#214120

September 2014
Commissioned by
Huawei Technologies Co., Ltd

Huawei CloudEngine 7800/6800/5800 Series Data Center Switch

Performance, Virtualization, Data Center Features and SDN Evaluation

EXECUTIVE SUMMARY THE BOTTOM LINE

Huawei CloudEngine 7800/6800/5800 series switches are 40GbE/10GbE/GbE Huawei CloudEngine 7800/6800/5800 Series Data Center
data center switches developed by Huawei Technologies Co., Ltd. Switch:
Huawei commissioned Tolly to evaluate their CE7800/6800/5800 series
switch performance, virtualization capability, features and SDN
1 Supports FCoE (FCF, NPV, FSB modes and DCB) -
with CE7800/6800
functionalities.
2 Supports virtualization with iStack - virtualize 16
physical switches into 1 logical switch, and SVF
Tolly engineers verified that the CloudEngine switches provided high
performance with low power consumption, virtualization capability with (vertical virtualization) - virtualize multiple
Huawei’s iStack and SVF technologies, as well as numerous data center homogeneous or heterogeneous physical switches
features including VEPA, TRILL, FCoE (FCF, NPV, FSB modes and DCB), and into 1 logical switch with local forwarding on leaf
Huawei nCenter interoperation with VMware vCenter. nodes

Tests also show that the CloudEngine switches supported OpenFlow SDN 3 Supports large layer 2 TRILL network with 512
nodes and active-active TRILL edge
including interoperability with Huawei Agile Controller and the third party
controller “Ryu”, L2/L3 line-rate forwarding, multiple flow tables, policy-based 4 Supports OpenFlow SDN with topology discovery,
L2/L3 line-rate forwarding, multiple flow tables,
routing, and dynamic traffic engineering (TE).
policy-based routing and dynamic traffic
engineering with interoperability with Huawei
Agile Controller and third party SDN controllers

Huawei CloudEngine 7800/6800/5800 Series Data Center Switch Layer 2 Throughput

(as reported by Ixia IxNetwork 7.22.9.9.9EA)
Throughputt (% line rate))
Frame Sizes 64-Byte 128-Byte 256-Byte 512-Byte 1024-Byte 1280-Byte 1518-Byte 9216-Byte
CE7850-32Q-EI
- 100% 100% 100% 100% 100% 100% 100%
32 x 40GbE ports
CE6850-48S6Q-HI
100% 100% 100% 100% 100% 100% 100% 100%
48 x 10GbE + 6 x 40GbE ports
CE6810-48S4Q-EI
100% 100% 100% 100% 100% 100% 100% 100%
48 x 10GbE + 4 x 40GbE ports
CE5850-48T4S2Q-HI
100% 100% 100% 100% 100% 100% 100% 100%
48 x GbE + 4 x 10GbE + 2 x 40GbE ports
CE5810-48T4S-EI
100% 100% 100% 100% 100% 100% 100% 100%
48 x GbE + 4 x 10GbE ports
CE5810-24T4S-EI
100% 100% 100% 100% 100% 100% 100% 100%
24 x GbE + 4 x 10GbE ports
Note: The same type of ports were in snake traffic topology. For example, CE5810-24T4S-EI has 24 GbE ports in snake topology and four 10GbE ports in snake topology.
Source: Tolly, May 2014 Table 1

© 2014 Tolly Enterprises, LLC Tolly.com Page 1 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

