Surprise Quiz EE384Z: McKeown, Prabhakar ”Your Worst Nightmares in Packet Switching Architectures”, 3 units [Total time = 15 mins, Marks: 15, Credit is.

Slides:



Advertisements
Similar presentations
1 Maintaining Packet Order in Two-Stage Switches Isaac Keslassy, Nick McKeown Stanford University.
Advertisements

Nick McKeown Spring 2012 Lecture 4 Parallelizing an OQ Switch EE384x Packet Switch Architectures.
Sundar Iyer Winter 2012 Lecture 8a Packet Buffers with Latency EE384 Packet Switch Architectures.
Optimal-Complexity Optical Router Hadas Kogan, Isaac Keslassy Technion (Israel)
Lecture 12. Emulating the Output Queue So far we have shown that it is possible to obtain the same throughput with input queueing as with output queueing.
Courtesy: Nick McKeown, Stanford 1 Intro to Quality of Service Tahir Azim.
Modeling the Interactions of Congestion Control and Switch Scheduling Alex Shpiner Joint work with Isaac Keslassy Faculty of Electrical Engineering Faculty.
Nick McKeown CS244 Lecture 6 Packet Switches. What you said The very premise of the paper was a bit of an eye- opener for me, for previously I had never.
Routers with a Single Stage of Buffering Sundar Iyer, Rui Zhang, Nick McKeown High Performance Networking Group, Stanford University,
Worst-case Fair Weighted Fair Queueing (WF²Q) by Jon C.R. Bennett & Hui Zhang Presented by Vitali Greenberg.
Algorithm Orals Algorithm Qualifying Examination Orals Achieving 100% Throughput in IQ/CIOQ Switches using Maximum Size and Maximal Matching Algorithms.
Making Parallel Packet Switches Practical Sundar Iyer, Nick McKeown Departments of Electrical Engineering & Computer Science,
Analysis of a Statistics Counter Architecture Devavrat Shah, Sundar Iyer, Balaji Prabhakar & Nick McKeown (devavrat, sundaes, balaji,
Generalized Processing Sharing (GPS) Is work conserving Is a fluid model Service Guarantee –GPS discipline can provide an end-to-end bounded- delay service.
1 Input Queued Switches: Cell Switching vs. Packet Switching Abtin Keshavarzian Joint work with Yashar Ganjali, Devavrat Shah Stanford University.
1 Comnet 2006 Communication Networks Recitation 5 Input Queuing Scheduling & Combined Switches.
Analyzing Single Buffered Routers Sundar Iyer, Rui Zhang, Nick McKeown (sundaes, rzhang, High Performance Networking Group Departments.
Analysis of a Packet Switch with Memories Running Slower than the Line Rate Sundar Iyer, Amr Awadallah, Nick McKeown Departments.
Using Load-Balancing To Build High-Performance Routers Isaac Keslassy, Shang-Tse (Da) Chuang, Nick McKeown Stanford University.
048866: Packet Switch Architectures Dr. Isaac Keslassy Electrical Engineering, Technion MSM.
CSIT560 by M. Hamdi 1 Course Exam: Review April 18/19 (in-Class)
048866: Packet Switch Architectures Dr. Isaac Keslassy Electrical Engineering, Technion The.
1 Internet Routers Stochastics Network Seminar February 22 nd 2002 Nick McKeown Professor of Electrical Engineering and Computer Science, Stanford University.
Nick McKeown 1 Memory for High Performance Internet Routers Micron February 12 th 2003 Nick McKeown Professor of Electrical Engineering and Computer Science,
1 EE384Y: Packet Switch Architectures Part II Load-balanced Switches Nick McKeown Professor of Electrical Engineering and Computer Science, Stanford University.
Fundamental Complexity of Optical Systems Hadas Kogan, Isaac Keslassy Technion (Israel)
Maximum Size Matchings & Input Queued Switches Sundar Iyer, Nick McKeown High Performance Networking Group, Stanford University,
COMP680E by M. Hamdi 1 Course Exam: Review April 17 (in-Class)
Ph. D Oral Examination Load-Balancing and Parallelism for the Internet Stanford University Ph.D. Oral Examination Tuesday, Feb 18 th 2003 Sundar Iyer
1 Achieving 100% throughput Where we are in the course… 1. Switch model 2. Uniform traffic  Technique: Uniform schedule (easy) 3. Non-uniform traffic,
1 Netcomm 2005 Communication Networks Recitation 5.
Analysis of a Memory Architecture for Fast Packet Buffers Sundar Iyer, Ramana Rao Kompella & Nick McKeown (sundaes,ramana, Departments.
