COMUTER NETWORKING// do not study from this. it's unreliable1. BASICS> network: conneced computers> Internet: A popular network; computers (end systems) connected by links and packet switches> packet: data segments> protocols: formats of exchanging packets> hosts: end systems which run applications; servers and clients; "network edge"> end systems are connected to "edge routers" > access networks allow access to Internet> DSL (Digital Subscriber Line) and cable are popular in residential areas > DSL uses existing phone lines > telephone company provides modem (hence, ISP = telephone company) > modem converts encodes signals from phone and comps at different frequencies >> high speed download, high freq >> medium speed upload, med freq >> two way phone, low freq (unchanged) > sends analog signals to DSL Accsess Multiplexer (DSLAM) at company's central office (CO) > customer uses splitter to differentiate bw incoming signals > "frequency division multiplexing"> cable system uses existing tv cables > similar to DSL, except cables are shared > hence, congestion reduces speed > house to common node -> coaxial; node to Cable Modem Termination System (CMTS) -> fibre > hence, Hybrid Fibre Coax (HFC)> fibre To The Home (FTTH) > new; house connected directly by fibre; fast (Gbps)> Satellite = slow> Dial-up: similar structure to DSL, but very slow> LAN: all computers in a region are connected to a single edge router> popular LAN = ethernet (wired & wireless), uses twisted copper wire> Moving data> packet switching > common method: store and forward; switch stores complete message before sending it to link > storing = buffering; can lead to delays > packets to be sent are stored in the output buffer/queue > if buffer is full, leads to clashes, packet drops, queuing delays > paths are determined using forwarding tables, etc. not fixed> circuit switching > paths are reserved for a certain session; constant speed; eg. telephones > ps is more efficient> Network structure> tier 1 ISPs, regional ISPs, access ISPs> pop (point of presence) : other isps can connect to tier 1 isps> ixp (internet exchange point) : isps can connect to each other> Delays> processing delay, queuing delay> transmission delay: time to send all the packets of a message into a link (depends on internet speed)> propagation delay: time it takes for bits to reach other switches; depends on wire used > throughput: rate at which data is received by a host > layers: application, transport, network, link, physical> each layer's frame: header + data (like the previous layer's frame)> uses encapsulation /* 2. APPLICATION LAYER */> network application architecture: p2p or server-client> processes in different end-systems communicate> send messages through "sockets" (interface bw application processes and transport layer protocols) > transport protocols may provide services (reliable data transfer, constant throughput, timing, security)> internet provides tcp and udp; applications can choose either; ssl provides encryption to tcp (if inmplemented)> application layer protocols: http, smtp, ftp, telnet, etc.> HTTP, port 80, HTML transfer> uses tcp; web servers, browsers; > http servers do not store data; they just serve; hence http is "stateless"> web servers are always on; constant ip > persistent connection: many reuests/ replies are sent over the same tcp connection > non persistent: opposite > http uses persistent by default > in non persistent, connections may be parallel or serial (one after another) > most browsers - 5 to 10 parallel connections; configurable > non persistence may cause load on servers (multiple connections to be maintained) > persistent connections can use pipelining > pipelining: multiple requests without waiting for replies> http message: 1 request line (get, post, delete, etc.), multiple header lines, data> web servers can use cookies to identify users; included in headers> cookies are managed by the browser and by the server> Web cache/proxy servers > satisfies requests in place of web server; keeps copies of recently requested objects > browsers can be setup to use proxies first; if object is not in cache, cache connects to server > caches are usually installed by isps; browsers are configured to point to the cache > fast, reduce traffic > usually uses conditional get statements (if modified since) > Content Distribution Networks (CDNs) companies can install caches; popular> FTP, port 20 for connection control, port 21 for data, file transfer> uses two parallel tcp connections; control info is separate from data (file); "out-of-band" > one file per data connection; maintains state info about user; opens new data connection if more files are requested> commands: stor, retr, user, passwd, etc. > SMTP, port 25, internet mail> user agent (eg. MS Outlook), mail server, smtp; uses TCP> each user has a mailbox in the mail server > A -> A's mail server -> B's mail sever -> B; no intermediate mail servers> entire message must be in 7-bit ascii> smtp is a "pull" protocol; can only upload mail; other protocols are used to access mail (pop3 (simple), imap (complex)) > web based mail (gmail, yahoo, etc.):> A -(HTTP or custom protocol)-> A's mail server -(SMTP)-> B's mail server -(HTTP/custom)-> B> Domain Name Systems (DNS), used by application protocols> translates readable site name to ip address> host name: eg. relay1.west-coast.gmail.com (canonical name), www.gmail.com (alias name)> if a canonical host name has multiple servers, each time a different server ip is provide by DNS; distributes load > Heirarchy: Root servers > Top Level Domain servers (org, com, net, in, etc.) -> Authoritative server (contains actual ip address)> User -> local dns server (provided by isp) -> Root -> TLD (and other intermediate servers) -> Authoritative> local DNS server can cache ip addresses to be more efficient> DNS servers store resource records (RRs)> (name , value, type, time to live); if type is A, then authoritative record; else, value contains ip of next dns server to go to > DNS messages contain headers, RRs, queries, replies, additional data, etc.> Peer to Peer (P2P)> peers upload to and download from each other > no need of complex architecture; can cause heavy load on isp's servers > less distribution time (minimum time required for all peers to get the file/data) than client-server, because everyone uploads > bittorrent, ex of p2p> uses trackers (special servers); registers users when they use torrents> peers give priority to other peers who give them priority> uses dht's; each active peer has parts of a hash table which contains info about which peer has which file's data > processes send messages into sockets; each socket has an associated port number and are created by applications > user datagram protocol (UDP): messages are sent without handshaking> transmission control protocol (TCP): handshaking; client sends message to server's welcoming socket; server acknowledges > and creates a connection on a new port with the client's port (connection socket) /* 3. TRANSPORT LAYER */> provides a logical connection bw two processes (NOT HOSTS) on different hosts> UDP only provides data delivery and error checking; TCP also provides congestion control and reliable data transfer (even though IP is unreliable)> Extrending host-to-host delivery to process-to-process delivery is transport layer de/multiplexing> demultiplexing: deliver data in transport layer segment to correct socket > multiplexing: obtain data from sockets, add relevant headers> ports nos. : 0 to 65535; data in transport segment = (source port, destination port, headers, data (applicaton segments, etc.))> udp demultiplexes using only destination ip address and destnation port number> tcp uses source and destination ip addresses and port numbers> hence each process may have multiple tcp sockets (uses multi-threading = using light-weight sub-processes)> UDP> connectionless; bare minimum; provides a checksum for error checking> TCP> header contains additional data, such as sequence numbers, ACK numbers, timers, etc.> is full duplex; hosts can send messages to each other at the same time