1. Supporting Multimedia Communication over a Gigabit Ethernet Network VARUN PIUS RODRIGUES

2. Need for Gigabit Networks  Streaming Multimedia  High performance distributed computing  Virtual Reality  Distance learning

3. Development and Design Challenges  Hardware Components:  Gigabit Network Interface Card  A buffered hub  Gigabit routing switch  File Server within LAN: Building pilot workgroups within LAN  Integration of gigabit switches  Design Challenges:  Simplicity  End-to-end solutions  Extended to include emerging multimedia applications

4. Design Features of Gigabit NIC  802.3 Frame compatibility  Design challenge of the MAC ASIC to operate the GNIC  Reducing host CPU utilization through Descriptor-based DMA

5. IEEE 802.3z Standard Standard for MAC and PHY layers  PHY Layer:  Fiber: 1000 BASE-SX (multi-mode) and 1000 BASE-LX (single-mode)  Copper-based: 1000 BASE-CX (twinax cable)  MAC Layer:  Identical to one defined for 10 Mbps and 100Mbps Ethernet  Fields: DA, SA, LEN, DATA, FCS

6. Designing GNIC  Architecture Consideration: 64-bit 66 MHz GNIC-II  Theoretical bandwidth: 4 Gbps vs 1 Gbps  Practical bandwidth: 3 Gbps vs 800 Mbps  Consists of:  Application-specific Integrated Circuit (ASIC)  Packet buffer memory  Serializer/Deserializer chip  Physical layer components

7. ASIC  Consists of:  PCI  Pair of DMA for controllers: for Rx and Tx  Pair of FIFO connected to external FIFO interface  GNIC  Reducing CPU utilization:  Accesses host memory directly through Descriptor-based DMA  Transfer Chaining: transferring arbitrary nos of packets from host memory to GNIC  Adaptability of interrupt rate of host to network load

8. Design Features of Buffered Gigabit Hub  Full-duplex: for eliminating CSMA/CD collision  Congestion Control: to avoid frame dropping  Round-Robin scheduling: to prevent “packet clumping”

9. Performance issues with CSMA/CD  Performance highly dependent on ratio of propagation delay to average packet transmission time  Two ways to solve it:  Increase minimum packet length  Decrease length  Length increased through virtual collisions

10. Full Duplex Repeater  Achieves switching and shared design concepts for switch-like performance while maintaining the cost of shared hub  Provides maximum throughput, collisionless forwarding, and congestion control  Logical flow of frames: Input -> Forwarding path -> Output  Input: Passes though buffer PHY and MAC before being queued in buffer; congestion control informs end system to slow down incase capacity of buffer is about to be reached  Forwarding path: Implements round robin scheduling to determine which port will send data  Output: includes a buffer, MAC and PHY with congestion notification

11. Design Features of Gigabit Routing Switch  Architecture issues  Parallel access shared memory architecture  Priority Queue Design

12. Architecture of Gigabit Routing Switch PE-4884  Consists of 12 channel cards, 1 EMM card for the chassis to function and 1 EMM card management, policy and routing table redundancy  Channel cards: Supports connectivity to Ethernet, fast Ethernet, FDDI etc  Sending and receiving data through physical interfaces and system packet memory  Performing routing and switching address lookups  Enforcing layer 4 policies; collecting management statistics, etc  Every channel card connected to central memory through 2 full-duplex gigabit channels Contd…

13. Architecture of Gigabit Routing Switch PE-4884  Memory Architecture:  Uses parallel memory architecture for high speed performance  Limitations of cross-bar architecture:  Port-based memory  Head-of-line blocking  Difficult to provide QoS support  Shared memory bus architecture overcomes these limitations  Packet flow:  Address Resolution Logic ASIC on channel card evaluates the destination address  ARL signals memory control cards that a packet must be sent to which port  Frame Management

14. Supporting Distributed Multimedia Applications  Challenges in On-Demand Video:  Large data size  Real-time constraint  Supporting concurrent accesses  Consideration for connection setup:  Implementation of RSVP scheme  New emerging standards such as IEEE 802.1p and 802.1Q  Integrated solutions with policy-based QoS

15. Experimental Results  Two major goals:  To test how a gigabit LAN performed in the basic metrics of throughput  To examine how concurrent video delivery is supported over gigabit LAN  Experiments designed to determine bottlenecks in current system and identify bounds on performance  End-to-end performance is evaluated

16. Experimental Setup  Hardware setup:  Pentium II 233 Mhz and Pentium Pro 300 Mhz  Running either Linux or NT  Benchmarking utilities: netperf (for max throughput) and netbench (for avg throughput)  Netperf:  Consists of 2 process: netserver and netperf  Netperf connects to remote system running an instance of netserver and uses control connection to send parameters  TCP connection using BSD sockets  Separate connection for measurement Contd..

17. Experimental Setup  Netbench:  Measures how well a file server handles I/O requests  Each client tallies how many data moves to and from server

18. Results  Maximum throughput results:  Message packet size varied from 2048 bytes to 4 Mbytes.  After 16KByte packet size is reached, there is slight drop in performance and then again increase to give peak throughput of 190Mbps  Performance peak in Linux system; NT system peak throughput around 90Mbps  To investigate, NT experiment performed on different machine; peak throughput of 180 Mbps attained  SUN’s Solaris attained peak performance of 488 Mbps  Raw device testing attained performance of 700-800 Mbps at hardware level  Average throughput results:  Maximum server throughput of 157 Mbps with 3 clients

19. High Quality Streaming Videos Criteria for experiment to measure concurrent access for on-demand video:  Buffering Scheme:  Used 2 buffer scheme at server: one to retrieve video frames from server system and other to transmit it to client  Client also uses 2 buffer scheme: one for the network and other to display the frame  Performance metrics for the measurement:  Need to determine maximum number of concurrent accesses that can be supported by the network  Need to calculate jitters (number of miss deadline retrievals)

20. Results: On-Demand Video Streaming  Range specifications of MPEG-2 between 4 and 32 Mbps were emulated  Number of active processes were increased until system could no longer provide acceptable QoS (jitter less than 1%)  Due to bus contention, 128 Mbps achieved for 32 streams of 4 Mbps videos  Buffer size is another performance bottleneck in addition to network throughput

21. Questions?  You may send me a mail on my UF mail in case of any issues you need to communicate with me  UF mail: varunpius@ufl.eduvarunpius@ufl.edu