Transcript SDN Intelligent Networking

SDN INTELLIGENT
NETWORKING
IMPLICATIONS FOR END-TO-END
INTERNETWORKING
Simone Mangiante
Senior research scientist
EMC Centre of Excellence, Ireland
© Copyright 2015 EMC Corporation. All rights reserved.
HEPTech, Budapest, 30/03/2015
1
THE EMC FEDERATION OF COMPANIES
SOFTWARE-DEFINED DATA CENTER
ADVANCED
SECURITY
SOFTWARE-DEFINED ENTERPRISE
SOFTWARE-DEFINED STORAGE
© Copyright 2015 EMC Corporation. All rights reserved.
2
COMPLETING THE PORTFOLIO
End User Computing
Next Generation Cloud & Big / Fast Data Apps
Content, Process & Collaboration Apps
The Software-Defined Data Center
Information Infrastructure
© Copyright 2015 EMC Corporation. All rights reserved.
3
EMC EUROPE RESEARCH FOCUS
SW Defined
Infrastructure
Co-develop
new technologies
Security
Analytics for
Infrastructure/
Security
Identify new
trends
© Copyright 2015 EMC Corporation. All rights reserved.
Solve difficult
problems
Leverage
Worldwide
Innovation
Network
4
ACTIVE EU PROJECTS
Project
Main Partners
Focus Area
SOLAS
Data mobility & SDDC Architectures
SPECS
Security SLAs for CSPs
SPARKS
Security Analytics for Smart Grids
NEAT
SDN for IoT and Big Data Apps
ESCUDO
Data Privacy & Security
SAFEcrypto
Key Management
© Copyright 2015 EMC Corporation. All rights reserved.
5
THE SOFTWARE DEFINED DATA CENTRE(S)
Virtualization Layer - Cloud Operating System
Leverage Intelligence
COMPUTE
© Copyright 2015 EMC Corporation. All rights reserved.
STORAGE
NETWORK
6
INDUSTRY TRENDS
• Industry trending to IP networking
• Storage networks, inter-site connections moving to IP
• Workload mobility between datacenters
• IP networking enables virtualization
• Virtualization shifts workloads between sites
• Increased competing demands on limited WAN
resources
• Explosion of wireless/mobile IP devices
• Poor or unreliable performance
• Expensive solutions
© Copyright 2015 EMC Corporation. All rights reserved.
7
PROBLEM: TRADITIONAL DATACENTER
NETWORKS
• There is an increasing volume of data traffic between data
centers
• Growing need to transfer large datasets and manage them
efficiently
• Data transfers are slow and costly
• Static network resource allocation
data
© Copyright 2015 EMC Corporation. All rights reserved.
data
8
USE CASES NEW AND OLD
• Big Data Analytics
• vCDN’s
• Backup
• Replication
• Dynamic Workload Migration
• Hybrid Cloud
© Copyright 2015 EMC Corporation. All rights reserved.
9
EMC KEY DISTANCE PRODUCTS
• VPLEX
• SRDF
• RecoverPoint
Hybrid Cloud
© Copyright 2015 EMC Corporation. All rights reserved.
10
A NEW APPROACH TO NETWORKING
Network virtualization is all about abstracting network services
and operations from the underlying hardware.
This moves network configuration decisions to a central location
and enables you to manage your network at the speed of software
Configuration is done with knowledge of the entire network state
© Copyright 2015 EMC Corporation. All rights reserved.
11
IMAGINE A NETWORK THAT REACTS TO YOUR
APPLICATION…
Network Demand!
Leveraging SDN we can allow Storage Functions to signal changes in
bandwidth demand
© Copyright 2015 EMC Corporation. All rights reserved.
12
VIEW OF DYNAMIC BANDWIDTH ALLOCATION
Bandwidth
New VPLEX
task started
Additional
bandwidth
requested
Additional
bandwidth
removed
Path Capacity
Available
Capacity
Reserved
Reserved
Reserved
Reserved
Reserved
for
for
for
for
for
VPLEX
SRDF
RecoverPoint
VMware Clustering
Other Network Traffic
Time
© Copyright 2015 EMC Corporation. All rights reserved.
13
PROOF OF CONCEPT ARCHITECTURE
Data Center A
NVP Controller
Data Center B
10G
Network
`
Service Node
NVP Logical Switch
NVP Gateway
NVP Gateway
WAN
40KM
Packet
Switch
Optical
Switch
Optical
Switch
Packet
Switch
VPlex
VPlex
SAN
SAN
VPLEX Distributed logical volume
VPLEX Stretched Cluster
VNX
© Copyright 2015 EMC Corporation. All rights reserved.
ESXi Cluster
ESXi Cluster
VNX
14
MULTIPLE APPROACHES – PHASE II
HOW THE NETWORK REACTS TO DEMAND: PATH CREATION
Multiple connections between
multiple sites
Single path utilized in
normal operation
VPLEX
© Copyright 2015 EMC Corporation. All rights reserved.
When demand increases diverse
separate paths are added, flows are
split between paths
VPLEX
VPLEX
Increased
Demand!
VPLEX
15
VIEW OF DYNAMIC BANDWIDTH ALLOCATION
USING MULTIPLE PATHS – PHASE II
New VPLEX
task started
Bandwidth
Separate
paths
removed
Separate
paths
requested
Path 3
Reserved for VPLEX
Path 2
Reserved for VPLEX
Path 1
Reserved
Reserved
Reserved
Reserved
Reserved
for
for
for
for
for
VPLEX
SRDF
RecoverPoint
VMware Clustering
Other Network Traffic
Time
© Copyright 2015 EMC Corporation. All rights reserved.
16
CURRENT WORK
• Software Defined Network (SDN) solution that allows
– Dynamic end-to-end allocation of resources
– Awareness of both network resources and application
demand
– Hardware isolation of each flow in the network
• SDN controller configures and manages the different
network paths
© Copyright 2015 EMC Corporation. All rights reserved.
17
THREE NETWORKED DATA CENTERS
• Multiple Physical Networks
• High Capacity Fiber
Data Center
• Networking Platform for:
Evaluating optical platforms,
topologies, protocols
– Next Generation connectivity,
Dark Fiber Cloud
– Engineering, Tech Support &
Advanced customer solutions
–
UCC
IP Fabric
Data Center
© Copyright 2015 EMC Corporation. All rights reserved.
Data Center
18
SCENARIOS
• Management of large and long data flows:
–
–
–
–
Data replication
Big Data
Video streaming
Research experiments
• Isolated paths and guaranteed bandwidth for specific
QoS requirements
• Internet of Things
© Copyright 2015 EMC Corporation. All rights reserved.
19
CONNECTIVITY SERVICES FOR STORAGE
• Discovery, Dynamic Provisioning, Pool management
• Path configuration
• Scheduling of n/work resources
• Indicating available network capacity (to help e.g. a video
application decide whether to switch its encoding).
• Congestion control, avoidance and reporting
• Security requirements, preferred network interface, etc.
• All at the right levels of abstraction!
© Copyright 2015 EMC Corporation. All rights reserved.
20
FURTHER WORK - RESEARCH CHALLENGES
• Traffic classification  traffic engineering
– Classify packets into high and low priority queues/flows
• Management of storage related network traffic
• Traffic prediction: track and analyze flows
• Fault handling in SDN network
• Scalability
• Policy
• Scheduling
© Copyright 2015 EMC Corporation. All rights reserved.
21
For further questions feel free to contact:
 [email protected]