Une très intéressante présentation autour de la virtualisation des réseaux contenant des explications détaillées autour des VLAN, VXLAN, mais aussi d'NVGRE et surtout de GENEVE (Generic Network Virtualization Encapsulation) supporté pour la première fois sur la dernière carte 40 GbE d'Intel (XL710)
Boost Fertility New Invention Ups Success Rates.pdf
Virtualizing the Network to enable a Software Defined Infrastructure (SDI)
1. Virtualizing the Network to enable a
Software Defined Infrastructure (SDI)
Brian Johnson – Solution Architect, Intel Corporation
Jim Pinkerton – Windows Server Architect, Microsoft
DATS002
2. 2
•Transforming The Network For The Cloud
•Accelerating Network Virtualization Overlays
•Next Generation Servers With Integrated Ethernet
Agenda
3. 3
•Transforming The Network For The Cloud
•Accelerating Network Virtualization Overlays
•Next Generation Servers With Integrated Ethernet
Agenda
13. 13
Developing New Technologies for the Virtualized NetworkDelivering Network Optimizations for Intel® Xeon® processor E5-2600 v3 Based Servers
Networking infrastructure needs to address business and infrastructure requirements
Network Functions Virtualization
Optimized small packet fast-paths with SR-IOV and Intel® Data Plane Development Kit
Network Virtualization OverlaysHardware assisted acceleration of VXLAN overlays for multi-core servers
Software-Defined NetworkingProgrammatic traffic steering withIntel® Ethernet Flow Director
Network Functions Virtualization
VM1
VM2
VM3
High Volume Servers
Dedicated appliances
Network Virtualization
Physical network
SDN Controller
Trends and Challenges
Intel®Ethernet Solutions
Reducing Service Deployment from 6 weeks to minutes…
14. 14
Intel® Xeon® processorE5-2600 v3 Family
Intel® Data Directed I/O makes theprocessor cache the primary destination and source of I/O data rather than main memory
Intel® Ingredients for Workload Optimization
Storage
Intel® Solid State DriveDC P3700 Family
PCI Express* brings extreme data throughput directly toIntel Xeon processors
Intel® Ethernet Controller XL710 Family40GbE & 10GbE connectivity for Enterprise, Cloud and Communications
Intel Ethernet ConvergedNetwork Adapter XL710 / X710 Family
Intel® QuickAssistTechnologyOffloads packet processing technology thereby reserving processor cycles for application and control processing
Intel QuickAssist Adapter 8950-SCCP
Intel® Solid-State Drive DC P3700 Series Family
Intel® Communications Chipset 89xx
Intel® C610 Series Chipset
Chipset
Network
Acceleration
Software
Intel® Data Plane Development KitPacket Processing Software create the foundation for NFV / SDN, server virtualization and vSwitchoptimizations
Compute
15. 15
Intel® Ethernet Controller XL710
Intel® Ethernet Controller XL710 Technical Details
SMBus
NC-SI
2x40GbE or 4x10GbE/1GbE
PCI Express 3.0 x8/x4/x1
MCTP
VF0
VF1
VFn
VF127
In-band
Mgmt
PF0
PF1
PF2
PF3
PCI Express 3.0 x8 –SR-IOV
Queue Mgmt, Scheduler
Protocol Acceleration / Offloads
Q1536
Q0
Q1
Q2
Q3
Qn
VEB, DCB Traffic Classifier
2x40GbE or 4x10GbE MAC
40GbE: KR4/XLAUI/CR4/XLPPI
10GbE: KR/SFI/XAUI/KX4
1GbE: KX/SGMII
Broad Offering of Physical Interfaces
•Low typical power at 3.8W for 2x40GbE single chip design for PCI Express* 3.0 x8
•Software configurable Ethernet Port Speed for up to 2x40GbE or up to 4x10GbE
•Interfaces for Converged Network Adapters, backplanes and LAN on Motherboard
Server I/O Virtualization assistants and by-pass
•VMDq for VMware*Netqueue* and Microsoft DVMQ*
•SR-IOV (Single Root I/O Virtualization), VEB (Virtual Ethernet Bridge)
•Edge Virtual Bridging / 802.1Qbg
Network Virtualization Overlay Accelerators and Offloads
•Abstract the network for cloud flexibility with performant network overlays
