background image

B–Data Center Bridging Technology

Data Center Bridging (DCB)

B-2

51031-00  A

Enhanced Transmission Selection (ETS) protocol addresses the bandwidth 
allocation issues among various traffic classes to maximize bandwidth usage. The 
IEEE 802.1Qaz standard specifies the protocol to support allocation of bandwidth 
amongst priority groups. ETS allows each node to control bandwidth per priority 
group. When the actual load in a priority group does not use its allocated 
bandwidth, ETS allows other priority groups to use the available bandwidth. The 
bandwidth-allocation priorities allow the sharing of bandwidth between traffic 
loads, while satisfying the strict priority mechanisms already defined in IEEE 
802.1Q that require minimum latency.

Bandwidth allocation is achieved as part of a negotiation process with link 
peers—this is called DCB Capability eXchange protocol (DCBX). It provides a 
mechanism for Ethernet devices (bridges, end stations) to detect the DCB 
capability of a peer device. It also allows configuration and distribution of ETS 
parameters from one node to another. 

ETS and DCBX simplify the management of DCB nodes significantly, especially 
when deployed end-to-end in a converged data center. The DCBX protocol uses 
Link Layer Discovery Protocol (LLDP) defined by IEEE 802.1AB to exchange and 
discover DCB capabilities.

Priority Flow Control

A fundamental requirement for a high performance storage network is guaranteed 
data delivery. This requirement must be satisfied to transport critical storage data 
on a converged Ethernet network with minimum latency. Another critical 
enhancement to conventional Ethernet is lossless Ethernet. IEEE 802.3X PAUSE 
defines how to pause link traffic at a congestion point to avoid packet drop. IEEE 
802.1Qbb defines Priority Flow Control (PFC), which is based on IEEE 802.3X 
PAUSE and provides greater control of traffic flow. PFC eliminates lost frames 
caused by congestion. PFC enables the pausing of less sensitive data classes, 
while not affecting traditional LAN protocols operating through different priority 
classes.

Figure B-1

 shows how PFC works in a converged traffic scenario.

Summary of Contents for Storage Networking (Unified Fabric Pilot)

Page 1: ...51031 00 A Installation Guide Storage Networking Unified Fabric Pilot QLogic 8100 Series Adapters QLogic 5800 Series Switches Cisco Nexus 5000 Series Switches JDSU Xgig Platform...

Page 2: ...gic Corporation reserves the right to change product specifications at any time without notice Applications described in this document for any of these products are for illustrative purposes only QLog...

Page 3: ...tecture Description 3 1 Architecture Overview 3 1 Equipment Details 3 3 4 Installation Determine Configuration 4 1 Equipment Installation and Configuration 4 2 Install the Converged Network Adapter Ha...

Page 4: ...5 20 Exchange Completion Time Latency Monitor 5 25 Validation Step 6 iSCSI Function and I O Test 5 26 Validation Process 5 26 Validation Results 5 27 Validation Step 7 Virtualization Verification 5 2...

Page 5: ...FIP Keep_Alive Frames 5 9 5 9 Time Difference Between Two Adjacent FIP Discovery Advertisement Multicast Frames to All_E_Nodes 5 10 5 10 Link De instantiated by the Switch 20 Seconds after the Last Ke...

Page 6: ...Flow Control B 3 B 2 FCoE Mapping Illustration Source FC BB 5 Rev 2 0 B 3 List of Tables Table Page 3 1 Converged Network Equipment Inventory 3 3 3 2 JDSU Converged Network Validation Equipment 3 3 5...

Page 7: ...es basic knowledge of Enhanced Ethernet and the associated standards If the you are not familiar with Fibre Channel over Ethernet FCoE and Enhanced Ethernet review the documents listed in Related Mate...

Page 8: ...d links to Web sites are shown in underlined blue For example Table 9 2 lists problems related to the user interface and remote agent See Installation Checklist on page 3 6 For more information visit...

Page 9: ...e agreements refer to the QLogic Software End User License Agreement What are shortcut keys To enter the date type mm dd yyyy where mm is the month dd is the day and yyyy is the year Topic titles betw...

Page 10: ...Preface Documentation Conventions x 51031 00 A...

Page 11: ...ed market acceptance because it delivers excellent performance reduces data center total cost of ownership TCO and protects current data center investments A unified fabric with FCoE preserves existin...

