Create a dynamic datacenter with software-defined networking.

Create a dynamic datacenter with software-defined networking

• Agility in deploying new services• IT is frequently the bottleneck various Business Groups• How do you become the HERO!?

• Flexibility in deployment• Once a workload is deployed, it can be hard to move it around within the

datacenter and across clouds• How can you unshackle your workloads?

• Availability• Infrastructure issues frequently cause services to go down causing SLAs to get

missed• How can you provide an even better SLA on the same infrastructure!?

• Security• Attacks frequently gets in through one host (an unpatched one!), and then spreads

across the rest of the infrastructure.• How can you model your network to make this MUCH harder?

Common Customer Challenges

We faced the same challenges

At a much higher scale!

Fortune 500 using Azure

>57%

>50 TRILLION

storage objects

425MILLIONAAD users

1 out of 5 VMs

are Linux VMs

>5 MILLION

requests/sec

1 TRILLIONEvent Hubs

events/month

1,400,000SQL databases in Azure

Azure Scale Momentum

>18BILLIONAzure Active Directory authentications/week

New Azure customers a month

>90,000

So, how did we do it?

Start by finding the right abstractions

SDN: Building the right abstractions for ScaleAbstract by separating

management, control, and data planes Azure

Resource Manager

Controller

Switch (Host)

Management Plane

Control Plane

Management plane

Create a tenant ACL

Control plane Plumb these tenant ACLs to these switches

Data plane Apply these ACLs to these flows

Example: ACLs

• Data plane needs to apply per-flow policy to millions of VMs

• How do we apply billions of flow policy actions to packets?

• If every host performs all packet actions for its own VMs, scale is much more tractable

• Use a tiny bit of the distributed computing power of millions of servers to solve the SDN problem• If millions of hosts work to implement billions of flows, each host only

needs thousands

• Build the controller abstraction to push all SDN to the host

Solution: Host Networking

Virtual Networks on the Host• A Virtual Network is essentially a set

of mappings from a customer defined address space (CAs) to provider addresses (PAs) of hosts where VMs are located

• Separate the interface to specify a Virtual Network from the interface to plumb mappings to switches via a Network Controller

• All CA<-> PA mappings for a local VM reside on the VM’s host, and are applied there

Azure Frontend

Controller

Customer Config

VNet Description (CAs)

L3 Forwarding Policy(CAs <-> PAs)

VMSwitchVMSwitch

Blue VMsCA Space

Green VMs

CA Space

Northbound API

Southbound API

Hyper-V Hyper-V

Frontend

Controller

Node1: 10.1.1.5

Blue VM110.1.1.2

Green VM1

10.1.1.2

Azure VMSwitch

Node2: 10.1.1.6

Red VM110.1.1.2

Green VM2

10.1.1.3

Azure VMSwitch

Node3: 10.1.1.7

Green S2S GW10.1.2.1

Azure VMSwitch

Green Enterprise Network10.2/16

VPN GW

Customer Config

VNet Description

L3 Forwarding Policy

SecondaryControllers

ConsensusProtocol

Controllers

Forwarding Policy: Traffic to on-prem

Node1: 10.1.1.5

Blue VM110.1.1.2

Green VM1

10.1.1.2

Azure VMSwitchSrc:10.1.1.

2Dst:10.2.0.

9

Src:10.1.1.2

Dst:10.2.0.9

Policy lookup:10.2/16 routes toGW on host withPA 10.1.1.7

Controller

Src:10.1.1.5

Dst:10.1.1.7

GRE:GreenSrc:10.1.1.

2Dst:10.2.0.

9

L3 Forwarding Policy

Node3: 10.1.1.7

Green S2S GW

10.1.2.1

Azure VMSwitch

Green Enterpise Network10.2/16

VPN GW

Src:10.1.1.2

Dst:10.2.0.9

L3VPN PPP

Site-to-Site VPN• S2S connectivity to branch offices• Connecting Virtual Networks in other

Azure Sites• BGP for route updates• Transit routing for resiliency

Contoso HQ

Exchange

AD/DNS

IIS ServersSQL Farm

Monitoring

Contoso virtual networks/VMs

Internet

Services on public IPs

VPN Gateway(Internet Edge)

ExpressRoute

Microsoft

WAN

Corp HQ

Branch office 1

Branch office 2

Public internet

ExpressRoute provides a private, dedicated, high-throughput network

connection to Microsoft

Security

Lower cost

Predictable performance

High throughput

IaaS VM

• All infrastructure runs behind an LB to enable high availability and application scale

• How do we make application load balancing scale to the cloud?

• Challenges:• Load balancing the load balancers• Hardware LBs are expensive, and cannot

support the rapid creation/deletion of LB endpoints required in the cloud

• Support 10s of Gbps per cluster• Support a simple provisioning model

Cloud Load Balancing

Internet

LB

Web Server VM

Web Server VM

SQL Service

IaaS VM

SQL Service

NAT

• Goal of an LB: Map a Virtual IP (VIP) to a Dynamic IP (DIP) set of a cloud service

• SDN controller abstracts out LB/vswitch interactions

• Two steps: Load Balance (select a DIP) and NAT (translate VIP->DIP and ports)

• Pushing the NAT to the vswitch makes the LBs stateless (ECMP) and enables direct return

All-Software Load Balancer:Scale using the Hosts

LB MUX

VM DIP10.1.1.