•Support for standard and custom network headers
•NVGRE, IPinGRE, VXLAN, MACinUDP, GENEVE
Advanced Hardware Traffic Steering
•Intel® Ethernet Flow Director –8000 perfect match filters stored on die
•User configurable to direct specific flows to targeted CPU optimizing cache utilization
•1536 queues / Physical Function (PF), 64 RSS / PF and 256 VMDq/ PF
Converged Networking
•Simplifying deployments by consolidating LAN, SAN (FCoE, iSCSI)
•Intelligent offloads optimized to accelerate software initiators
•Reduce infrastructure and cabling costs
16. 16
•Transforming The Network For The Cloud
•Accelerating Network Virtualization Overlays
•Next Generation Servers With Integrated Ethernet
Agenda
17. 17
Network Virtualization: Abstracts Physical Network
Server Virtualization
Hypervisor
Virtual Switch
PhysicalHardware
Network Virtualization
PhysicalIP Network
Virtual Network Abstraction using tunnel overlays e.g., VXLAN, Geneve and NVGRE
Open Virtual Switch
Open Virtual Switch
Open Virtual Switch
Open Virtual Switch
Network Virtualization Controller using VMware* NSX
Virtual Network 2
Virtual Network 3
Virtual Network 1
35. 35
Network Virtualization Assists and Offloads
NVGRE Encapsulated Task Offloads
•Large Send Offload (LSO)
•Checksum Tasks
•Virtual Machine Queue (VMQ)
CustomerAddress
ProviderAddress
VSID
192.168.10.20
192.168.10.60
10.0.0.5
10.0.0.7
MAC
GRE Key 5001
192.168.10.20
192.168.10.60
10.0.0.5
10.0.0.7
VXLAN NI(VNI) 5001
Outer UDP Header
CustomerAddress
VTEPAddress
VNI
NVGRE
VXLAN
VXLAN Encapsulated Offloads
•Large Send Offload (LSO)
•Checksum Tasks
•Receive Side Scaling (RSS)
Encapsulation and Decapsulationof packets is performed by the hypervisor and virtual switch in conjunction with the network adapter
36. 36
What is Unique between Hosts when using NVGRE?
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Optional
Outer
802.1Q
Outer Dest MAC
Outer Source MAC
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
RSVD
Protocol
type
VSID
FCS
Flow ID
NVGRE Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
GRE header
8 bytes
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
Optional Outer 802.1Q
EtherType
37. 37
What is Unique between Hosts when using NVGRE?
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Optional
Outer
802.1Q
Outer Dest MAC
Outer Source MAC
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
RSVD
Protocol
type
VSID
FCS
Flow ID
NVGRE Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
GRE header
8 bytes
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
Layer 2
Optional Outer 802.1Q
EtherType
38. 38
What is Unique between Hosts when using NVGRE?
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Optional
Outer
802.1Q
Outer Dest MAC
Outer Source MAC
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
RSVD
Protocol
type
VSID
FCS
Flow ID
NVGRE Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
GRE header
8 bytes
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
192.168.100.20
192.168.100.10
Layer 2
Layer 3
Optional Outer 802.1Q
EtherType
39. 39
What is Unique between Hosts when using NVGRE?
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Optional
Outer
802.1Q
Outer Dest MAC
Outer Source MAC
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
RSVD
Protocol
type
VSID
FCS
Flow ID
NVGRE Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
GRE header
8 bytes
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
192.168.100.20
192.168.100.10
5001
5002
ca:f1:ea:bc:51:4b
3a:50:3c:94:c9:45
d6:b3:69:8c:d7:462a:e4:d2:12:bd:46
Layer 2
Layer 3
Unique
Optional Outer 802.1Q
EtherType
40. 40
What is Unique between Hosts when using NVGRE?
