VirtualizationAbdullah Aldhamin
June 3, 2013
2
Outline Virtualization: what and why?
Server Virtualization Offerings
Shortcoming and Challenges on SV
Sustaining Performance while Virtualizing
Future Perspective on SV
Data Center Network Virtualization, what and why?
Overview of VN Projects and how they compare with each others
Future Perspective on DC Network Virtualization
Conclusions
3
Virtualization Virtualization
Techniques, methods, approaches to create a virtual instance of a computing resource.
Why? Advances in h/w technologies, e.g. multi-core
H/W optimization Cloud computing Economic factors
4
Server Virtualization Key features:
Multiple Isolated
5
Server Virtualization: Offerings Improved H/W utilization
Consolidation VMs are just files
Faster provisioning and deployment Better availability and BC solutions
6
Server Virtualization: Offerings Lower TCO
Longer h/w purchase cycle Smaller datacenter Lower maintenance, power, cooling … etc
7
Server Virtualization: Challenges Performance:
CPU and Network Overhead to do virtual-related emulations A single NIC = shared bandwidth Not good for I/O bound apps Propagation delay latency
Storage Latency Needs enough physical storage power to support the
multiple VMs Memory
Amount needed for the application
8
Server Virtualization: Challenges Scalability
How many VMs can each physical server accommodate?
Asset management Tracking a dynamic environment Service management burdens
Security Malicious access to the hypervisor means you’re
screwed up! Single-point of failure?
9
Server Virtualization and Performance:Can we have both?
10
Virtualization and Performance Selective virtualization Direct assignment of physical NICs to VMs
Isolation and and better I/O− Cost for more NICs and cabling− Flexibility constraints
11
Virtualization and Performance Firmware-based I/O virtualization
Virtual I/O channels connected to the same NIC No true isolation, a channel will impact all other
channels H/W-based I/O virtualization
I/O channels are built in the H/W- Vendor-specific NICs
12
Virtualization and Performance Consolidation in a shared resource pool
Allocate VMs to a resource pool instead of physical server
Dynamic resource allocation to VMs based on workload
Complexity and poor visibility on how performance is managed
Adds challenges to asset management
13
Virtualization and Performance Management tools
Help manage performance, track SLAs and set priorities
Optimize resource allocation Proactive failure assessment
Advances in processors and memory Multiple logical CPUs and improved instructions
faster comm. between hypervisor and VMs Integrate I/O memory mgmt and DMA remapping
better I/O virtualization
14
Highlights on SV Trends Open-source tools Joint collaboration between server virtualization
vendors and processors’ manufacturers Hardware-based hypervisor Convergence of server, storage, desktop and
apps virtualization.
15
So… Server virtualization Pros and Cons How can we overcome (or minimize) the impact
on performance Future perspectives
16
Is this enough? Does it solve everything we want
from Virtualization?
17
Server Virtualization is NOT enough! No performance isolations Security Poor application deployability Limited management flexibility No support network innovation
18
Data Center Network Virtualization Multiple virtual networks (VNs) on top of physical
network. Challenges
Addressing schemes Scalability Failure tolerance Security Resource management
19
Examples of DC Network Topologies Conventional DC network topology
20
Examples of DC Network Topologies Clos topology
Path diversity
21
Examples of DC Network Topologies Fat-tree topology
K=4
22
Network Basic Functionalities/Features Packet forwarding schemes Bandwidth guarantees Relative bandwidth sharing Multipathing techniques
23
Examples of Virtual NetworkProposals/Implementations
24
Traditional Data Center Objective: isolation between tenants Pros:
Uses VLANs to provide means of isolations Uses commodity switches and popular hypervisors SPs can have their own L2 & L3 address spaces
Cons: Imposes scalability concerns
25
Diverter Objective: to provide software-based VN solution
for packet forwarding L3 network virtualization to allow tenants control
their addresses Software module VNET on every physical machine
Replaces VM MAC with physical MAC Special IP scheme (10.tenant.subnet.vm)
Lacking QoS
26
NetLord Objective: virtual network designed to provide
solution for scalability of tenants, in order to increase resource utilization and revenue
Utilizes L2 and L3 virtualization: L2+L3 encapsulation Full address-space virtualization
Uses Diverter forwarding Uses SPAIN multipathing
27
NetLord
28
NetLord Pros:
Takes advantage of already implemented schemes to provide basic functionalities
Cons Not all commercial off-the-shelf switches support IP
forwarding Large packet encapsulation results in more drops
and fragmentation No bandwidth guarantee
29
Oktopus Issues with multi-tenant data center networks:
Difficult application performance management Unpredictable network performance decrease app.
