Operating Systems CMPSCI 377 Networks

1
Operating SystemsCMPSCI 377Networks

• Emery Berger and Mark Corner
• University of Massachusetts Amherst

2
Networks and OS

• Networks and OS are strongly linked
• Want to know about networking guts?
• Take 453
• Concentrate on OS abstractions
• Multiple layers of abstraction accessible
• Pick the simplest one
• Pick the one with the functionality you need
• Raw, TCP, simple socket, RPC, RMI , etc.

3
Hardware Reality

• Many kinds of networks
• Ethernet, FDDI, Cable, DSL, ATM
• Only delivery locally
• Multiple hops to get to destination
• Destination named by MAC address
• Small Messages
• Machine to Machine communication
• Unordered, unreliable, insecure

4
Abstractions Provided by OS

• Network/device independence
• Routing to final destination (not the OS)
• Destinations have readable names
• Large messages
• Process-Process communication
• Ordered, reliable
• Streams of bytes or even function calls
• Security
• More (distributed file systems, shared mem)

5
Communication Protocols

• Protocol agreed-upon rules for communication
• Protocol stack layers that comprise networking
  software
• Each layer N provides service to layer N1

6
Abstraction and Protocol Layers
NFS
HTTP
Ssh
SMTP
RPC
TCP
UDP
IP
ATM
Ethernet
7
UDP

• User Datagram Protocol
• Not much of an abstraction
• Lossy, unordered, individual messages
• Why use this?

8
TCP

• Transmission Control Protocol
• Lowest layer typically used by systems
• Provides
• Reliability
• Ordering
• Byte Stream
• no message boundaries!
• Remember reality
• Losses, reordering, individual messages

9
Process to Process

• Process to process communications
• Reality is machine to net to net to machine
• Machines have addresses
• Typically IP addresses
• Port
• endpoint sub address
• Think apartment building is IP
• Apartment number is port number
• How do I know the right port number?

10
Sockets

• Standard API for network programming
• Duplex communication (2-way)
• OS Abstraction for a connection
• Can configure sockets to use UDP or TCP
• UDP best for lossy communication
• Pings, video, audio, other multimedia
• Sends datagrams (packets)
• TCP most stuff
• Sends reliable stream of bytes

11
Socket communication

• Sockets (in TCP) send stream of bytes
• Multiple sends may be combinedinto one received
  message
• To send multiple application-level messages, must
  have application level protocol
• E.g., 4-bytes message type, 4-bytes message
  length, plus message

12
Client-side using sockets

• Creates socket
• int fd socket (AF_INET, SOCK_STREAM, 0)
• Converts hostname to IP address
• getaddrinfo (name, portStr, NULL, portNum)
• Connects to host
• connect (fd, addr, len) // from getaddrinfo
• Once connected, can send or recv
• send (fd, data, amount, 0)
• recv (fd, buf, len, 0)

13
Server-side using sockets

• As before
• Create socket as before
• Convert hostname to IP address with getaddrinfo
• getaddrinfo (name, portStr, hint, addrinfo)
• Binds socket to address
• bind (fd, addr, len) // from getaddrinfo
• Listens for connection
• listen (fd, 1) // number of waiting msgs
• Accepts connection ) new socket
• int newfd accept (fd, addr, len)
• Why does this return a socket number?

14
Exercise

• Lamest web server ever
• Wait for connections on localhost port 80
• Read message until first newline (\n)
• Send lthtmlgtltbodygtHello worldlt/bodygtlt/htmlgt
• Use
• int fd socket (AF_INET, SOCK_STREAM, 0)
• getaddrinfo (name, portStr, hint, addrinfo)
• bind (fd, addr, addrlen)
• int v listen (fd, 1)
• int newfd accept (fd, addr, len)

15
Systems Patterns

• Two dominant forms of building systems
• Client-Server (ie. web server and client)
• Peer-to-peer (ie. BitTorrent)
• Not completely distinct
• Google is which?
• More generally Distributed Systems

16
Distributed Systems

• Physically separate processors connected by one
  or more communication links
• no shared clock or memory
• Most systems today distributed in some way
• email, file servers, printers, remote backup,
  web..

P2
P1
P4
P3
17
Parallel vs. Distributed Systems

• Tightly coupled parallel processing
• Processors share clock, memory, run one OS
• Frequent communication
• Loosely coupled distributed computing
• Each processor has own memory, runs independent
  OS
• Infrequent communication

18
Advantages

• Resource sharing
• Computational speedup
• Reliability
• Communication
• Close to some physical resource

19
Advantages

• Resource sharing
• Resources need not be replicated
• Shared files
• Expensive (scarce) resources can be shared
• Poster-size color laser printers
• Present same environment to user
• Keeping files on file server

20
Advantages, continued

• Computational speedup
• n processors n times computational power
• SETI_at_home
• Problems must be decomposable
• Trivial embarrassingly parallel
• Coordination communication required between
  cooperating processes
• Synchronization
• Exchange of results

21
Advantages, continued

• Reliability
• Replication of resources provides fault tolerance
• One node crashes, user works on another one
• Performance degradation but system available
• Must avoid single point of failure
• Single, centralized component of system
• Example central file servers

22
Advantages, continued

• Communication
• Users/processes on different systems can
  communicate
• Mail, transaction processing systems like
  airlines banks, www

23
Abstractions for Building

• RPC Remote Procedure Call
• Lots of systems built on this
• Children RMI, CORBA, DCOM
• Programmer doesnt see socket
• Sees blocking function call
• Difficult to get right when things fail
• Lockups, etc.