depending upon the device and model. In

Test Results
Tolly engineers benchmarked the
performance and feature set of a range of
Huawei CloudEngine 7800/6800/5800
Series Data Center top-of-rack (ToR)
Switches outfitted with Gigabit Ethernet,
10GbE and 40GbE ports.
The feature evaluation included:
virtualization, data center functionality and
OpenFlow capabilities. Test results are
summarized below and detailed in the Test
Setup and Methodology section. See Table
4 for the list of all verified items.
Performance
Layer 2 Throughput &
Latency
For each device under test, the Layer 2
throughput was measured individually
across a range of frame sizes from 64-byte
through 9216-bytes.
Testing encompassed combinations of
Gigabit Ethernet, 10GbE and 40GbE ports
all cases, traffic of a particular topology was Huawei
snaked from port to port. Technologies, Co.,
As shown in Table 1, all models tested of Ltd
the 5800/6800 delivered line rate at every
frame size tested from 64-byte to 9216- CloudEngine
byte jumbo frames. The 7850 outfitted with
7800/6800/5800
32 40GbE ports delivered line rate
throughput at all frame sizes tested from Series Data Center Tested
128-byte to 9216-byte frames. Switches May
Tolly engineers measured the latency at the 2014
same frame sizes in both 40GbE and 10
Performance
GbE configurations. Evaluation and
In tests of 40GbE ports on the CE7850
Feature Validation
switch, latency ranged from 0.60 μs to 0.73
μs. In tests of 10GbE ports on the CE6850 Engineers benchmarked various
switch, latency ranged from 0.87 μs to 0.95 combinations of ports across the
μs. In tests of 10GbE ports on CE6810 CloudEngine 7800/6800/5800 family. Tests
switch, latency ranged from 0.81 μs to 1.37 were carried out according to the ATIS
μs. See Table 2 for detailed results. recommendations and the results can be
found In Table 3.
Power Consumption
To assist network architects in determining
operational costs of the data center
switches, Tolly engineers measured the
power consumption of the devices.

Huawei CloudEngine 7800/6800 Series Data Center Switch Layer 2 Latency

(as reported by Ixia IxNetwork 7.22.9.9.9EA and Spirent TestCenter)
Laten
ncy (μs)

Frame Sizes 64-Byte 128-Byte 256-Byte 512-Byte 1024-Byte 1280-Byte 1518-Byte 9216-Byte

CE7850-32Q-EI
0.60 0.62 0.63 0.68 0.73 0.73 0.73 0.73
40GbE port 1 to port 2 - Cut-through
CE6850-48S6Q-HI
0.87 0.87 0.93 0.94 0.95 0.94 0.94 0.92
10GbE port 1 to port 2 - Store-and-forward
CE6810-48S4Q-EI
0.81 0.86 0.96 1.16 1.38 1.37 1.38 1.37
10GbE port 1 to port 2 - Cut-through
Note: Line-rate traffic was used to test. Cut-through latency measured FIFO latency while store-and-forward latency measured LIFO latency. Thus, store-and-forward
results reported here do not include the time required to store the frame.
Source: Tolly, May 2014 Table 2

© 2014 Tolly Enterprises, LLC Tolly.com Page 2 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

aggregation switches to function as one

Features
Virtualization
iStack
iStack is Huawei’s technology to virtualize
multiple ToR switches to one logical switch.
Tolly engineers verified that 16
CE6850-48T4Q-EI switches were stacked
with a ring or line topology using the iStack
technology.
Super Virtual Fabric
Super Virtual Fabric (SVF) is Huawei’s
technology for vertical virtualization which
can virtualize the access switches and core/
logical switch.
Data Center Features
With the use of server virtualization and
CE6850 switches were stacked together
using iStack and served as the aggregation
cloud computing in data centers,
traditional networks face challenges
switch. Then, the stacked switch was
including Layer 2 network scalability issues,
virtualized with multiple CE5810 switches
limited 4,094 VLANs, increased demands
which served as access switches. See Figure
on switch MAC tables, network
1 for the topology.
requirements for FCoE traffic, and difficulty
of enforcing network policies on virtual
Tolly engineers verified that switches in SVF
machines (VMs) while they “live migrate” to
supported local forwarding on the leaf
different hosts or even different data
nodes and that the stacking links
centers.
supported link aggregation and load
balancing.
Tolly engineers verified a few features on
Tolly engineers swapped all CE5810 Huawei CE7800/6800/5800 switches to
switches in the test bed with CE6810 solve these problems. TRILL was verified to
switches and verified the same features. expand the Layer 2 network. DCB features
were verified to provide lossless Ethernet