048866: Packet Switch Architectures Dr. Isaac Keslassy Electrical Engineering, Technion Scheduling.
CS144, Stanford University Error in Q3-7. CS144, Stanford University Using longest prefix matching, the IP address will match which entry? a /8.
Pipelined Two Step Iterative Matching Algorithms for CIOQ Crossbar Switches Deng Pan and Yuanyuan Yang State University of New York, Stony Brook.
Localized Asynchronous Packet Scheduling for Buffered Crossbar Switches Deng Pan and Yuanyuan Yang State University of New York Stony Brook.
1 IP routers with memory that runs slower than the line rate Nick McKeown Assistant Professor of Electrical Engineering and Computer Science, Stanford.
Nick McKeown CS244 Lecture 7 Valiant Load Balancing.
Summary of switching theory Balaji Prabhakar Stanford University.
EE384y EE384Y: Packet Switch Architectures Part II Scaling Crossbar Switches Nick McKeown Professor of Electrical Engineering and Computer Science,
Designing Packet Buffers for Router Linecards Sundar Iyer, Ramana Kompella, Nick McKeown Reviewed by: Sarang Dharmapurikar.
Winter 2006EE384x1 EE384x: Packet Switch Architectures I Parallel Packet Buffers Nick McKeown Professor of Electrical Engineering and Computer Science,
Routers. These high-end, carrier-grade 7600 models process up to 30 million packets per second (pps).
Packet Forwarding. A router has several input/output lines. From an input line, it receives a packet. It will check the header of the packet to determine.
Nick McKeown1 Building Fast Packet Buffers From Slow Memory CIS Roundtable May 2002 Nick McKeown Professor of Electrical Engineering and Computer Science,
1 Performance Guarantees for Internet Routers ISL Affiliates Meeting April 4 th 2002 Nick McKeown Professor of Electrical Engineering and Computer Science,
An Introduction to Packet Switching Nick McKeown Assistant Professor of Electrical Engineering and Computer Science, Stanford University
Nick McKeown Spring 2012 Lecture 2,3 Output Queueing EE384x Packet Switch Architectures.
Winter 2006EE384x1 EE384x: Packet Switch Architectures I a) Delay Guarantees with Parallel Shared Memory b) Summary of Deterministic Analysis Nick McKeown.
Scheduling Determines which packet gets the resource. Enforces resource allocation to each flows. To be “Fair”, scheduling must: –Keep track of how many.
Techniques for Fast Packet Buffers Sundar Iyer, Ramana Rao, Nick McKeown (sundaes,ramana, Departments of Electrical Engineering & Computer.
048866: Packet Switch Architectures
Buffered Crossbars With Performance Guarantees Shang-Tse (Da) Chuang Cisco Systems EE384Y Thursday, April 27, 2006.
SNRC Meeting June 7 th, Crossbar Switch Scheduling Nick McKeown Professor of Electrical Engineering and Computer Science, Stanford University
1 How scalable is the capacity of (electronic) IP routers? Nick McKeown Professor of Electrical Engineering and Computer Science, Stanford University
Techniques for Fast Packet Buffers Sundar Iyer, Ramana Rao, Nick McKeown (sundaes,ramana, Departments of Electrical Engineering & Computer.
Scheduling algorithms for CIOQ switches Balaji Prabhakar.
The Fork-Join Router Nick McKeown Assistant Professor of Electrical Engineering and Computer Science, Stanford University
Techniques for Fast Packet Buffers Sundar Iyer, Nick McKeown Departments of Electrical Engineering & Computer Science, Stanford.
Tel Hai Academic College Department of Computer Science Prof. Reuven Aviv Markov Models for data flow In Computer Networks Resource: Fayez Gebali, Analysis.
1 Building big router from lots of little routers Nick McKeown Assistant Professor of Electrical Engineering and Computer Science, Stanford University.
scheduling for local-area networks”
EE384Y: Packet Switch Architectures Scaling Crossbar Switches
Packet Forwarding.
Parallelism in Network Systems Joint work with Sundar Iyer
EE384x: Packet Switch Architectures
Write about the funding Sundar Iyer, Amr Awadallah, Nick McKeown
Techniques and problems for
EE384Y: Packet Switch Architectures II
Techniques for Fast Packet Buffers
Presentation transcript:

