三网合一优于单网

Click to edit Master title style,Click to edit Master text styles,Second Level,Third Level,Fourth Level,Fifth Level,Traffic,management,An Engineering Approach to Computer Networking,11/14/96,1,Multimedia network communications requirements,For video,sustained bandwidth of at least 64 kbps,low loss rate,For voice,sustained bandwidth of at least 8 kbps,low loss rate,For interactive communication,low delay ( S/a, u0,reflects that user feels worse off than no such transfer if time taken too long for the transfer.,S,and,a,can be experimentally determined,More Examples,For video or audio transfer, we may set utility function for the transfer delay as,u = if (t,d = 1.25 C,as a result,u = 4-1.25C, v=8-2.5C, sum of utilities,u + v = 12-3.75 C,If Bs delay reduced to,0.5C, according to the law, then As delay =,2C,Sum,of utilities,= 12 - 3C.,which is larger than the former,12-3.75 C,.,increased the social,welware,Increase in social welfare need not benefit everyone,A loses utility, but may pay less for service by the network manager providing lower price for it.,Some economic principles,A single network that provides heterogeneous,QoS,is better than separate networks for each,QoS,unused capacity is available to others,三网合一优于单网,Lowering delay of delay-sensitive traffic increased welfare,can increase welfare by matching service menu to user requirements,BUT need to know what users want (signaling),For typical utility functions, welfare increases more than linearly with increase in capacity,individual users see smaller overall fluctuations,generally, we can increase welfare by increasing capacity,Principles applied,A single wire that carries both voice and data is more efficient than separate wires for voice and data,ADSL,IP Phone,Moving from a 20% loaded10 Mbps Ethernet to a 20% loaded 100 Mbps Ethernet will still improve social welfare,increase capacity whenever possible,Better to give 5% of the traffic lower delay than all traffic low delay,should somehow mark and isolate low-delay traffic,The two camps,Can increase welfare either by,matching services to user requirements,or,increasing capacity blindly,Which is cheaper?,no one is really sure!,small and smart vs. big and dumb,It seems that smarter ought to be better,otherwise, to get low delays for some traffic, we need to give,all traffic,low delay, even if it doesnt need it,But, perhaps, we can use the money spent on traffic management to increase capacity,Anyway, we will study traffic management, assuming that it matters!,Traffic models,traffic model tries to summarize the expected behavior of users or aggregates of users,e.g. how long a user uses a modem,e.g. average size of a file transfer,e.g. aggregate: how are the % of text and % of file in a chat process,Models change with network usage,now, more bandwidth available makes BT available,We can only guess about the future,Two ways to obtain models,measurements,maybe over millions of connections over years,mathematical analysis in an amendable way,but very few can reach this,Telephone traffic models,Call arrival model,How are calls placed?,studies show that time between calls is drawn from an exponential distribution,call arrival process is therefore,a Poisson Process,memoryless,process: the fact that a certain amount of time has passed since the last call gives no information of time to next call,Call holding-time model,How long are calls held?,usually modeled as exponential,however, measurement studies show it to be,heavy tailed,means that a significant number of calls last a very long time,Internet traffic modeling,A few common apps account for most of the traffic,WWW,email,FTP or other file transfer app,telnet, now web service based more,changes along with time,A common approach is to model apps with a combination of 4 parameters,time between app invocations,connection duration,# bytes transferred,packet,interarrival,distribution in a connection,Little consensus on models of common apps,some think this, some that,But two important features,Internet traffic models: features,LAN connections differ from WAN connections,Higher bandwidth (more bytes/call),longer holding times,also changes now! BT makes very long holding time,Many parameters are heavy-tailed,examples,# bytes in connection,connection duration,means that a,few,connections are responsible for most of the traffic,these connections must be well-managed,New models appear all the time, to account for rapidly changing traffic mix,Outline,Economic principles,Traffic classes,Time scales of TM,TM Mechanisms,Some open problems,Traffic classes,Try to categorize various applications into a set of widely used applications application classes,Try to categorize various services of networks into a set of network services service classes,Try to merge the app and service classes into one type of classes - Traffic Classes,Traffic classes - details,A basic division:,guaranteed service (GS),and,best effort (BE),like flying with reservation or standby,Guaranteed-service,utility is zero unless app gets a minimum level of service quality,typical parameters,bandwidth, delay, loss,open-loop flow control with admission control,e.g. telephony, remote sensing, interactive multiplayer games,Best-effort,send and pray,closed-loop flow control,e.g. many found on the Internet today, email, FTP, WWW, etc,GS vs. BE (cont.),Degree of synchrony,time scale at which peer endpoints interact,GS are typically,synchronous,or,interactive,interact on the