Page 12: ...1 Overview 1 2 51031 00 A...

Page 13: ...nd validation process accordingly Organizational Ownership Fibre Channel Storage Ethernet Networking A critical factor for successfully implementing a unified data center fabric is the stability of ne...

Page 14: ...e most common implementation in today s data centers because the Fibre Channel switch and SAN storage are typically already in place Converged Network Adapter DCB switch FCF Fibre Channel switch Fibre...

Page 15: ...based on JDSU hardware and software Screen shots and trace data were captured to show the results from the seven step validation process and to demonstrate the various DCB functions Reference Archite...

Page 16: ...3 Architecture Reference Architecture Description 3 2 51031 00 A Figure 3 1 Reference Architecture Diagram...

Page 17: ...t server Windows 2008 UCS C200 M1 1 NetApp FAS3040 storage system 10Gb iSCSI storage and 10Gb native FCoE storage FAS3040 1 HDS AMS500 Storage Array 4Gb Fibre Channel storage AMS500 1 QLogic 5800V 8Gb...

Page 18: ...3 Architecture Reference Architecture Description 3 4 51031 00 A...

Page 19: ...Ie Gen1 x8 or PCIe Gen2 x4 Hardware platforms IA32 x86 Intel64 AMD64 x64 IA64 SPARC PowerPC Operating systems Windows Server 2003 2008 2008 R2 targeted Red Hat EL AP 5 x Novell SLES 10 x 11 VMware ESX...

Page 20: ...r and find an empty PCIe x8 bus slot Gen1 or PCIe x4 bus slot Gen2 4 Pull out the slot cover if any by removing the screw or releasing the lever 5 Install the low profile bracket if required 6 Grasp t...

Page 21: ...http driverdownloads qlogic com QLogicDriverDownloads_UI default aspx Figure 4 1 Driver Download Page Model Selection 2 In the table at the bottom of the page select Converged Network Adapters the ada...

Page 22: ...I default aspx 2 In the table at the bottom of the page Figure 4 2 select Converged Network Adapters the adapter model your operating system and then click Go Figure 4 2 Driver Download Page Driver Li...

Page 23: ...ntation for zoning instructions FCoE Switches QLogic and Cisco have jointly developed the QLogic and Cisco FCoE Design Guide for implementing a unified data center using Cisco Nexus 5000 Series switch...

Page 24: ...ty 3 class map type qos class nic match cos 0 class map type qos class fcoe match cos 3 3 Configure queuing class maps class map type queuing class nic match qos group 2 4 Configure network qos class...

Page 25: ...ng class fcoe bandwidth percent 50 class type queuing class default bandwidth percent 0 7 Configure network qos policy maps and set up the PFC for no drop traffic class policy map type network qos pol...

Page 26: ...work qos policy maps policy map type network qos policy1 class type network qos class nic match qos group 2 pause no drop class type network qos class fcoe match qos group 1 pause no drop mtu 2240 cla...

Page 27: ...ass map queuing class nic match any Match qos group 2 bandwidth percent 50 Class map queuing class fcoe match any Match qos group 1 bandwidth percent 50 Class map queuing class default match any Match...

Page 28: ...tch under both Ethernet eth1 16 and Fibre Channel vfc116 protocols Figure 4 4 shows the Device Manager interface for the Cisco Nexus 5000 FCoE switch Figure 4 4 Device Manager Interface Verify Server...

Page 29: ...rs to your storage so that the servers can discover the LUNs Figure 4 6 shows an example of the discovered FCoE Converged Network Adapters and the created initiator groups on a NetApp Filer after zoni...

Page 30: ...Nsurfer Management Validation Screen 1 Figure 4 8 SANsurfer Management Validation Screen 2 7 Use the operating system tools to create disk partitions and volumes on your servers using the assigned LUN...

Page 31: ...sion used by each network component Priorities assigned to each protocol Priorities with the per PFC enabled Bandwidth assigned to each priority Validation Process 1 Place the Analyzer into the two pa...

Page 32: ...4 Configure the FCoE switch with the following parameters FCoE priority 3 PFC enabled ETS bandwidth 50 percent All other classes of service lossy ETS bandwidth 50 percent 5 Capture the trace and veri...