24
AJAX

• Asynchronous JavaScript and XML
• Ie. google maps
• Move data asynchronously
• Ask for data, when it shows up, great.
• Interact with data locally
• Multiple data sources, display is fast

25
The End
26
Networks

• Goal Efficient, correct, robust message passing
  between two separate nodes
• Local area network (LAN) nodes in single
  building, fast reliable (Ethernet)
• Media twisted-pair, coax, fiber
• Bandwidth 10-1,000MB/s
• Wide area network (WAN) nodes across large
  geographic area (Internet)
• Media fiber, microwave links, satellite channels
• Bandwidth 1.544MB/s (T1), 45 MB/s (T3)

27
Principles of Net Communication

• Data broken into packets ( 1Kb)
• Basic unit of transfer
• Computers routers control packet flow
• Road analogy
• Packets cars
• Network roads
• Computer traffic lights (intersection)
• Too many packets on shared link/node traffic jam

28
Traditional Layers

• Application layer applications that use the net
• Presentation layer data format
  conversion(big/little endian)
• Session layer implements communication
  strategy(e.g., RPC)
• Transport layer reliable end-to-end
  communication
• Network layer routing congestion control
• Data link control layer reliable point-to-point
  communication over unreliable channel
• Physical layer electrical/optical signaling
  across wire

29
TCP/IP Protocol Stack

• Internet standard protocol stack
• TCP reliable protocol packets received in
  order
• UDP (user datagram protocol) unreliable
• No guarantee of delivery

30
Packet Format

• All info needed to recreate original message
• Packets may arrive out of order
• need sequence number
• Data segment contains headers for higher protocol
  layers application data

31
Network Topologies

• Connection of nodes impacts
• Maximum average communication time
• Fault tolerance
• Expense
• Two basic topologies
• Bus (for LANs)
• Point-to-point

32
Bus Network Topologies

• Bus nodes connect to common network
• Linear bus single shared link
• Nodes connect directly to each other via bus
• Inexpensive (linear in of nodes)
• Tolerant of node failures
• Ethernet LAN

33
Bus Network Topologies

• Ring bus single shared circular link
• Same technology tradeoffs as linear bus

34
Point-to-Point Network Topologies

• Fully-connected
• Each message takes one hop
• Node failure no effect on communication with
  others
• Expensive impractical for WANs

35
Point-to-Point Network Topologies

• Fully-connected
• Each message takes one hop
• Node failure no effect on communication with
  others
• Expensive impractical for WANs

36
Point-to-Point Network Topologies

• Fully-connected
• Each message takes one hop
• Node failure no effect on communication with
  others
• Expensive impractical for WANs

37
Point-to-Point Network Topologies

• Fully-connected
• Each message takes one hop
• Node failure no effect on communication with
  others
• Expensive impractical for WANs

38
Point-to-Point Network Topologies
2
3
1
4

• Partially connected
• Links between some, but not all nodes
• Less expensive, less tolerant to failures
• Single node failure can partition network
• Several hops requires routing

39
Point-to-Point Network Topologies
I want to goto Node 4!
2
3
1
4

• Partially connected
• Links between some, but not all nodes
• Less expensive, less tolerant to failures
• Single node failure can partition network
• Several hops requires routing

40
Point-to-Point Network Topologies
I want to goto Node 4!
2
3
1
4

• Partially connected
• Links between some, but not all nodes
• Less expensive, less tolerant to failures
• Single node failure can partition network
• Several hops requires routing

41
Point-to-Point Network Topologies

• Tree structure network hierarchy
• Messages fast between direct descendants
• Max message cost?
• Not failure tolerant
• Any interior node fails network partitioned

42
Point-to-Point Network Topologies

• Star network all nodes connect to central node
• Each message takes how many hops?
• Not failure tolerant
• Inexpensive sometimes used for LANs

43
Point-to-Point Network Topologies

• One-directional ring
• Given n nodes, max hops?
• Inexpensive
• Fault-tolerant?

44
Point-to-Point Network Topologies

• Bi-directional ring
• Given n nodes, max hops?
• Inexpensive
• Fault-tolerant? One node? Two?

45
Point-to-Point Network Topologies

• Doubly-connected ring nodes connected to
  neighbors one more distant
• Given n nodes, max hops?
• Fault-tolerant?
• More expensive