timescale of a round trip time,e.g. telephone conversation, telnet, or game,BE are typically,asynchronous,or,non-interactive,interact on longer time scales,e.g. Email,Sensitivity to time and delay,GS apps are,real-time,performance depends on wall clock,BE apps are typically indifferent to real time,elastic,Traffic subclasses,ATM Forum,based on sensitivity to bandwidth,GS,CBR, VBR,BE,ABR, UBR,IETF,based on sensitivity to delay,GS,intolerant,tolerant,BE,interactive burst,interactive bulk,asynchronous bulk,ATM Forum GS subclasses,Constant Bit Rate (CBR),constant, cell-smooth traffic,mean and peak rate are the same,e.g. telephone call evenly sampled and uncompressed,model telephone service but more flexible,constant bandwidth, variable quality (not just * 64kbps),Variable Bit Rate (VBR),long term average with occasional bursts,try to minimize delay,can tolerate loss and higher delays than CBR,e.g. compressed video or audio where bit rates vary with the degree of achievable compression,ATM Forum BE subclasses,Available Bit Rate (ABR),users get whatever is available,zero loss if TM signals sent from network to user are obeyed,no guarantee on delay or bandwidth,Unspecified Bit Rate (UBR),like ABR, but no feedback,no guarantee on loss,presumably cheaper,closest to the current service on Internet,IETF GS subclasses,based on apps sensitivity to delay,Tolerant GS,requires a nominal mean delay, but can tolerate “occasional” violation,not specified how often and how much the violation is,uses,controlled-load,service,even at “high loads”, admission control assures a source that its service “does not suffer”,imprecise!,Intolerant GS,requires a worst case delay bound,equivalent to CBR+VBR in ATM Forum model,Both implicitly require either a constant or variable bit-rate b/w, and low loss,Note, IETF has defined this, but does not mean the Internet MUST take or even HAVE taken this,in fact, current Internet is still very hard to take any,GSs, even the,GSs,are lower than ATM CBR specification.,IETF BE subclasses,Still based on the delay sensitivity,Interactive burst,bounded asynchronous service, where bound is qualitative, but pretty tight (at least at human time scales),e.g. paging, messaging, email,Interactive bulk,bulk, but a human is waiting for the result,e.g. FTP,Asynchronous bulk,junk traffic,e.g Usenet( Internet news),Some notable points,The only thing out there is CBR and asynchronous bulk!,telephone network is CBR.,the Internet today is still asynchronous bulk in general.,These are application requirements. There are also organizational requirements,like link sharing,Users needs,QoS,for other things than b/w, delay, and loss too!,billing,privacy,reliability and availability,Outline,Economic principles,Traffic classes,Time scales,Mechanisms,Some open problems,Time scales of TM,Some actions are taken once per call,tell network about traffic characterization and request resources - signaling,in ATM networks, finding a path from source to destination,Other actions are taken during the call, every few round trip times,feedback flow control,Still others are taken very rapidly,during the data transfer,scheduling,policing and regulation,Traffic management mechanisms must deal with a range of traffic classes at a range of time scales,Summary of mechanisms at each time scale,Less than one round-trip-time (cell or packet - level),Scheduling and buffer management,Regulation and policing,Policy routing (datagram networks),always open-loop,One or more round-trip-times (burst-level),Feedback flow control in BE,Retransmission in BE,Renegotiation in GS,closed-loop,Summary (cont.),Session (call-level),Signaling for GS,Admission control for GS,Service pricing,Routing in connection-oriented networks,but connectionless networks do not recognize multiple RTT and session time scale,Day,Peak load pricing,Weeks or months, or even longer,Capacity planning,increase the capacity,Signaling,Signaling,the process by which a user requests the network to set up, tear down, or renegotiate a call (connection).,can be understood as how a source tells the network its utility function,include two parts,how to carry the signaling messages reliably (signaling transport),how to interpret the messages (signaling semantics),Useful to separate these mechanisms,make it possible to change signaling meaning without changing the transportation method.,Signaling semantics,Classic scheme: sender-initiated signaling,now also have receiver-initiated signaling,typical signaling messages,SETUP, SETUP_ACK, SETUP_RESPONSE,Steps,source sends SETUP,switches (or routers) in the path either accept or reject the setup,by use of admission control mechanism,when accepted, resources are reserved for GS,in ATM, VPI or VCI is assigned.,telling source by sending back the SETUP_ACK,destination either accept or reject the setup,by sending back the SETUP_RESPONSE,the ensure the transmission, an ACK is also needed for the response message,Simplex or duplex setup,reserve resources for one way or duplex ways,ATM signaling specifies duplex, IETF signaling simplex,Only a basic model, doesnt work for multicast, and also individual signaling protocols differ from this greatly.,Resource translation,easy for bandwidth, but much hard for delay descriptors,Application asks for end-to-end quality, but how to translate to per-hop requirements?,E.g. end-to-delay bound of 100 ms,What should be bound at each hop?,Two-pass,forward pass: each switches maximize and tells