Page 33: ...onfigured at the switch console If IEEE DCBX version 1 4 or later is used then verify the following Converged Network Adapter and Target Willing bit W is ON while the Switch W bit is OFF pfc_enable va...

Page 34: ...ers 5 4 51031 00 A Validation Results Figure 5 3 shows one trace captured between the adapter and FCoE switch exchanging DCB parameters using the DCBX protocol Both peers use version 1 01 Figure 5 3 D...

Page 35: ...they share the remaining 50 percent of the bandwidth If the switch cannot propagate its DCB parameters to the peer devices then use the JDSU Load Tester to negotiate different versions of the DCBX pro...

Page 36: ...switch FCF or between the FCoE target and the FCoE switch FCF The FIP test setup is shown in Figure 5 5 Figure 5 5 FIP Test Setup Within TraceView normal FIP operation begins with FIP discovery advert...

Page 37: ...to the switch with the FIP FLOGI message 6 The switch responds with an Accept FLOGI or a Reject The Accept FLOGI contains all the Fibre Channel characteristics It also contains the new MAC address to...

Page 38: ...CoE Switch From the captured trace you can verify the following FIP version used by both devices is 1 FCoE VLAN ID returned by the switch is 1002 The link is capable of delivering 2181 byte frames The...

Page 39: ...of the variable FKA_ADV_PERIOD To validate that the switch disconnects end devices after 20 seconds use the JDSU Jammer to remove all FIP Keep Alive messages from the link after an end device has suc...

Page 40: ...ding FIP Keep Alive messages every eight seconds and FIP Discovery Advertisements every eight seconds Figure 5 9 Time Difference Between Two Adjacent FIP Discovery Advertisement Multicast Frames to Al...

Page 41: ...1031 00 A 5 11 Figure 5 11 shows the time difference between the last FIP Keep Alive Sent message and the FIP Clear Virtual Link message from the switch Figure 5 11 Time Difference Between the Last FI...

Page 42: ...nverged Network Adapter host FCoE switch Fibre Channel switch Fibre Channel storage Figure 5 12 I O Test with Fibre Channel SAN Storage Setup Validation Process The following validation steps are used...

Page 43: ...rity 5 Set up the Analyzer s TraceControl to capture FCoE frames over the Ethernet VLAN by setting these frame types as trigger conditions The FCoE EtherType is 0x8906 6 Use Expert to analyze the trac...

Page 44: ...rcent of the total available bandwidth The bandwidth may have been limited by the storage device in this case whereas it was limited by the network in the other two cases This conclusion is based on t...

Page 45: ...following The hosts have storage LUNs mapped and can generate I O to storage devices The Analyzer connects to the FCoE switch with a 10 Gbps link and the FCoE switch connects to the Fibre Channel stor...

Page 46: ...gured and the link is not lossless In this case ensure that PFC is enabled and restart the validation at Step 1 When TraceControl stops capturing it means that at least one PFC frame crossed the link...

Page 47: ...a credit In such a case the switch will probably drop the frames which will cause the Fibre Channel protocol to abort the sequence and force the initiator to re issue the entire sequence If this is a...

Page 48: ...Process The test assumes that the hosts have mapped storage device LUNs and can generate I O to storage devices The test should include two traffic classes one for FCoE and one for combined traffic i...

Page 49: ...igure 5 19 Verifying ETS with Read Operations from the Storage 3 Start the FCoE and other traffic TCP iSCSI simultaneously in the MLTT application to saturate the link 4 Use TraceControl to monitor li...

Page 50: ...e Figure 5 20 Throughput per Priority Validation Results Table 5 3 lists the port settings to verify the ETS setup ETS allocates 50 percent of the bandwidth to FCoE and 50 percent of the bandwidth to...

Page 51: ...ine rate 1176MBps Because the traffic reached full line rate the switch enforced ETS parameters The FCoE traffic decreases to 600MBps as seen on port GE Storage 1 1 1 The difference between the full l...

Page 52: ...switch In addition Expert reports aborted sequences and retransmitted FCoE iSCSI and TCP traffic Expert reports all symptoms and network anomalies indicating frame losses and performance degradation F...

Page 53: ...izes the throughput achieved for the previous three tests The test continues cycling FCoE traffic and TCP traffic on and off The results show that the switch continues to enforce the ETS configuration...