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Optional
Outer
802.1Q
Outer Dest MAC
Outer Source MAC
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
RSVD
Protocol
type
VSID
FCS
Flow ID
NVGRE Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
GRE header
8 bytes
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
192.168.100.20
192.168.100.10
5001
5002
ca:f1:ea:bc:51:4b
3a:50:3c:94:c9:45
d6:b3:69:8c:d7:462a:e4:d2:12:bd:46
Layer 2
Layer 3
Intel® Ethernet Converged Network Adapter XL710
Intel Ethernet Converged Network Adapter X710
Unique
Optional Outer 802.1Q
EtherType
41. 41
Receive Side Scaling for VXLAN
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Outer Dest MAC
Outer Source MAC
Optional VXLAN Type
Optional Outer 802.1Q
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
Source Port
Dest Port (8472)
UDP Length
UDP Check Sum
VXLAN Flags
RSVD
VXLAN NI (VNI)
FCS
RSVD
VXLAN Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
Outer UDP Header
8 bytes
VXLAN Header
8 bytes
EtherType
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
42. 42
Receive Side Scaling for VXLAN
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Outer Dest MAC
Outer Source MAC
Optional VXLAN Type
Optional Outer 802.1Q
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
Source Port
Dest Port (8472)
UDP Length
UDP Check Sum
VXLAN Flags
RSVD
VXLAN NI (VNI)
FCS
RSVD
VXLAN Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
Outer UDP Header
8 bytes
VXLAN Header
8 bytes
EtherType
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
Layer 2
43. 43
Receive Side Scaling for VXLAN
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Outer Dest MAC
Outer Source MAC
Optional VXLAN Type
Optional Outer 802.1Q
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
Source Port
Dest Port (8472)
UDP Length
UDP Check Sum
VXLAN Flags
RSVD
VXLAN NI (VNI)
FCS
RSVD
VXLAN Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
Outer UDP Header
8 bytes
VXLAN Header
8 bytes
EtherType
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
192.168.100.20
192.168.100.10
Layer 2
Layer 3
44. 44
Receive Side Scaling for VXLAN
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Outer Dest MAC
Outer Source MAC
Optional VXLAN Type
Optional Outer 802.1Q
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
Source Port
Dest Port (8472)
UDP Length
UDP Check Sum
VXLAN Flags
RSVD
VXLAN NI (VNI)
FCS
RSVD
VXLAN Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
Outer UDP Header
8 bytes
VXLAN Header
8 bytes
EtherType
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
192.168.100.20
192.168.100.10
8472
Unique
Layer 2
Layer 3
Layer 4
45. 45
Receive Side Scaling for VXLAN
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Outer Dest MAC
Outer Source MAC
Optional VXLAN Type
Optional Outer 802.1Q
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
Source Port
Dest Port (8472)
UDP Length
UDP Check Sum
VXLAN Flags
RSVD
VXLAN NI (VNI)
FCS
RSVD
VXLAN Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
Outer UDP Header
8 bytes
VXLAN Header
8 bytes
EtherType
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
192.168.100.20
192.168.100.10
8472
Unique
Layer 2
Layer 3
Layer 4
Intel Ethernet Converged Network Adapter X520
Intel Ethernet Converged Network Adapter X540
46. 46
Receive Side Scaling for VXLAN
Inner Dest MAC
Inner Source MAC
Optional Ether Type
Optional Inner 802.1Q
IP Header
TCP/UDP
Application Data
Inner Ethernet Frame
Outer Dest MAC
Outer Source MAC
Optional VXLAN Type
Optional Outer 802.1Q
IP Header Data†
IP Protocol
Header Check Sum
Outer Source IP
Source Port
Dest Port (8472)
UDP Length
UDP Check Sum
VXLAN Flags
RSVD
VXLAN NI (VNI)
FCS
RSVD
VXLAN Encapsulated Frame
Outer Ethernet Header
14 bytes
Outer IP Header
20 bytes
Outer UDP Header
8 bytes
VXLAN Header
8 bytes
EtherType
Outer Dest IP
†IP Header Data = Version, IHL, TOS, Length, ID
68:05:ca:27:ab:b9
68:05:ca:27:af:9d
192.168.100.20
192.168.100.10
5001
5002
8472
Unique
ca:f1:ea:bc:51:4b
3a:50:3c:94:c9:45
d6:b3:69:8c:d7:462a:e4:d2:12:bd:46
Layer 2
Layer 3
Layer 4
Intel Ethernet Converged Network Adapter X520
Intel Ethernet Converged Network Adapter X540
Intel® Ethernet Converged Network Adapter XL710
