trading

• Sources are in various formats, like Reuters and FIX events (Financial Information eXchange protocol)
• Adapters translate heterogeneous formats to a common internal format

Web Search Technology

• Web search emerged as a major growth industry as monthly global search volumes increased dramatically
• This requires indexing the entire contents of the World Wide Web
• Google has invested significantly in sophisticated distribution system infrastructure to support web search (ex. Google Earth)
• This requires a large amount of networked computers located at data centers and parallel and distributed computations/storage providing fast access to large datasets.

Pervasive Networking

• The modern Internet is a vast collection of different types of networks like WiFi, WiMAX, Bluetooth and mobile phone networks.
• Programs interact through message passing with a common means of communication.
• Internet enables users to access WWW, email, and file transfer services wherever they are.
• Intranets are company subnetworks protected by firewalls that link through backbones with high transmission capacity.

Mobile and Ubiquitous Computing

• Ubiquitous computing connects small, cheap computational devices present in users' environments
• Computing appears anytime and everywhere, and can occur using any device on any location and with any format.
• Mobile devices such as laptops, handhelds (mobile phones, smart phones, GPS), pagers, PDAs, video/digital cameras, and smart watches are ubiquitous now.
• Devices embedded in appliances such as washing machines, hi-fi systems, cars and refridgerators is another example.

HTTP Protocol

• Hypertext Transfer Protocol (HTTP) is the application layer protocol for the WWW.
• It follows a client-server model where the client is the browser and the server is the WWW server, that sends objects when requested
• http1.0: RFC 1945 and http1.1: RFC 2068
• It leverages connections providing downloads of images, scripts, etc.
• Over TCP providing socket connection at port 80
• The server maintains no information about past client requests, this is known as stateless
• When protocols "state" are maintained, the past history is complex with inconsistent views that must be reconciled

Simple Mail Transfer Protocol

• For email messages on the internet, the underlying protocol is always the same: TCP
• For remote terminal access, the underlying protocol is also the same: TCP
• For web requests, the underlying protocol is also the same: TCP
• For file transfer, the underlying protocol is again: TCP

HTTP Example

• A HTTP client initiates a TCP connection to HTTP server at www.cs.uiuc.edu at port 80
• This is the default port for HTTP servers
• HTTP Client send a HTTP request message that contains the URL into the TCP connection socket
• The HTTP server receives the request message, and then sends a response message containing the requested object (like the index.html) into a socket
• The next step is that the HTTP client receivesthe response with the html file, and it parses to find all the image refs
• All the steps are repeated for each image

HTTP Example Continued

• non-persistent connection - for fetching referenced objects, there is only one object that is fetched via the TCP connection
• persistent object - there are multiple objects transferred within the same TCP connection

Telnet Protocol

• You can interact with any server via the command line program "telnet"
• type "telnet www.google.com 80" to connect to the google web server
• and you can also send HTTP commands to the HTTP web server you're connected to
• By typing a single line containing "GET /index.html" will instruct the target web server to send you data

Cloud Services

• Defined as a set of Internet-based applications, storage and computing services sufficient to distribute local data storage and application software.
• Computing is delivered on demand over the internet and on a pay per use basis.
• Renting computing infrastructure or data centers are an alternative for companies on the internet now.
• Firms can avoid upfront IT costs, and providers can benefit from economies of scale, including consumer services such as Gmail.

Cloud Service Delivery Models

• SaaS delivers software as an application, rather than complex software installation on the computer, it can accessed and utilized over the internet.
• PaaS delivers an development environment as a service, which means the user can run and deploy the application built.
• IaaS infrastructure is used by the clients (can be used for storage, network or firewalls) providing multiple resources like load balancing and networking.