the destination,reverse pass: relax, destination tells,still open problem anyway!,Signaling: transport,Telephone network uses Signaling System 7 (SS7),Carried on Common Channel Interoffice Signaling (CCIS) network,CCIS is a datagram network, logically distinct from the circuit-switch network that is for voice calls.,SS7 protocol stack is loosely modeled on ISO (but predates it),Signaling in ATM networks uses Q.2931 standard,part of User Network Interface (UNI),complex,layered over SSCOP ( a reliable transport protocol) and AAL5,Internet (IETF) signaling transport : RSVP (RFC 2205),Main motivation is to efficiently support multipoint multicast with resource reservations,easy for,unicast,also,Progression,Unicast,signaling messages: s-d-s,Nave multicast,Intelligent multicast,Nave multipoint multicast,RSVP,RSVP motivation,Multicast reservation styles,Nave multicast signaling (source initiated),source contacts each receiver in turn,s-d1-s, s-d2-s,wasted signaling messages,Intelligent multicast signaling (merge replies),SETUP,msg,is duplicated to all,dests,and responses are merged.,source needs to know all receivers in advance,doesnt scale,Nave multipoint multicast signaling,a multicast group, users in the group may be sender and receiver,like nave multicast signaling,still source initiated,RSVP,Two features applied,Receiver initiated,Reservation state per group, instead of per connection,PATH and RESV messages,PATH sets up next hop towards source(s),but does not make reservations,multicast the PATH,sent from source,RESV makes reservation,sent from receivers,go back to source along the path specified by PATH,includes resources requests of the receiver,every node reserves resources for this receiver,node forwards RESV,msg,only the,msg,requires larger than that already reserved,many other issues related,Why is signaling hard ?,Complex services,Feature interaction,call screening + call forwarding,Tradeoff between performance and reliability,Extensibility and maintainability,Outline,Economic principles,Traffic classes,Mechanisms at each time scale,Faster than one RTT,One RTT,Session,Signaling,Admission control,Day,Weeks to months,Some open problems,Admission control,Admission control,Admission control,Whether admitting a call (connection),Signaling,Carry signaling messages and make resource reservations for the call,Admission control comes first, signaling comes secondly,CBR Admission control,Simple,Call i of CBR can be described by a single parameter, the rate,r(i,),A link with Capacity C has admitted L to calls, then,a new call i can be admitted, L+r(i)=C,may also have delay bounds,on failure: try again,reroute(try,using another path), or hold till available,BE Admission control,Best-effort admission control,usually not reject any calls,may assign different priority levels depending on the BE service priority,e.g., a scheduler may has 3 BE priority levels individually for email, FTP, and Usenet,If BE call requires some min. b/w, switch controller may performs a test like a CBR admission control,VBR admission control,VBR,peak rate differs from average rate =,burstiness,if we reserve bandwidth at the peak rate, wastes bandwidth,if we reserve at the average rate, may drop packets during peak,key decision: how much to overbook,Four known approaches,peak rate admission control,worst-case admission control,admission control with statistical guarantees,measurement-based admission control,1.,Peak-rate admission control,Reserve at a connections peak rate,Pros,simple (can use FIFO scheduling),connections get zero delay and zero loss,works well for a small number of sources,Cons,wastes bandwidth,peak rate may increase because of scheduling jitter,time,rate,2.,Worst-case admission control,Characterize source by average rate and burst size (LBAP),Use WFQ or rate-controlled discipline to reserve bandwidth at average rate,Pros,may use less bandwidth than with peak rate,can get an end-to-end delay guarantee (WFQ),Cons,for low delay bound, need to reserve at more than peak rate! (WFQ),implementation complexity,3.,Admission with statistical guarantees,Key insight is that as number of calls increases, probability that multiple sources send a burst decreases,sum of connection rates is increasingly smooth,With enough sources, traffic from each source can be assumed to arrive at its average rate,E.g., 10 identical sources with rate 1.0 and probability 0.1 on.,Then, the probability of n sources ON is:,4.,Measurement-based admission,For traffic that cannot describe itself,e.g., renegotiated traffic, the descriptors change unpredictably,Key: measure,real average load,Users are described by peak, average, and LBAP, are sure of the peak but not very sure of others,switch measures the average rate of existing calls,If a new call, and its peak + measured average capacity, admit,Over time, new call becomes part of average,Problems:,assumes that past behavior is indicative of the future,how to measure,how long to measure?,when to forget about the past?,Outline,Economic principles,Traffic classes,Mechanisms at each time scale,Faster than one RTT,One RTT,Session,Day,Weeks to months,Some open problems,Peak load pricing,Problems with cyclic demand,Service providers want to,avoid overload,use all available capacity,Hard to do both with cyclic demand,if capacity C1, then waste capacity,if capacity C2, overloaded part of the time,Peak load pricing,Traffic shows strong cyclical behavior at the time scale of a day and at a