Intel Ethernet Converged Network Adapter X710
47. 47
Intel® Virtualization Technology
CPU utilization per core
Core 1
Core 2
Core 3
Core 4
Core5
Core N
VXLAN Network Virtualization Optimizations using Receive Side ScalingVTEP / Virtual Switch
Without Receive Side Scaling
SingleRx Queue
48. 48
Intel® Virtualization Technology
CPU utilization per core
Core 1
Core 2
Core 3
Core 4
Core5
Core N
CPU utilization per core
Core 1
Core 2
Core 3
Core 4
Core 5
Core N
VXLAN Network Virtualization Optimizations using Receive Side ScalingVTEP / Virtual SwitchVTEP / Virtual Switch
Receive Side Scaling for VXLAN Traffic
Without Receive Side Scaling
SingleRx Queue
MultipleRx Queues
49. 49
Intel® Virtualization Technology
Feature
Intel® Ethernet Products
EnablingTechnology
Acceleration for VXLAN Traffic
Intel® Ethernet ConvergedNetwork Adapter X520
Intel Ethernet Converged Network Adapter X540
ReceiveSide Scaling for VXLAN Traffic
(scale Rx/Txtraffic based on the VXLAN Outer SRC UDP Port [Layer 4] )
Advanced Acceleration for VXLAN Traffic with Stateless Offloads
Intel® Ethernet Converged Network Adapter XL710
Intel Ethernet Converged Network Adapter X710
Receive Side Scaling for VXLAN Traffic
(scale Rx/Txtraffic based Inner or Outer header information Plus Stateless Offloads)
CPU utilization per core
Core 1
Core 2
Core 3
Core 4
Core5
Core N
CPU utilization per core
Core 1
Core 2
Core 3
Core 4
Core 5
Core N
VXLAN Network Virtualization Optimizations using Receive Side ScalingVTEP / Virtual SwitchVTEP / Virtual Switch
Receive Side Scaling for VXLAN Traffic
Without Receive Side Scaling
Linux*enable commands: # ethtool-N “device ID” rx-flow-hash udp4 sdfn
(Enabled by default only on XL710/X710) # ethtool-N “device ID” rx-flow-hash tcp4 sdfn
SingleRx Queue
MultipleRx Queues
50. 50
Network Functions Virtualization (NFV)
Router
VPN
Firewall
Load Balancer
Network Services
Switch
Current Model
•Services in dedicated hardware or physical boxes that are Network Topology dependent
•Inflexible deployment model, requires changing forwarding behavior
Today IT delivers a network service by utilizing ordered sets of cooperating network applications known as Service Function Chain (SFC)
51. 51
Network Functions Virtualization (NFV)
Hypervisor
Virtual Switch
PhysicalHardware
Hypervisor
Virtual Switch
PhysicalHardware
Router
VPN
Firewall
Load Balancer
Network Services
Switch
Current Model
•Services in dedicated hardware or physical boxes that are Network Topology dependent
•Inflexible deployment model, requires changing forwarding behavior
Today IT delivers a network service by utilizing ordered sets of cooperating network applications known as Service Function Chain (SFC)
52. 52
Network Functions Virtualization (NFV)
Hypervisor
Virtual Switch
PhysicalHardware
Hypervisor
Virtual Switch
PhysicalHardware
Router
VPN
Firewall
Load Balancer
Network Services
Switch
Current Model
•Services in dedicated hardware or physical boxes that are Network Topology dependent
•Inflexible deployment model, requires changing forwarding behavior
NFV is about dynamic provisioning of services
•Virtualizing service functions on Intel® Architecture based servers in VMs
Today IT delivers a network service by utilizing ordered sets of cooperating network applications known as Service Function Chain (SFC)
53. 53
Metadata for Network Function Virtualization (NFV)
ServiceClassifier
NetworkForwarder
SFCProxy
SFCAware
Service Function
SFCUnaware
Service Function
IETF*Service Function Chaining
Service Forwarder
https://datatracker.ietf.org/wg/sfc/documents/
54. 54
Metadata for Network Function Virtualization (NFV)
NSH: Network Services Header
Geneve: Generic Network Virtualization Encapsulation
ServiceClassifier
NetworkForwarder
SFCProxy
SFCAware
Service Function
SFCUnaware
Service Function
Service Function Chaining (SFC)
Internet Engineering Task Force (IETF)
IETF*Service Function Chaining
Outer
Ethernet Header
Outer IP Header
Outer UDP Header
Geneve Base Header
GeneveOptions
Inner Payload
Outer CRC
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
VER|O|C|R|R|R|R|R|R|R|Length | MD Type = 1 | Next Protocol
ServicePath ID | Service Index
Mandatory Context Header
Mandatory Context Header
Mandatory Context Header
Optional Variable Length Context Headers
