Recap

Distributed Systems

• Definition: more than one machine working together to solve a problem

• Examples:

  • client / server: web server and web client

  • cluster: page rank computation

• Why go distributed?

  • More computing power, more storage capacity, fault tolerance, data sharing

• New challenges

  • System failures: need to worry about partial failure

  • Communication failure: links unreliable

    • bit errors

    • packet loss

    • node/link failure

  • Why are network sockets less reliable than pipes?

• Communication overview

  • Raw messages: UDP

  • Reliable messages: TCP

  • Remote procedure call: RPC

• Raw messages: UDP

  • UDP: user datagram protocol

  • API:

    • Reads and writes over socket file descriptors

      • Socket: a communication connection point (endpoint) that you can name and address in a network

    • Messages sent from / to ports to target a process on machine

  • Provide minimal reliability features

    • Messages may be lost

    • Messages may be reordered

    • Messages may be duplicated

  • Only protection: checksums to ensure data not corrupted

  • Advantages

    • Lightweight

    • Some applications make better reliability decisions themselves (e.g., video conferencing programs)

  • Disadvantages

    • More difficult to write applications correctly

• Reliable messages: layering strategy

  • TCP: Transmission Control Protocol

  • Using software to build reliable logical connections over unreliable physical connections

  • Techniques

    • ACK: sender knows messages was received

    • TIMEOUT: how long to wait?

      • Too long: system feels unresponsive

      • Too short: messages needlessly re-sent, messages may have been dropped due to overloaded server. Resending makes overload worse!

    • Sequence numbers

      • Senders gives each message an increasing unique seq number

      • Receiver knows it has seen all messages before N

    • Buffer messages so arrive in order, timeouts are adaptive

• RPC: remote procedure call

  • Approach: create wrappers so calling a function on another machine feels just like calling a local function! (simplify application development)

  • Help with two components

    • Runtime library

      • Thread pool

      • Socket listeners call functions on server

    • Stub generation

      • Create wrappers automatically

        • Wrappers must do conversions

          • Client arguments to message

          • Message to server arguments

          • Convert server return value to message

          • Convert message to server return value

          • Marshaling / unmarshaling, or serializing / desterilizing

      • Many tools available (rpcgen, thrift, protobufs)

  • RPC over UDP:

    • Use function return as implicit ACK

    • If function takes a long time, then send a separate ACK