Huawei CloudEngine 7800/6800/5800 Series Data Center Switch Power Consumption

(as reported by Chroma Programmable AC Source 6560)

Power Consum
mption (Watts)

ATIS TEER ATIS Weighted

ATIS Weighted (Gbps/Watts) Watts/Gbps
0% Traffic 30% Traffic 100% Traffic
Power

CE7850-32Q-EI
277.7 290.6 320.5 292.3 4.38 0.23
32 x 40GbE ports
CE6810-48S4Q-EI
124 130 136 130.0 4.92 0.20
48 x 10GbE + 4 x 40GbE ports
CE5850-48T4S2Q-HI
110.6 118.3 128.3 118.5 1.42 0.71
48 x GbE + 4 x 10GbE + 2 x 40GbE ports
CE5810-48T4S-EI
69.9 70.2 72.0 70.4 1.25 0.80
48 x GbE + 4 x 10GbE ports
CE5810-24T4S-EI
49.8 50.3 50.7 50.3 1.27 0.79
24 x GbE + 4 x 10GbE ports
Note: 1. Switches were fully loaded with fans and power supplies.
2. White cells are measured results. Green cells are calculated results.
3. Alliance for Telecommunications Industry Solutions (ATIS) weighted power = (power consumption with 0% traffic) x 0.1 + (power consumption with 30% traffic) x 0.8 +
(power consumption with 100% traffic) x 0.1. ATIS Telecommunication Energy Efficiency Ratio (TEER) = (maximum demonstrated throughput) / (ATIS weighted power).
4. ATIS weighted Watts/Gbps = 1 / (ATIS TEER).
5. iMIX traffic (5% 49 bytes frames, 20% 576 bytes frames, 42% 1,500 bytes frames and 33% 49-1500 bytes frames) was used.
Source: Tolly, May 2014 Table 3

© 2014 Tolly Enterprises, LLC Tolly.com Page 3 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

for FCoE. VEPA was verified to direct all
network traffic of VMs to the physical
switch for easier management. Huawei
nCenter and VSI Manager were evaluated
to provide network policy migration
following VMs’ live migration.
Tolly engineers also verified that Huawei
CE7850 and CE6850HI switches could act
as the VXLAN overlay network tunnel
endpoint and gateway.
Tolly engineers further verified that the
Huawei CE12800 data center core switch
supported Huawei’s Ethernet Virtual
Network (EVN) to provide L2 connectivity
across the L3 WAN network. This feature is
discussed in the Tolly Test Report #214119.
TRILL & High-Availability
Tolly engineers verified that Huawei
CloudEngine switches supported
transparent interconnection of lots of links
(TRILL) with a large Layer 2 TRILL network
that consisted of 512 nodes. Additionally,
engineers verified support for high
availability with active-active TRILL edge.

Huawei CloudEngine 7800/6800/5800 Series Data Center Switch

Tolly Verified Performance and Features
Tolly Certified Performance and Features
Performance

Line-rate forwarding

10GbE port latency (cut-through) as low as 0.8μs, 40GbE port latency (cut-through) as low as 0.6μs

Low power consumption

Virtualization

iStack, virtualize 16 physical switches into one logical switch
Super Virtual Fabric (SVF) vertical virtualization (virtualize aggregation and access switches into one) with local


forwarding on leaf nodes
Data Center Featuress
Transparent Interconnection of Lots of Links (TRILL): support a large L2 network with up to 512 nodes,


High availability with TRILL: active-active TRILL edge - two nodes as a DFS group with one pseudo TRILL nickname
FCoE (FCF, NPV, FSB modes)

Data Center Bridging (DCB) - PFC, ETS, DCBX
(not include CE5800 switches)

802.1Qbg Virtual Edge Port Aggregator (VEPA)
Network Policy Migration: Controlled by Huawei nCenter, interoperate with VMware vCenter to implement in-service


policy migration with virtual machine live migration
Network Policy Migration: Controlled by Huawei VSI Manager, interoperate with VMware vCenter and IBM 5000V


virtual distributed switch using VEPA to implement in-service policy migration with virtual machine’s live migration