Version | Option Length | OAM | Critical Options | Reserved | Protocol Type
VirtualNetwork Identifier (VNI) | Reserved
VariableLength Options
Service Forwarder
https://datatracker.ietf.org/wg/sfc/documents/
55. 55
Generic Network Virtualization Encapsulation (Geneve)
R –Reserved
Geneve Option: Type, Length, Value (TLV) Format
Outer
Ethernet Header
Outer IP Header
Outer UDP Header
Geneve Base Header
GeneveOptions
Inner Payload
Outer CRC
Geneve Header:
Co-authored by
Version | Option Length | OAM | Critical Options | Reserved | Protocol Type
VirtualNetwork Identifier (VNI) | Reserved
VariableLength Options
OptionClass | Option Type | R | R | R | Length
VariableLength Options
Geneve overview:
•Geneve is UDP encapsulation for overlays
•Unifies VXLAN, NVGRE, STT formats
•Extensible to support future control planes
•Options infrastructure to carry metadata/ context for network virtualization & service chaining
•Options use TLV format for flexibility
Motivation for Geneve:
•Metadata (system state, service context)
Example usage for metadata
•Service Chaining: Sharing service context between service functions e.g., FW, LB, DPI, NAT, VPNhttps://datatracker.ietf.org/doc/draft-gross-geneve/
56. 56
Getting 40Gb/s between Two Hosts using Geneve
Demo of Geneve Overlay at 40Gbps in IDF Showcase Booth 121
57. 57
Software and Hardware for NFV
First Open 40GbE Driver
DPDK.org
Common Network Elements
Intel®Architecture based servers for Communications and Storage Virtual Appliances
Migration from closed, tightly integrated architecture to open architecture with Linux* packet processing interface
+ Intel Ethernet Converged Network Adapter XL710 / X710 Family
Intel Data Plane Development Kit
1Source as of Aug 2014: Intel® Data Plane Development Kit (Intel® DPDK) / Intel® Ethernet CNA X710 4x10GbE IPv4 Layer 3 Forwarding Performance -Routing Control Unit (RCU) bypass improved 128B performance from 31Gbps (80% line rate) to 38 Gbps(95% line rate). SUT: Rose City CRB, E5-2658v2 UP, DDR3-1867 ECC 1DPC [XL710 (rev 01) 4x10GBE, EETrackID: 124D]
40Gbps
128B
256B
512B
1024B
0 Gbps
64B
Line-Rate Above 128B1
Optimized Network Drivers
igb, ixgbe, and i40e
58. 58
Physical Server Networking Connectivity
1GbE
10GbE
40GbE
Transitioning to Different Ethernet Speeds
10000BASE-T
SR/LR Optics
10GBASE-T
Direct Attach Copper
SR/LR Optics
No BASE-T Option
Direct Attach Copper
SR/LR Optics
59. 59
Introducing Low-cost QSFP+ Optics withIntel® Ethernet Modular Optics and Cable Solution (MOCs)
Intel® Ethernet CNAXL710-QDA1
Intel Ethernet CNAXL710-QDA2
Intel® Ethernet QSFP+ SR Optics
Intel Ethernet Modular Optic and Cable Solution
Source as of Aug 2014: Pricing from CDW website –SR4 Optics FTL410QD2C ($585 x2) + MPO Cable PRO-MPOMPO-10M5OM3 ($209), AOC #: MC2210310-010 ($512), Intel Ethernet MOT ($107 x2) + Intel Ethernet MOC ($97) = $311
Intel® Ethernet Modular Optical Transceiver
Low cost option to 40GBASE-SR4
Modular alternative to AOC cables
Low power with RoHS compliant lenses
Intel Ethernet Modular Optical Cable
Thinner and lighter cable than CR4
Robust and flexible Fiber cables
7mm bend radius
Intel® Ethernet Optics
60. 60
Introducing Low-cost QSFP+ Optics withIntel® Ethernet Modular Optics and Cable Solution (MOCs)
Intel® Ethernet CNAXL710-QDA1
Intel Ethernet CNAXL710-QDA2
Intel® Ethernet QSFP+ SR Optics
Intel Ethernet Modular Optic and Cable Solution
CR4
(Passive Copper)
AOC
(Active Optical)
SR4
(Optical)
Intel® Ethernet MOCs
(Optical)
MaxReach
7m
100m
150m
100m
Bend Radius
98mm
35mm
35mm
7mm
Modular Design
No
No
Yes
Yes
10Meter + Optics
N/A
$512
$1379
$311
Comparing QSFP+ Options
Source as of Aug 2014: Pricing from CDW website –SR4 Optics FTL410QD2C ($585 x2) + MPO Cable PRO-MPOMPO-10M5OM3 ($209), AOC #: MC2210310-010 ($512), Intel Ethernet MOT ($107 x2) + Intel Ethernet MOC ($97) = $311
Intel® Ethernet Modular Optical Transceiver
Low cost option to 40GBASE-SR4
Modular alternative to AOC cables
Low power with RoHS compliant lenses
Intel Ethernet Modular Optical Cable
Thinner and lighter cable than CR4
Robust and flexible Fiber cables
7mm bend radius
Intel® Ethernet Optics
61. 61
•Transforming The Network For The Cloud
•Accelerating Network Virtualization Overlays?