performance unhappy customers revenue lo$$e$ Objectives: to design virtual network abstractions and
to explore the trade-off between the guarantees offered to tenants, the tenant cost and provider revenue
Key design concept: two virtual network abstractions: Virtual cluster Virtual oversubscribed cluster
30
Oktopus Cluster
31
Oktopus Virtual cluster
32
Oktopus Virtual oversubscribed cluster
33
Oktopus Pros:
Increases application performance Flexibility to InP Balance between higher application performance
and lower cost Cons
Deployability: only for tree-like physical network topologies
34
Gatekeeper Objectives: a virtual network to provide bandwidth
guarantees, and achieves high bandwidth utilization Design features:
Scalable in terms of number of VMs Predictable in terms of network performance Flexible based on minimum and maximum bandwidth
guarantees Design elements:
Minimum guaranteed rate Maximum allowed rate for each VM
35
Gatekeeper
36
Gatekeeper Pros:
Flexible bandwidth guarantees Cons:
Other performance features are not addressed Still under development Scale of experimental evaluation is small
37
Seawall Objective: a virtual network to provide bandwidth
sharing in a multi-tenant data center network
38
Seawall Pros:
Policy enforcement so no one tenant can consume all bandwidth
Dynamic nature allows for tenant requirement change adjustment
Cons: Deployability: currently for Windows and Hyper-V
39
ProjectFeature
Forwarding Scheme
Bandwidth Guarantee
Multipathing Relative BW Sharing
Traditional DC ✔ ✔
Diverter ✔NetLord ✔ ✔VICTOR ✔
VL2 ✔ ✔PortLand ✔ ✔
SEC2 ✔ ✔SPAIN ✔
Oktopus ✔SecondNet ✔ ✔Gatekeeper ✔CloudNaaS ✔ ✔
Seawall ✔NetShare ✔ ✔
Classification of Projects by Feature
40
Highlights on Research Direction Virtualized Edge Data Center: concerned about placement of
small data centers at the edge of the network to be closer to end-users
Benefits: Better QoS for delay sensitive applications Reduces network communication cost by reducing network traffic
across network providers Lower construction cost compared to large remote data centers
Challenges: Service placement problem: finding optimal trade-off between
performance and construction cost How to efficiently manage services hosted in multiple data centers?
41
Highlights on Research Direction Virtual data center embedding algorithms: to
accommodate a high number of VDCs in data centers Dependent on how efficient virtual resources are
mapped to physical ones Challenges:
Need to consider all physical resources, in addition to servers
Dynamic nature of the demand for data center applications
Energy efficiency: how to optimize the placement of VMs and VNs for energy efficiency?
42
Highlights on Research Direction Network performance guarantees:
The design and implementation for bandwidth allocation schemes
Find a good trade-off between maximizing network utilization and guaranteed network performance
Data center management: Monitoring Energy management Failure detection and handling
43
Highlights on Research Direction Security:
Mitigating security vulnerabilities Monitoring and auditing infrastructure Support for multi-layer security depending on
tenants needs
44
Conclusions Server virtualization: what and why? Techniques to address performance issues related
to server virtualization Datacenter network virtualization: what and why? Recent virtual network architectures: no ideal
project to solve all problems!
Future perspectives on server and network virtualization research: we still have a lot to do…
45
Thank You!
46
References Fauzi Kamoun, “Virtualizing the Datacenter Without Compromising Server
Performance”, ACM Ubiquity, Vol 2009, Issue 9. Md. Faizul Mari, et. al., “Data Center Network Virtualization: A Survey”, in IEEE
Communications Surveys & Tutorials, Vol 15, pg. 909-928, Sep 2012. [online] http://en.wikipedia.org/virtualization “Virtualization”, May 2013. T. Benson, et. al., “CloudNaaS: A Cloud Networking Platform for Enterprise
Applications”, SOCC’11, 2011. A. Edwards, et.al., “Diverter: A New Approach to Networking Within Virtualized
Infrastructures”, WREN’09, 2009. J. Mudigonda, et. al., “SPAIN:COTS Data-Center Ethernet for Multipathing over Arbitrary
Topologies,” in Proceedings ACM USENIX NSDI, April 2010. J. Mudigonda, et. al., “NetLord: A Scalable Multi-Tenant Network Architecture for
Virtualized Datacen- ters,” in Proceedings ACM SIGCOMM, August 2011. F. Hao, et. al., “Enhancing Dynamic Cloud-based Services using Network Virtualization,”
in Proceedings ACM VISA, August 2009.