Underlying network for VXLAN or NVGRE overlay network
CE7850 and CE6850HI acted as the VXLAN Tunnel Endpoint (VTEP)


CE7850 and CE6850HI acted as the VXLAN overlay network gateway

OpenFlow SDN

Controlled by the Huawei Agile Controller or Third Party SDN Controllers (tested Ryu)
Topology discovery, L2/L3 line-rate forwarding, multiple flow tables, policy-based routing, dynamic traffic


engineering (TE)

Source: Tolly, May 2014 Table 4

© 2014 Tolly Enterprises, LLC Tolly.com Page 4 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

FCoE/DCB
Huawei Super Virtual Fabric (SVF) Test Bed
Engineers verified support for a key set of
data center functionality in the areas of
Fibre Channel over Ethernet (FCoE) with
data center bridging (DCB). The Huawei
CloudEngine switches supported FCF, NPV
and FSB modes for FCoE. Engineers also
verified the interoperability between the
Huawei CloudEngine switches and the
CNAs from major vendors including
Emulex, QLogic and Intel. See Table 4 and
the Test Methodology section for additional
details.

VEPA and Network Policy Migration

Tolly engineers verified interoperability
between Huawei nCenter and VMware
vCenter to implement in-service policy
migration with virtual machine migration.
When a virtual machines was live migrated
to another host, the network policy (ACL
rules and QoS policies) for the virtual
machine group also migrated to the proper
different switch or port.
Engineers also verified interoperability
between Huawei VSI Manager, VMware
vCenter and the IBM 5000V virtual
distributed switch using 802.1Qbg Virtual
Edge Port Aggregator (VEPA) to implement
in-service policy migration with virtual
machine live migration.
Overlay Network - VXLAN and
NVGRE
Two major data center overlay network
technologies are Virtual Extensible LAN
(VXLAN) and Network Virtualization using
Generic Routing Encapsulation (NVGRE).
The overlay network technologies can
provide Layer 2 connectivity for tunnel
endpoints (e.g virtual switches) over a
physical Layer 3 network. It can expand the
Layer 2 network for the virtual machines,
Source: Tolly, May 2014
overcome the limitation of VLAN numbers
by adding a new Layer 2 network segment
header (VNI for VXLAN and VSI for NVGRE),
and reduce the demands of the MAC tables
on the physical switches.
As the underlying physical network only
needs to provide Layer 3 connectivity for
the tunnel endpoints (e.g. virtual switches),
the physical switches do not need to
change much. Huawei CE7800/6800/5800
acted as the underlying network in the
VXLAN and NVGRE overlay network
environment during the test.
To allow the virtual environment using
VXLAN or NVGRE to communicate with
other non-VXLAN or non-NVGRE endpoints
as well as provide Layer 3 connectivity for
VXLAN or NVGRE endpoints in different
network segments, a gateway is needed.
Tolly engineers verified that the Huawei
CE7850 and CE6850-HI switches could act
as the gateway for the VXLAN overlay
Figure 1
network while Huawei CE12800 switch
could act as the gateway for the VXLAN or
NVGRE overlay network
OpenFlow Software
Defined Networking
Tolly engineers verified various capabilities
in the area of software defined networking
(SDN).
Topology Discovery
The Huawei Agile Controller supports
displaying the whole network topology by
the LLDP topology discovery capability
from the CE7800/6800/5800 switches.
Third-Party SDN Controller with
Multiple Flow Tables
In addition to verifying SDN management
via the Huawei Agile Controller, engineers
also verified that the Huawei devices could

