基于LSTM的SDN流量预测与负载均衡python源代码+数据+详细注释

""" Mininet: A simple networking testbed for OpenFlow/SDN! author: Bob Lantz ([email protected]) author: Brandon Heller ([email protected]) Mininet creates scalable OpenFlow test networks by using process-based virtualization and network namespaces. Simulated hosts are created as processes in separate network namespaces. This allows a complete OpenFlow network to be simulated on top of a single Linux kernel. Each host has: A virtual console (pipes to a shell) A virtual interfaces (half of a veth pair) A parent shell (and possibly some child processes) in a namespace Hosts have a network interface which is configured via ifconfig/ip link/etc. This version supports both the kernel and user space datapaths from the OpenFlow reference implementation (openflowswitch.org) as well as OpenVSwitch (openvswitch.org.) In kernel datapath mode, the controller and switches are simply processes in the root namespace. Kernel OpenFlow datapaths are instantiated using dpctl(8), and are attached to the one side of a veth pair; the other side resides in the host namespace. In this mode, switch processes can simply connect to the controller via the loopback interface. In user datapath mode, the controller and switches can be full-service nodes that live in their own network namespaces and have management interfaces and IP addresses on a control network (e.g. 192.168.123.1, currently routed although it could be bridged.) In addition to a management interface, user mode switches also have several switch interfaces, halves of veth pairs whose other halves reside in the host nodes that the switches are connected to. Consistent, straightforward naming is important in order to easily identify hosts, switches and controllers, both from the CLI and from program code. Interfaces are named to make it easy to identify which interfaces belong to which node. The basic naming scheme is as follows: Host nodes are named h1-hN Switch nodes are named s1-sN Controller nodes are named c0-cN Interfaces are named {nodename}-eth0 .. {nodename}-ethN Note: If the network topology is created using mininet.topo, then node numbers are unique among hosts and switches (e.g. we have h1..hN and SN..SN+M) and also correspond to their default IP addresses of 10.x.y.z/8 where x.y.z is the base-256 representation of N for hN. This mapping allows easy determination of a node's IP address from its name, e.g. h1 -> 10.0.0.1, h257 -> 10.0.1.1. Note also that 10.0.0.1 can often be written as 10.1 for short, e.g. "ping 10.1" is equivalent to "ping 10.0.0.1". Currently we wrap the entire network in a 'mininet' object, which constructs a simulated network based on a network topology created using a topology object (e.g. LinearTopo) from mininet.topo or mininet.topolib, and a Controller which the switches will connect to. Several configuration options are provided for functions such as automatically setting MAC addresses, populating the ARP table, or even running a set of terminals to allow direct interaction with nodes. After the network is created, it can be started using start(), and a variety of useful tasks maybe performed, including basic connectivity and bandwidth tests and running the mininet CLI. Once the network is up and running, test code can easily get access to host and switch objects which can then be used for arbitrary experiments, typically involving running a series of commands on the hosts. After all desired tests or activities have been completed, the stop() method may be called to shut down the network. """ import os import re import select import signal import random import time from sys import exit # pylint: disable=redefined-builtin from time import sleep from itertools import chain, groupby from math import ceil from mininet.cli import CLI from mininet.log import info, error, debug, output, warn from mininet.node import ( Node, Host, OVSKernelSwitch, DefaultController, Controller ) from mininet.nodelib import NAT from mininet.link import Link, Intf from mininet.util import ( quietRun, fixLimits, numCores, ensureRoot, macColonHex, ipStr, ipParse, netParse, ipAdd, waitListening, BaseString ) from mininet.term import cleanUpScreens, makeTerms # Mininet version: should be consistent with README and LICENSE VERSION = "2.3.0" class Mininet( object ): "Network emulation with hosts spawned in network namespaces." # pylint: disable=too-many-arguments def __init__( self, topo=None, switch=OVSKernelSwitch, host=Host, controller=DefaultController, link=Link, intf=Intf, build=True, xterms=False, cleanup=False, ipBase='10.0.0.0/8', inNamespace=False, autoSetMacs=False, autoStaticArp=False, autoPinCpus=False, listenPort=None, waitConnected=False ): """Create Mininet object. topo: Topo (topology) object or None switch: default Switch class host: default Host class/constructor controller: default Controller class/constructor link: default Link class/constructor intf: default Intf class/constructor ipBase: base IP address for hosts, build: build now from topo? xterms: if build now, spawn xterms? cleanup: if build now, cleanup before creating? inNamespace: spawn switches and controller in net namespaces? autoSetMacs: set MAC addrs automatically like IP addresses? autoStaticArp: set all-pairs static MAC addrs? autoPinCpus: pin hosts to (real) cores (requires CPULimitedHost)? listenPort: base listening port to open; will be incremented for each additional switch in the net if inNamespace=False waitConnected: wait for switches to Connect? (False; True/None=wait indefinitely; time(s)=timed wait)""" self.topo = topo self.switch = switch self.host = host self.controller = controller self.link = link self.intf = intf self.ipBase = ipBase self.ipBaseNum, self.prefixLen = netParse( self.ipBase ) hostIP = ( 0xffffffff >> self.prefixLen ) & self.ipBaseNum # Start for address allocation self.nextIP = hostIP if hostIP > 0 else 1 self.inNamespace = inNamespace self.xterms = xterms self.cleanup = cleanup self.autoSetMacs = autoSetMacs self.autoStaticArp = autoStaticArp self.autoPinCpus = autoPinCpus self.numCores = numCores() self.nextCore = 0 # next core for pinning hosts to CPUs self.listenPort = listenPort self.waitConn = waitConnected self.hosts = [] self.switches = [] self.controllers = [] self.links = [] self.nameToNode = {} # name to Node (Host/Switch) objects self.terms = [] # list of spawned xterm processes Mininet.init() # Initialize Mininet if necessary self.built = False if topo and build: self.build() def waitConnected( self, timeout=None, delay=.5 ): """wait for each switch to connect to a controller timeout: time to wait, or None or True to wait indefinitely delay: seconds to sleep per iteration returns: True if all switches are connected""" info( '*** Waiting for switches to connect\n' ) time = 0.0 remaining = list( self.switches ) # False: 0s timeout; None: wait forever (preserve 2.2 behavior) if isinstance( timeout, bool ): timeout = None if timeout else 0 while True: for switch in tuple( remaining ): if switch.connected(): info( '%s ' % switch ) remaining.remove( switch ) if not remaining: info( '\n' )