•Next Generation Servers With Integrated Ethernet
Agenda
62. 62
Creating Server Optimized Network Services
Characteristics of optimized network services –beyond just virtualization
-Design point is Private Cloud
-Current goal is full utilization of physical resources with VMs
5-50 VMs per physical host can be typical
New requirements for high VM density for Private Cloud
1.Lower network and storage CPU overhead
2.Higher throughput requirements due to high VM density
3.Low variance for latency & throughput (95thpercentile)
4.Transparent hardware fault tolerance for network
5.VM workload isolation
A solution: SMB3 and SMB Direct (RDMA support)
63. 63
The Origins of SMB3
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
64. 64
File server cluster
The Origins of SMB3
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
65. 65
File server cluster
The Origins of SMB3
SMB
Microsoft SQL Server
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
66. 66
File server cluster
The Origins of SMB3
SMB
Microsoft SQL Server
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
67. 67
File server cluster
The Origins of SMB3
SMB
Microsoft SQL Server
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
68. 68
File server cluster
The Origins of SMB3
SMB
Windows* virtualizedstorage
Tiered physicalstorage
Storage space
Storage space
Storage space
SSD
HDD
Microsoft SQL Server
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
69. 69
The Origins of SMB3
SMB
Microsoft SQL Server
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
70. 70
Third Party
SMB3 File Servers
The Origins of SMB3
SMB
Microsoft SQL Server
•File sharing semantics rather than block semantics
-Increased flexibility, easier provisioning and management
-Easy deployment of encryption & signing
•Enterprise class RAS
-No application downtime for planned maintenance or unplanned failures
-Extremely fast failover (<10 sec)
•Excellent Performance
-Near line rate performance for both small and large IOs w/ RDMA
-Very low CPU utilization w/ RDMA
•Can work with existing storage investments
•Works on existing network infrastructure and next generation infrastructure
71. 71
SMB3 and SMB Direct Workloads
SMB3 Workloads
•Storage for Hyper-V, SQL Server, HPC
•Storage for desktops/laptops/slates (LAN/WAN)
•Hyper-V Live Migration between hosts
The Design Point is Private Cloud
•Almost all IOs are small (<64 KB)
-Throughput is significantly less due to CPU saturation
-RDMA enables near line rate with small IOs
SMB3 Multi-Channel Enables Linear Scaling
•Linear 10GbE scaling with TCP/IP
•4300 Mbps with 4x10GbEhttp://go.microsoft.com/fwlink/p/?LinkId=227841
Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark*and MobileMark*, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products.
72. 72
What is Remote Direct Memory Access (RDMA) ?