© 2014 Tolly Enterprises, LLC Tolly.com Page 5 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches
be managed by a third-party controller - in
this test, the Ryu SDN Framework. Multiple
flow tables were applied to the
CloudEngine switches from the Ryu
controller.
Flow Table Performance
Engineers verified that the CloudEngine
switches delivered line-rate Layer 2 and
Layer 3 performance with two 10GbE ports
using SDN-based flow tables.
Policy-based Routing
Tolly engineers verified that policy controls
could be used to route traffic through
specific switches as configured via SDN.
Dynamic Traffic Engineering
Tolly engineers verified that dynamic traffic
engineering (TE) could be used to adjust
the forwarding path dynamically based on
traffic load.
Test Setup &
Methodology
Performance
Throughput
CE7850-32Q-EI, CE6850-48S6Q-HI,
CE6810-48S4Q-EI, CE5850-48T4S2Q-HI,
CE5810-48T4S-EI and CE5810-24T4S-EI were
tested using the RFC2544 throughput test
suite in the Ixia IxNetwork. For each device
under test (DUT), all available same type of
ports were in snake topology. See Table 1
for results.
Latency
Cut-through latency (FIFO) of the
CE7850-32Q-EI and CE6810-48S4Q-EI was
measured as port to port with line-rate
© 2014 Tolly Enterprises, LLC
traffic using the RFC2544 latency test suite
in the Ixia IxNetwork. Store-and-forward
latency (LIFO) was measured for the
CE6850 switch using the RFC2544 latency
test suite in the Spirent TestCenter. Thus,
store-and-forward results reported here do
not include the time required to store the
frame. See Table 2 for results.
Power Consumption
The power consumption was measured
using the same traffic topology as the
throughput test. According to the ATIS
standard for data center switches, the
power consumption of 0% traffic, 30%
traffic and 100% traffic were measured
using iMIX traffic (5% 49 bytes frames, 20%
576 bytes frames, 42% 1,500 bytes frames
and 33% 49-1500 bytes frames). Then the
ATIS weighted power, ATIS TEER and ATIS
weighted Watts/Gbps of the switch were
calculated. See the notes of Table 3 for
additional details.
ATIS standard refers to the “Energy
Efficiency for Telecommunications
Equipment: Methodology for
Measurement and Reporting for Router
and Ethernet Switch Products” document
published by Alliance for Telecommunications
Industry Solutions (https://www.atis.org/
docstore/product.aspx?id=25324).
Virtualization
iStack
iStack is Huawei’s technology to virtualize
multiple ToR switches to one logical switch.
Tolly engineers verified that 16
CE6850-48T4Q-EI switches were stacked as
a ring or line topology using the iStack
technology.
Tolly.com
#214120
SVF
Super Virtual Fabric (SVF) is Huawei’s
technology for vertical virtualization which
can virtualize the access switches and core/
aggregation switches to be one logical
switch.
CE6850 switches were stacked together
using iStack and acted as the aggregation
switch. Then the stacked switch was
virtualized with multiple CE5810 switch
which acted as access switches. See Figure
1 for the topology.
Tolly engineers verified that switches in SVF
support local forwarding. Also, the stacking
links between the aggregation switch and
the access switches supported load
balancing.
Tolly engineers swapped all CE5810
switches in the test bed into CE6810
switches and verified the same features.
Data Center Features
TRILL
Transparent Interconnection of Lots of
Links (TRILL) uses Layer 3 routing
techniques to build a large Layer 2
network. Engineers used Spirent TestCenter
to simulate one TRILL node on one port
and 510 TRILL nodes on the other port.
Both ports were connected to the
CloudEngine switch under test. Tolly
engineers verified that the switch under
test showed all 511 TRILL neighbors.
Together with the switch under test, the
whole TRILL network included 512 nodes.
CE7850, CE6850 and CE5850 switches were
all tested.
Engineers also configured two CE6850
switches in active-active status as a DFS
group with a pseudo TRILL nickname for
Page 6 of 13
 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

