Software-Defined Networks (SDN): Bridging the application-network divide

Transcription

1 Software-Defined Networks (SDN): Bridging the application-network divide Inder Monga Chief Technologist and Area Lead, Engineering, Research and Software development CHEP 2013

2 Shared historical moment

3 Thank you! HEP community helped create ESnet 27 years ago

4 Global Transfer Activities (LHC/ATLAS)

5 Clev. 27 years and 5 generations later. World s first G Science Network Int l peerings Int l peerings JGI SUNN SNLL LLNL PNNL Capability to scale to 13.2 Tbps AMES 1 BNL PPPL GFDL PU Physics Int l peerings Int l peerings 1 SNLA Geographical representation is only approximate SF Bay Area Chicago New York Amsterdam SUNN STAR AofA AMST G testbed

6 Back to the future First workshop report for ESnet on intersite networking, 1986 Network community is still struggling to meet application requirements captured in 1986! Brute force approach (add more bandwidth) is not going to meet those requirements

7 What is common between modern networks and analog phone switches?

8 Labor-intensive, nearly static, error prone

9 What is SDN? Loose definition: separation of data-plane from control plane In essence: enables programmability programmable Network Provisioning Network Monitoring HEP Applications HEP Applications Network Apps Network Virtualization Network Controller(OS) Statistics Topology Provisioning Protocols (SNMP, TL1) Protocol(s) (OpenFlow,?) Control Software Firmware Network ASICs Network Element control control control Control Firmware control Software control Firmware Firmware Network Firmware Network Firmware ASICs Network ASICs Network ASICs Network ASICs ASICs Network Element

10 Simple programming constructs OpenFlow 1.0 standard Rule Action Stats Packet + byte counters 1. Forward packet to zero or more ports 2. Encapsulate and forward to controller 3. Send to normal processing pipeline 4. Modify Fields 5. Any extensions you add! Switch Port VLAN ID VLAN pcp MAC src MAC dst Eth type IP Src IP Dst IP ToS IP Prot L4 sport L4 dport + mask what fields to match /17/2013 Inder Monga, WLCG GDB Slide courtesy Srini Seetharaman

11 OpenFlow usage Classic model Controller OpenFlow Alice s Rule Switch Alice s App PC Decision? OpenFlow Protocol OpenFlow Alice s Rule Switch OpenFlow Alice s Rule Switch OpenFlow offloads control intelligence to remote software Alice /17/2013 Inder Monga, WLCG GDB 11

12 Layer 3 Layer 2 Layer 1 What is the paradigm difference? Internet today: - Built-in control in each layer - Multiple management domains SDN Approach: - Network-wide cross-layer view - Unclear how management evolves Layer 3 Control? Control (Network-wide view) Layer 3 Control OpenFlow OpenFlow Layer 2 Layer 1 Control Control Management Management /16/13 Inder Monga, Layer123 SDN 12

13 Claim #1: Programmability will lead to greater predictability Burst movement of data using PhEDex Analysis triggered data movement (PD2P) HEP increasingly needs to deal with high performance, any-any bursts of data SDN enables multi-layer control packet and optical layer Control over individual flows ex. Route science flows around packet bottlenecks Many NRENs have access to fiber, optical and packet platforms.

14 ESnet leading development of SDN control for packet-optical demonstration /16 OTS Config Manager L0/L1 Topology Advanced Reservation System (OSCARS) Multi-Layer Topology App Multi-Layer Path Engine Multi-Layer Provisioning Multi-Layer SDN Control Layer Traffic Optimization Engine SDN Controller Floodlight OpenFlow & REST/JSON OpenFlow Host A Host B OTS Virtualization WDM/ OTN/ Packet Infinera DTN-X Presented at SDN World Congress In Bad Homburg

15 SDN is about network abstraction Complexity of the Internet infrastructure black box : It s strength and weakness Image by Matt Britt (used with permission under Creative Commons Attribution 2.5 )

16 Constructing simplicity through abstractions App 1 App n NB API Modeled as a Virtual WAN Network Element Network Virtualization simple complex Network slice Network Controller(OS)

17 Claim #2: Virtualization will simplify how HEP applications program the network Going from the Monarc model to full multi-point mesh would be as simple as changing the multiple point to point link abstraction with a virtual switch abstraction Tier Tier 0 0 Tier 1 Tier Tier 1 1 Tier 2 Tier 2 Virtual Switch Virtual Switch policy Tier 2 Virtual Switch Tier 2 Tier 2 Tier 2 Tier 2 Tier 2 Tier 2 Tier 2 Tier 2 Tier 2 Tier 2 Tier 2 The complexity is absorbed by software hypervisor of the underlying multi-domain network

18 SDN is about system optimization Google s B4 SDN Network Utilization When the application and network work as a system, network resource optimization is possible Without knowledge of flows, networking can only do coarse characterization Fine discrimination of flows possible with SDN, meet application needs

19 Claim #3: SDN enables an opportunistic way to leverage all bandwidth without extra investment exploits the fact In general it s much cheaper to transport data than to store it, for example, enables diskless Tier3

20 We are at the starting line! SDN concepts have made tremendous leaps in academia and industry There are three major leaps that still need to be made: Operationalization of the SDN model Manageability (for failures, failure recovery) and Scalability Security Next Challenge is to tackle the gaps with open-source toolsets So a small network (not size of Google) can enjoy the potential benefits

21 Network organizations need to tackle similar software challenges Layer Layer 8-9 People People (network engineers**, sysadmins, operators**) (network engineers, + sysadmins, operators) (software engineers/devops) Network Operating System (control) Management, Tools, + Measurement New tools, service plane and management Living with inefficient software is much more expensive - Torre Wenaus Keynote Layer 0-7 Network (control (API + and data data plane) plane) ** need to develop new skills

22 A fun peek into the future just imagine With silicon photonics integration, each chip will have a network interface That implies each chip could be network addressable If so, we could design servers without needing NIC cards no difference between communication within the motherboard or outside. With HEP applications like FAX, file systems or memory can be mounted remotely to my chip while streaming data for analysis. With SDN, can effectively route IP and non-ip protocols (like ROCE) SDN could revolutionize how computing is done, are we ready for that?

23 Thank you!

24 The wall must go! Researchers & Applications The Network Scientific Productivity Connections