2

VM DIP10.1.1.

3

Azure VMSwitch

StatelessTunnel

Edge Routers

Client

VIP

VIP

DIPDIP

DirectReturn:

VIP

VIP

LB MUX

VM DIP10.1.1.

4

VM DIP10.1.1.

5

Azure VMSwitch

NATController

Tenant Definition:VIPs, # DIPs

Mappings

NAT

Layered Security, Protection, and Isolation

DDoSProtection

Virtual Networ

kIsolatio

n

DMZ &NSG

VMFirewall

Cloud Services &

Virtual Machines InternetACLs

• Segment network to meet security needs

• 5 tuple ACLs on both directions

• Can protect Internet and internal traffic

• Enables DMZ subnets• Associated to

subnets/VMs and now NICs

• ACLs can be updated independent of VMs

Network Security Groups

Virtual Network

Backend10.3/16

Mid-tier10.2/16

Frontend10.1/16

VPN GW

Internet

On Premises 10.0/16

ExpressRouteand VPNs

√ √

√ √

• VMs that perform specific network functions • Focus: Security (Firewall, IDS , IPS), Router/VPN, ADC

(Application Delivery Controller), WAN Optimization

• Typically Linux or FreeBSD-based platforms• 1st and 3rd Party Appliances

• Azure Marketplace• Available through Azure Certified Program to ensure quality

and simplify deployment

Network Virtual Functions/Appliances

Network Virtual Appliance Ecosystem

• Hot-add / hot-remove vNIC• Add or remove a virtual NIC in a running Linux virtual

machine• Linux guest will add or remove the corresponding /dev

entry• New in Windows Server 2016 Hyper-V

• Network throughput• Implemented vRSS/vSSS and various TCP offloads• Instrumented and tightened code paths• All improvements are upstream in the main Linux

kernel

Linux Networking on Hyper-V

Hyper-V host #1 Hyper-V host #2

Linux guest (8 vCPUs)

Linux guest(8 vCPUs)

iperf3(16

threads)

10G Ethernet

iperf3(16

threads)9.4 Gbps throughput

Microsoft partners with distro vendors to get

improvements built-in. Feature grids in Linux

TechNet docs have version info.

WAP UI Portal

ARMAzure Resource

Manager

Authentication, Authorization, Role BasedAccess Control (RBAC), Template handling

CRPCompute Resource Provider

SRPStorage

Resource Provider

NRPNetwork Resource Provider

NCNetwork

Controller

xRP cache for resources

TemplateMicrosoft.Compute Resource 1 Resource 2Microsoft.Storage Resource 3Microsoft.Network Resource 4 Resource 5

Write Your App Once, Run Anywhere

Identical across Azure and other clouds

Azure Stack Demo

Storage needs networking as well as compute!

How do we make Azure Storage scale?

• Remote DMA primitives (e.g. Read address, Write address) implemented on-NIC• Zero Copy (NIC handles all transfers via DMA)• Zero CPU Utilization at 40Gbps (NIC handles all packetization)• <2μs E2E latency

• RoCEv2 enables Infiniband RDMA transport over IP/Ethernet network (all L3)• Enabled at 40GbE for Azure Storage, achieving massive COGS savings by

eliminating many CPUs in the rack

All the logic is in the host. Software Defined Storage now scales with the Software Defined

Network

RDMA: High Performance TransportNICNICMemory

Buffer A

Memory

Buffer B

Write local buffer at Address A to remotebuffer at Address B

Buffer B is filled

ApplicationApplication

Just so we’re clear…40Gbps of I/O with 0% CPU

It gets better – Converged Fabric

Network Performance MonitoringInterne

t

Fault domain (Subnet1)

Fault domain (Subnet2)

www.bing.com www.msn.com

Capabilities

Loss & latency monitoring (Intra/Inter/Subnet to Internet)

Impact assessment

Advanced algorithms

Network visibility

SCOM integration

Reduce MTTD and time to resolve issues through network performance monitoring, impact assessment, fault localization and health data

Infrastructure Demo

• Agility• You get a default gallery of applications (eg. SharePoint, Exchange, etc)• You get a self-service Portal that your customers/BGs use• Customer picks a workload and all the underlying requirements automatically get plumbed into the

infrastructure

• Flexibility• Workloads can get instantiated into overlays or virtual networks• You can move VMs around in the overlay without changing any IP addresses• You can move subnets to different clouds altogether and connect via gateways• Your apps can be written such that they work transparently in Azure or Azure Stack!

• Availability• Every component in the underlying system is designed to ensure it remains available• If one instance goes down, another one picks up

• Security• Use the Distributed Firewall, Network Security Groups, and Virtual Appliances for more dynamic security

• Any others? We want to hear from you!

Back to your challenges

• As a platforms company, we wanted to create a Cloud platform that would solve for real customer needs. We did – with Microsoft Azure!

• With Microsoft Azure we faced a lot of the same challenges you currently are.

• We did not want to take away choice from you. Instead of forcing you into Azure, we instead are bringing Azure to you as well! Now you get to choose where to put your workloads while benefiting from the years of innovation and more to come

• We would love to hear your use cases and challenges.

If you would like to take part in surveys that shape the future of Microsoft SDN. Please drop off a business card or send me contact info at grcusanz@microsoft.com.

Summary

© 2015 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.