RDMA
•Accelerated IO delivery model which works by allowing application software to bypass most layers of software and communicate directly with the hardware
RDMA benefits
•Low latency
•High throughput
•Zero copy capability
•OS / Stack bypass
RDMA Hardware Technologies
•iWARP: RDMA over TCP/IP
•RoCE: RDMA over Converged Ethernet
•InfiniBand*
RDMA support in Windows*network stack
•New Network Direct Kernel-mode Provider Interface (NDKPI), which abstracts the hardware
File Server
SMB Direct
Client
RDMA NIC
SMB Direct
Ethernet or InfiniBand
SMB Server
SMB Client
Memory
Memory
NDKPI
NDKPI
RDMA NIC
RDMA
73. 73
SMB3 with RDMA on 40GbE Chelsio*T580 PRELIMINARY
Large IO (512 KB)†in Gbps
†IOs are not written to non-volatile storage
Test configuration details in backup
QD R/ Thread
Read
IOPs
50thRead
99thRead
1
163,800
0.056
0.249
2
291,000
0.094
0.270
4
440,900
0.129
0.397
8
492,400
0.195
1.391
16
510,200
0.363
2.810
0.000
1.000
2.000
3.000
0
20
40
60
80
100
Latency(ms)
8K Read Incast -2 Threads / Server @ QD R / Th
1R
2R
4R
8R
16R
Near line-rate with small IOs
•33.4 Gbpsat 8 KB IO
•Excellent latency variance
8 KB IOPs†and Latency (ms)
Single client, 8 file servers
Large IO achieves near line-rate
Excellent Log Write Performance
•7.2M Read IOs†, 512 Byte, single outstanding IO
•3.3M Write IOs†, 512 Byte, single outstanding IO
512KB
Read BW
Write BW
1 Thread
1 IO
17.82
16.14
2 IO
29.74
23.13
2 Thread
2 IO/t
37.21
30.61
0.000
0.500
1.000
1.500
2.000
0
50
100
Latency (ms)
512KB Reads
1+1
1+2
2+2
Client-to-Server Performance
IncastPerformance
Percentage
Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark*and MobileMark*, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products.
74. 74
Next Generation PCH with Integrated iWARPEnabled LAN Controller
SMBus
NC-SI
Integrated I/O
MCTP
VF0
VF1
…
VFn
In-band
Mgmt
PF0
PF1
…
PFn
PCI Express*3.0 x8 –SR-IOV
Queue Mgmt, Scheduler
Protocol Acceleration / Offloads
Qm
Q0
Q1
Q2
Q3
…
VEB, DCB Traffic Classifier
MACs
Integrated PHYs
Key Capabilities
•Remote Direct Memory Access (RDMA) via iWARP
•Network Virtualization Offloads for Geneve, VXLAN and NVGRE
•Flexible filters with Intel® Ethernet Flow Director and Application Targeted Routing (ATR)
•SR-IOV support up to 4 Physical Functions (PFs) and 128 Virtual Functions (VFs)
•VEB (Virtual Ethernet Bridge), Edge Virtual Bridging / 802.1Qbg
•Data Center Bridging (DCB) with iSCSI stateless offloads
Demo of implemented in an FPGA running Microsoft*Windows*2012 R2 SMB Direct in IDF Showcase Booth 106
75. 75
System Solutions Powered By Intel® Technologies
Delivering optimal efficiency and commonly available building blocks
Intel Xeon processors
Hyper-Threading
Turbo boost
Intel® AVX 2.0
Intel® VT-x
Ecosystem Enabling
Open Data Center Alliance
Intel® Cloud Builders
Security
Intel TXT
Intel AES-NI
Intel Secure Key
Software
Open Attestation SDKIntel® Data Center ManagerIntel® Node ManagerIntel® Service Assurance Administrator
Compute
Intel® Xeon® processor E5 v3 family
Intel® Architecture
Intel® Virtualization Technologies (Intel® VT)
Ecosystem Enabling
Intel® Network Builders
Network Acceleration
Communications chipset
Intel® Ethernet CNAs
Intel® Ethernet Switch Silicon
Software
Intel® QuickAssist APIs
Intel® Data Plane Development Kit
Network
Intel® Ethernet CNA XL710/X710 family
Intel Ethernet CNA X520/X540 family
Solid State Drives
Ecosystem Enabling
HP
Red Hat*
Nexenta
Plus others…
Storage Accelerators & SoCs
Software
Storage acceleration libraries (ISA-L)
Intel®CAS -Cache Acceleration Software
Storage
Intel® Solid State Drive DC P3700
Intel® SSD DC S3700
Intel® SSD DC P3700
Intel® Advanced Vector Extensions (Intel® AVX)
Intel® Virtualization Technology (Intel® VT)
Intel® Trusted Execution Technology (Intel® TXT)
Intel® Advanced Encryption Standards New Instructions (Intel® AES-NI)
76. 76
•Intel® Ethernet Controller XL710
-Provides a flexible 10/40 gigabit Ethernet connection at 3.8W that is optimized for Intel® Xeon® Processor E5 v3 server platforms
-Improves performance of virtualized networks and cloud applications with Network virtualization Overlay stateless offloads for Geneve, VXLAN and NVGRE
-Optimized for Intel® Data Plane Development Kit (Intel® DPDK) to provide the platform of choice for Network Function Virtualization (NFV)
•Intel Is Implementing iWARP RDMA In Future Intel Xeon Processor Based Servers
-RDMA is an advanced networking technology that lowers the latency and improves the efficiency of network data transfers
-Intel intends to drive broad adoption of iWARP RDMA via Intel® Ethernet IP integration in to server silicon
Summary
77. 77
•Attend other sessions (or review materials) to learn more about Intel’s work on Software Defined Infrastructure
•IT administrators and developers should transition to the Intel® Ethernet Controller XL710 based adapters to evaluate Network Virtualization Overlay performance improvements for both 10GbE and 40GbE connections
•Developers should look for opportunities to use iWARP RDMA to take advantage of its broad deployment in future Intel® Xeon® processor based servers
Call to Action
78. 78
Additional Sources of Information
A PDF of this presentation is available from our Technical Session Catalog: www.intel.com/idfsessionsSF. This URL is also printed on the top of Session Agenda Pages in the Pocket Guide.