Page 54: ...Before the start of TCP traffic server ingress is equal to the FCoE traffic When the TCP traffic starts server ingress is the sum of the FCoE and TCP traffic The switch starts sending PFC pause reques...

Page 55: ...he read latency increases from about 10ms to 12 5ms Figure 5 27 Read Latency Measurement In this test the link congestion did not happen at the host with the oversubscription scenario Instead the swit...

Page 56: ...functions Figure 5 28 iSCSI Traffic Performance Validation Setup Validation Process 1 Enable the Converged Network Adapter to process FCoE and iSCSI data 2 Configure the Analyzer to capture any traffi...

Page 57: ...ined Fibre Channel and iSCSI storage trace and calculate the average minimum and maximum throughput results per priority 13 Compare performance results for FCoE storage iSCSI storage and converged FCo...

Page 58: ...Fibre Channel storage throughput is about 90 percent of wire speed and iSCSI storage throughput is about 35 percent of wire speed There were no data integrity errors during the test In summary this te...

Page 59: ...d one FCoE initiator in another virtual machine as shown in Figure 5 30 Figure 5 30 Setup for Verifying Virtualization Validation Process 1 Set up two virtual machines on one server and install MLTT o...

Page 60: ...roughput results per priority 9 Compare I O performance for the two virtual machines Repeat the entire test using separate hosts for the iSCSI initiator and the FCoE initiator and compare I O performa...

Page 61: ...creased because PFC paused FCoE traffic In contrast the iSCSI application gained bandwidth while FCoE traffic was paused These changes in throughput might have been caused by increased virtual OS proc...

Page 62: ...tiple factors may determine the behavior of virtual server based applications in the unified network environment The internal application balancing algorithm either among virtual servers or within the...

Page 63: ...ooling requirements significant cost savings in infrastructure software and hardware and reduced infrastructure management costs The Nexus 5000 Series switch uses cut through architecture supports lin...

Page 64: ...is the only protocol Analyzer supporting multiple protocols all within the same chassis For more information about the Xgig Analyzer visit http www jdsu com products communications test measurement pr...

Page 65: ...hich leads to faster application performance and greater consolidation in virtualized systems QLogic 5800V Series Fibre Channel Switch The QLogic SB5800V Series Fibre Channel switch supports 2Gb 4Gb a...

Page 66: ...A Hardware and Software QLogic 5800V Series Fibre Channel Switch A 4 51031 00 A...

Page 67: ...is built based on the following DCB protocols DCBX and ETS Priority Flow Control Fibre Channel Over Ethernet iSCSI DCBX and ETS Existing Ethernet standards cannot control and manage the allocation of...

Page 68: ...ity of a peer device It also allows configuration and distribution of ETS parameters from one node to another ETS and DCBX simplify the management of DCB nodes significantly especially when deployed e...

Page 69: ...hernet frame is transported over Enhanced Ethernet networks as shown in Figure B 2 Compared to other mapping technologies FCoE has the least mapping overhead and maintains the same constructs as nativ...

Page 70: ...was finalized by the Internet Engineering Task Force IETF in April 2004 A TCP IP connection ties the iSCSI initiator and target session components together Network portals identified by their IP addre...

Page 71: ...09 Ethernet Alliance 2009 retrieved from http www ethernetalliance org files static_page_files 281AD8C4 1D09 3519 AD 7AD835AD525E36 SC09 20white 20paper pdf Unified Data Center Fabric Primer Fcoe And...

Page 72: ...C References C 2 51031 00 A...

Page 73: ...ETS see enhanced transmission selection exchange completion time 5 25 F FCoE function validation 5 12 switch 4 6 FCoE initialization protocol validation 5 6 Fibre Channel forwarder 2 2 Fibre Channel...

Page 74: ...1 5 15 process summary 3 1 Q QLogic and Cisco FCoE Design Guide 4 5 R related materials vii S SANsurfer FC HBA Manager 4 4 server bus interface 4 1 storage 4 1 4 9 switch FCoE 4 5 4 6 Fibre Channel 4...

Page 75: ......

Page 76: ...d Windows Server are registered trademarks of Microsoft Corporation NetApp is a registered trademark of Network Appliance Inc HDS is a registered trademark of Hitachi Ltd and or its affiliates SPARC i...

Reviews: