Keynote -金耀辉--network service in open stack cloud-osap2012_jinyh_v4

Network Service in
OpenStack Cloud

Yaohui Jin
email: jinyh@sjtu.edu.cn
Sina Weibo: @bright_jin
(The slides will be shared in Sina Weipan & Slideshare )

Network & Information Center

© jinyh@sjtu

Acknowledgement

 Team: Dr. Xuan Luo, Pengfei Zhang, Xiaosheng Zuo,
Zhixing Xu, Xinyu Xu, Jianwen Wei, Baoqing Huang, etc.
 Prof. Hongfang Yu and team with UESTC
 Prof. Jianping Wang with CityU HK
 Engineers, discussion and slides from Intel, SINA, IBM,
Cisco, Dell, VMware/EMC, H3C, Huawei, IXIA, …
 OpenStack Community
 China OpenStack User Group (COSUG)
 China OpenStack Cloud League (COSCL)
 Technical blogs such as blog.ioshints.info, ipspace.net, …

© jinyh@sjtu 2

About me

 上海交通大学教授，以前做光通信的，现在改行
做云计算了。。。 
 上海交通大学网络信息中心副主任，其实就是
个苦逼的挨踢网管啊。。。 
 研究兴趣：数据中心网络，海量流式数据分析，
云计算架构

© jinyh@sjtu 3

OpenStack in Academia
for Research & Operation
 USC, Information Science Institute
 Purdue University
 University of Melbourne
 San Diego Supercomputer Center
 Brookhaven National Lab., DOE
 Argonne National Lab., DOE
 European Organization for Nuclear Research (CERN)
 Shanghai Jiao Tong University
 University of Science & Technology of China
 University of Electrical Science & Technology of China
 ……
© jinyh@sjtu 4

Agenda

 Introduction
 SDN and OpenFlow
 Network Virtualization
 Network Virtualization in OpenStack
 Our Work

© jinyh@sjtu 5

The Service Trend

 "Decoupling infrastructure management from service
management can lead to innovation, new business
models, and a reduction in the complexity of running
services. It is happening in the world of computing, and
is poised to happen in networking.“
Jennifer Rexford
Professor, Princeton University

 Last month, VMware paid $1.2B
to acquire Nicira for software
defined networking (SDN).

© jinyh@sjtu 6

Why is Nicira worth $1.2 billion?

© jinyh@sjtu 7

SDN and OpenFlow

© jinyh@sjtu

Software Defined Network (SDN)

 A network architecture in which the network control
plane (OS) is decoupled from the physical topology
using open protocols such as OpenFlow.

© jinyh@sjtu 9

Flow Table (v1.1)

 Rules: Ethernet, IP, MPLS, TCP/UDP any combination,
exact or wildcard
 Actions: Forward, Drop, Modify field (NAT)
 Statistics: Volume based billing, anti DDOS

© jinyh@sjtu 10

OpenFlow Implementation

 Hypervisor Mode
 Open vSwitch (OVS): XEN, KVM, …
 OVS other features: security, visibility, QoS,
automated control
 Hardware Mode
 OpenFlow Switch
 Hop by hop configuration

© jinyh@sjtu 11

Reality Check

 “OpenFlow doesn’t let you do anything you couldn’t do on a
network before” –Scott Shenker (Professor, UC Berkeley,
OpenFlow co-inventor)
 Frames are still forwarded, packets are delivered to hosts.
 OpenFlow 1.3 was recently approved.
 Major vendors are participating - Cisco, Juniper, Brocade,
Huawei, Ericsson, etc. It’s still early stage technology but
commercial products are shipping.
 OpenFlow led by large companies Google/Yahoo/Verizon
and lack of focus on practical applications in the enterprise.

© jinyh@sjtu 12

OpenFlow Interop

 Fifteen Vendors Demonstrate OpenFlow Switches at
Interop (May 8-12, 2011)

© jinyh@sjtu 13

Network Virtualization

© jinyh@sjtu

General Data Center Architecture

Cloud management system allows us dynamically provisioning VMs and virtual storage.

© jinyh@sjtu 15

What customers really want?

Virtual Network

 Requirements
 Multiple logical segments
 Multi-tie applications
 Load balancing and firewalling
 Unlimited scalability and mobility

© jinyh@sjtu 16

Multi-Tenant Isolation
 Making life easier for the cloud provider
 Customer VMs attached to “random” L3 subnets
 VM IP addresses allocated by the IaaS provider
 Predefined configurations or user-controlled firewalls
 Autonomous tenant address space
 Both MAC and IP addresses could overlap between two tenants, or
even within the same tenant
 Each overlapping address space needs a separate segment

© jinyh@sjtu 17

Scalability

 Datacenter networks have got much bigger (and getting bigger still !!)
 Juniper’s Qfabric ~6000 ports, Cisco’s FabricPath over 10k ports
 Tenant number dramatically increase as the IaaS experiences rapid
commoditization
 Forrester Research forecasts that public cloud today globally valued at
$2.9B, projected to grow to $5.85B by 2015.
 Server virtualization increase demand on switch MAC address tables
 Physical with 2 MACs -> 100 VMs with 2 vNIC need 200+ MACs!

© jinyh@sjtu 18

Possible Solutions (1)

 VLANs per tenant
 limitations of VLAN-id range (Only 12bits ID = 4K)
 VLAN trunk is manually configured
 Spanning tree limits the size of the network
 L2 over L2
 vCDNI(VMware), Provider Bridging(Q-in-Q)
 Limitations in number of users (limited by VLAN-id range)
 Proliferation of VM MAC addresses in switches in the network
(requiring larger table sizes in switches)
 Switches must support use of same MAC address in multiple
VLANs (independent VLAN learning)

© jinyh@sjtu 19

Possible Solutions (2): L2 over IP

 Virtual eXtensible LAN (VXLAN)
 VMware, Arista, Broadcom, Cisco, Citrix, Red Hat
 VXLAN Network Identifier (VNI): 24 bits = 16M
 UDP encapsulation, new protocol
 Network Virtualization Generic Routing Encapsulation
(NVGRE)
 Microsoft, Arista, Intel, Dell, HP, Broadcom, Emulex
 Virtual Subnet Identifier (VSID): 24 bits = 16M
 GRE tunneling, relies on existing protocol
 Stateless Transport Tunneling (STT)
 Nicira
 Context ID: 64 bits, TCP-like encapsulation

© jinyh@sjtu 20

VXLAN/NVGRE: How it Works?

without
overlay

using VXLAN

using NVGRE

© jinyh@sjtu 21

Dynamic MAC learning

 Dynamic MAC learning with L2 flooding over IP multicasting

Flooding does not scale when fabric gets bigger.

© jinyh@sjtu 22

Control Plane (Nicira)

 L2-over-IP with control plane
 OpenFlow-capable vSwitches
 IP tunnels (GRE, STT ...)
 MAC-to-IP mappings by OpenFlow
 Third-party physical devices
 Benefits
 No reliance on flooding
 No IP multicast in the core

© jinyh@sjtu 23

Transitional Strategy
Depends on Your Business
 100s tenants, 100s servers: VLANs
 1000s tenants, 100s servers: vCDNI or Q-in-Q
 Few 1000s servers, many tenants: VXLAN/NVGRE/STT
 More than that: L2 over IP with control plane

Open question: How to solve the co-existing scenarios in one cloud?
© jinyh@sjtu 24

Network Virtualization in
Openstack

© jinyh@sjtu

OpenStack Today

 Networking is embedded inside of Nova compute, and
un-accessible to application developers
 Details and differences associated with network
provisioning complicates a simple compute service
 Difficult to track changes in networking as Software-
defined Networking (SDN) comes into play

© jinyh@sjtu 26

With Quantum –
Networking becomes a Service
 Nova becomes simpler, easier to maintain and extend
 Developers have ability to create multiple networks for
their own purposes (multi-tier apps)
 May support provisioning of both virtual and physical
networks – differences captured through plugin’s

© jinyh@sjtu 27

Quantum API interactions

© jinyh@sjtu 28

Plug-in’s available today

 Open vSwitch
 Linux bridge
 Nicira NVP
 Cisco (Nexus switches and UCS VM-FEX)
 NTT Labs Ryu OpenFlow controller
 NEC OpenFlow
 Big Switch Floodlight

© jinyh@sjtu 29

Quantum in Horizon

 Create/delete private network
 Create “ports” and attach VM’s
 Assign IP address blocks (DHCP)

© jinyh@sjtu 30

Network Environment
 Data Center Network: 10 GE Switch (BNT&H3C) in 2 domains
 Control and Manage: GE Switch (DCRS)
 10GE connect to campus network
 Fat tree topology; L3: VRRP;
 L2: LACP+VLAG+MSTP
 Security control: SSH, NAT, ACL, VLAN
 NIC: Intel X520-DA2; Chelsio T420E-CR

© jinyh@sjtu 36

Transition:Co-existing VLAN/GRE

 VLAN solution: Openstack + Open vSwitch

 GRE solution: Openstack + Ryu

© jinyh@sjtu 37
43

QoS in Virtual Network

 Bandwidth upper bound for VMs
 With only OVS : 200Mbit/s
 With OVS and virtio: 8Gbit/s
 Bandwidth guarantee with Openstack + OVS
 User defined rate limitation
 Differential service level for tenants
 High bandwidth utilization
 Stable performance under dynamic traffic

© jinyh@sjtu 38

Visible Virtual Network by sFlow

Virtual Physical Virtual Physical
Machine Server Switch Switch

CPU Unicast

Disk Multicast

Port
Traffic
Traffic

© jinyh@sjtu
© jinyh@sjtu 39
45

Keynote -金耀辉--network service in open stack cloud-osap2012_jinyh_v4

More Related Content

What's hot (17)

Viewers also liked (20)

Similar to Keynote -金耀辉--network service in open stack cloud-osap2012_jinyh_v4 (20)

More from OpenCity Community (20)

Recently uploaded (20)

Keynote -金耀辉--network service in open stack cloud-osap2012_jinyh_v4