Surprise Quiz EE384Z: McKeown, Prabhakar ”Your Worst Nightmares in Packet Switching Architectures”, 3 units [Total time = 15 mins, Marks: 15, Credit is given for short answers] 1.(5pts) Find the conditions under which a centralized shared memory switch built using multiple slower parallel memories can emulate an OQ switch 2.(5 pts) For a shared memory crossbar switch, find a sufficient bound on the speedup of the crossbar, to emulate an OQ switch which performs WFQ. 3.(5 pts) Show that if the traffic to any output is leaky bucket constrained, an IQ switch, can emulate an OQ switch within a delay bound, with a speedup of 2

Using Constraint Sets to Analyze IQ Switches Sundar Iyer, Nick McKeown (sundaes, Departments of Electrical Engineering & Computer Science, Stanford University

Outline 1.Introduction 2.FIFO Scheduling 3.PIFO Scheduling 4.Comparison to Charny’s Thesis

The Constraint Set (CS) Technique A technique to analyze single buffered routers 1.Determine packet’s departure time 2.Define the Constraint’s on the system for both inputs and outputs (if applicable) –Buffer, Fabrics, Speedup etc. 3.Apply the Pigeonhole principle Constraint Sets can be used to analyze Parallel Packet Switches (FIFO, PIFO) Shared Memory with Bus Shared Memory with Crossbar Input Queued Switches.. and we expect in general any single buffered router

Characteristics of an IQ switch Arriving packets are immediately written into the input queue, on the same port as that of the switch The packet is sent to the output at or after, the time of its “ideal” departure. We shall assume that the crossbar has a speedup of “s”, where “s” is the number of packets which can be sent from one input to an output in a cell slot. Buffers Inputs Outputs

The Leaky Bucket “(  )” Regulator Tokens at rate,  Token bucket size,  Packet buffer Packets One byte (or packet) per token “(  R )”

Some Definitions M is the Leaky Bucket Size –A property of the traffic pattern –This is not in our hands K is the Relative Delay –The amount by which we can delay the departure of a cell –A property of the switch –This is in our hands, we can tweak it

Outline 1.Introduction 2.FIFO Scheduling 3.PIFO Scheduling 4.Comparison to Charny’s Thesis

Allocations as seen by the Output … DT + kDT-KDT c k Packet has a FIFO Departure Time = DT Allocated Departure Time (ADT) in (DT, DT + k) In the interval (DT, DT + k) –There is one cell which tries to get allotted in that interval. –No more than k cells get delayed and are allotted to that interval Number of Time Slots Available >= [k – k /S] The past comes to haunt you ….

Allocations as seen by the Input … DT + kDT-M-KDT c M + k DT-M Packet has a FIFO Departure Time = DT Allocated Departure Time (ADT) in (DT, DT + k) In the interval (DT, DT + k) –There is one cell which tries to get allotted in that interval –No cell which arrived before DT–M-k will be allotted to this interval Number of Time Slots Available >= [k – (k+M)/S]

Sufficiency Conditions on Speedup We are guaranteed a timeslot if –[k- k/S] + [k – (k+M)/S] > K –S > 2 + M/k Thus we can prove that –S > 2 The IQ switch has 100% throughput, by setting k  very large –S > 3 The IQ switch can emulate a FIFO-OQ switch within M slots by setting k=M