Demos in the showcase –
-Demo of implemented Geneve and VXLAN Network Virtualization Overlays with Intel® Ethernet Controller XL710 at 40Gbps IDF Showcase Booth 121
-Demo of implemented in an FPGA running Microsoft*Windows*2012 R2 SMB Direct in IDF Showcase Booth 106
More web based info: www.intel.com/go/ethernet
80. 80
Risk Factors
The above statements and any others in this document that refer to plans and expectations for the second quarter, the year and the future are forward- looking statements that involve a number of risks and uncertainties. Words such as “anticipates,” “expects,” “intends,” “plans,”“believes,” “seeks,” “estimates,” “may,” “will,” “should” and their variations identify forward-looking statements. Statements that refer to or are based on projections, uncertain events or assumptions also identify forward-looking statements. Many factors could affect Intel’s actual results, and variances from Intel’s current expectations regarding such factors could cause actual results to differ materially from those expressed in these forward-looking statements. Intel presently considers the following to be important factors that could cause actual results to differ materially from thecompany’s expectations. Demand for Intel's products is highly variable and, in recent years, Intel has experienced declining orders in the traditional PC market segment. Demand could be different from Intel's expectations due to factors including changes in business and economic conditions; consumer confidence or income levels; customer acceptance of Intel’s and competitors’ products; competitive and pricing pressures, including actions taken by competitors; supply constraints and other disruptions affecting customers; changes in customer order patterns including order cancellations; and changes in the level of inventory at customers. Intel operates in highly competitive industries and its operations have high costs that are either fixedor difficult to reduce in the short term. Intel's gross margin percentage could vary significantly from expectations based on capacity utilization; variationsin inventory valuation, including variations related to the timing of qualifying products for sale; changes in revenue levels; segment product mix; the timing and execution of the manufacturing ramp and associated costs; excess or obsolete inventory; changes in unit costs; defects or disruptions in the supply of materials or resources; and product manufacturing quality/yields. Variations in gross margin may also be caused by the timing of Intel product introductions and related expenses, including marketing expenses, and Intel's ability to respond quickly to technological developments and to introduce new products or incorporate new features into existing products, which may result in restructuring and asset impairment charges. Intel's resultscould be affected by adverse economic, social, political and physical/infrastructure conditions in countries where Intel, its customers or its suppliers operate, including military conflict and other security risks, natural disasters, infrastructure disruptions, health concerns and fluctuations in currency exchange rates. Intel’s results could be affected by the timing of closing of acquisitions, divestitures and other significant transactions. Intel's results could be affected by adverse effects associated with product defects and errata (deviations from published specifications), and by litigation or regulatory matters involving intellectual property, stockholder, consumer, antitrust, disclosure and other issues, such as the litigation and regulatory matters described in Intel's SEC filings. An unfavorable ruling could include monetary damages or an injunction prohibiting Intel from manufacturing or selling one or more products, precluding particular business practices, impacting Intel’s ability to design its products, or requiring other remedies such as compulsory licensing of intellectual property. A detailed discussion of these and other factors that could affect Intel’s results is included in Intel’sSEC filings, including the company’s most recent reports on Form 10-Q, Form 10-K and earnings release.
Rev. 4/15/14