Outline 1.Introduction 2.FIFO Scheduling 3.PIFO Scheduling 4.Comparison to Charny’s Thesis

PIFO Queues – Departure Order Timeline of departures Arrival Order, Cell 8 arrives first, Cell 4 arrives last “The cell number is the name of a cell. In this figure it also represents the final departure order of these cells”

What is the problem with PIFO? 1.The CS Technique depends on being able to predict the departure time and schedule it. –The departure time of a cell is not fixed in PIFO 2.The departure time of a cell can increase –Hence, at the input we can have very old cells and cannot bound the number of cells. 3.How do we solve this problem? We shall schedule cells based on their initial departure time

PIFO Queues – Initial Departure Time (IDT) Scheduling Timeline for Departures Arrival Order, Cell 8 arrives first, Cell 4 arrives last IDT for cells

Extreme Case for IDT – Backlogged Queues Arrival Order, Cell 8 arrives first, Cell 4 arrives last Scheduling Timeline for Departures IDT for cells are shown in red

Some Properties Lemma: 1.(Weak) “If an output has a bucket size of M then no more than M cells are allotted the same initial departure time ”. 2.(Strong) If an output has a bucket size of M, then no more than M+a cells are allotted an IDT in a time interval of size a Proof: –Consider a cell arriving at time = t. –Since, the bucket size is M, it’s initial departure time is within (t, t +M) –Hence proved.

past Allocations as seen by the Output Packet has a PIFO Initial Departure Time = PT Allocated Departure Time (ADT) in (PT, PT + k) In the interval (PT, PT + k) When a cell arrives at the switch, there are no more than k + M cells waiting in the switch for that output. Unlike FIFO, future allocations can interfere by pushing in to the interval Number of Time Slots Available >= [k – (k +M)/S] … PT + k PT c future M + k

Allocations as seen by the Input … PT + kPT-M-KPT c M + k PT-M Packet has a PIFO Initial Departure Time = PT Allocated Departure Time (ADT) in (PT, PT + k) In the interval (PT, PT + k) –There is one cell which tries to get allotted in that interval –No cell which arrived before PT–M-k will be allotted to this interval Number of Time Slots Available >= [k – (k+M)/S]

Sufficiency Conditions on Speedup We are guaranteed a timeslot if –[k- (k+M)/S] + [k – (k+M)/S] > K –S > 2 + 2M/k The IQ switch has the following properties –with S > 2, has 100% throughput, even with PIFO based scheduling, set k  very large. –Can emulate a PIFO-OQ switch with with S > 3, and a relative delay of 2M with S > 4, and a relative delay of M

Outline 1.Introduction 2.FIFO Scheduling 3.PIFO Scheduling 4.Comparison to Charny’s Thesis

Anna Done it… Theorems –(Weak) If S > 4, then any maximal matching policy will give 100% throughput –(Strong) If S > 2, then any maximal matching policy will give 100% throughput. –(Stronger) If S > 2, then there is bounded emulation of a FIFO-OQ switch.

Comparison FIFO –Charny’s analysis and CS are similar Anna’s done the FIFO analysis first!  The proof using CS is much simpler though QoS –Charny does not analyze QoS scheduling –Charny uses rate controlled inputs, with FIFO scheduling to give QoS –The CS Technique directly analyzes PIFO –Note: Charny’s proofs can be modified to analyze PIFO PIRO –No one has thought of working on this before. –But in fact, PIRO can be done with CS, with LIFO allocation of departure times.

Comparison Summary Switch Algorithm –Charny’s Work: Pros: –Works for any maximal algorithm Cons: –But the matching has to be calculated –QoS not analyzed directly –CS Technique: Cons: –Works for a specific algorithm only Pros: –The crossbar scheduling is automatic –FIFO, PIFO, PIRO can be analyzed directly for IQ switches

References Anna Charny, “Providing QoS guarantees in Input Buffered Crossbar Switches with Speedup”, Sep Internal References: –Papers on Constraint Sets –PPS papers, Shared Memory Paper