high availability and verified the fast
failover for switch and link failures.
FCoE
Tolly engineers verified that Huawei
CE6850 could act in FCF, NPV or FSB mode
for FCoE. CNAs from major vendors
including Emulex, QLogic and Intel were
used during the test to verify Huawei
CE6850 switch’s interoperability with them.
FCoE - FCF Mode
When an Fibre Channel over Ethernet
(FCoE) switch operates in the FCoE
Forwarder (FCF) mode, it encapsulates FC
frames in Ethernet frames and uses FCoE
virtual links to simulate physical FC links. It
provides standard FC switching capabilities
and features on a lossless Ethernet
network.
Tolly engineers verified that the
CE6850-48S4Q-EI switch supported FCF
mode single-hop as well as multi-hop. In
the single-hop test, only one CE6850
switch was used to connect the SAN
storage and the physical server. In the
multi-hop test, two CE6850 were used to
connect the SAN storage and the physical
server in serial. In both tests, the physical
server can mount and access the LUNs in
the SAN storage without any problem.
One Emulex OneConnect OCe11102-FM
dual-port 10G/s FCoE Converged Network
Adapter (CNA) was used on the physical

Huawei CloudEngine Switch Performance Test Bed

CE7850-32Q-EI

CE6850-48S6Q-HI

CE6850-48S4Q-EI

CE6810-48S4Q-EI

CE5850-48T4S2Q-HI

CE5810-48T4S-EI Ixia XM12 IP Performance Tester

CE5810-24T4S-EI

Note: The same type of ports on one switch were in snake topology: All available 40GbE ports in snake topology, all available 10GbE ports in snake
topology and all GbE ports in snake topology

Source: Tolly, May 2014 Figure 2

© 2014 Tolly Enterprises, LLC Tolly.com Page 7 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches
server to connect to the CE6850 switch
under test.
FCoE - NPV Mode
An Fibre Channel Storage Area Network (FC
SAN) needs a large number of edge
switches that are directly connected to
nodes (servers and storage). FCoE switches
in FCoE N-port Virtualization (NPV) mode
can expand the number of switches in an
FC SAN.
Fabric is the main network with FCoE
switches in FCF mode. NPV switches reside
between nodes and core FCoE FCF
switches on the edge of the fabric. NPV
switches forward FCoE traffic from it’s
connected nodes to the core FCF switch.
The NPV switch appears as an FCF switch to
nodes and as a node to the core FCF switch.
Tolly engineers verified that the Huawei
CE6850 switch could work in NPV mode
and interoperate with the Brocade
VDX6700 FCoE switch in FCF mode. One
physical server was connected to the
Huawei CE6850 switch while one SAN
storage was connected to the Brocade
VDX6700 FCoE switch. FCoE traffic was
forwarded to the Brocade VDX6700 switch
by CE6850 and the physical server access
the SAN storage without any problem.
One QLogic 8200 series 10Gbps CNA was
used on the physical server to connect to
the Huawei CE6850 switch.
FCoE - FSB Mode
FCoE switch in FCoE Initialization Protocol
Snooping Bridge (FSB) mode does not
support FC protocol itself. It uses FCoE
Initialization Protocol (FIP) snooping to
prevent attacks.
One port of Spirent TestCenter simulating a
server (FCoE initiator) was connected to
one Huawei CE6850 switch in FSB mode.
© 2014 Tolly Enterprises, LLC
The CE6850 switch was then connected to
one Huawei CE12800 switch in FCF mode.
Another port of Spirent TestCenter was
connected to the CE12800 switch on the
other end to simulate one SAN storage
(FCoE target).
The FCoE session was created between the
simulated FCoE initiator and target. The
CE6850 then stores the FCoE session
information by FIP snooping.
Tolly engineer then verified that only FCoE
traffic matching the MAC address of the
simulated SAN storage could be forwarded
to the FCoE initiator. FCoE traffic not
matching the MAC address of the FCoE
session or other types of non-FCoE traffic
matching the MAC address can neither be
passed to the FCoE initiator by the CE6850
switch in FSB mode.
Engineers also used real physical server and
SAN storage with the CE6850 in FSB mode
and CE12800 in FCF mode to verify the
connectivity. The physical server could
access the storage without any problem.
One Intel CNA was used on the physical
server to connect to the CE6850 switch.
Data Center Bridging (DCB)
Data Center Bridging (DCB) is a suite of IEEE
standards to provides many advantages for
data centers such as lossless Ethernet for
FCoE traffic. Tolly engineers verified DCBX,
PFC and ETS which are components of DCB
for the Huawei CE6850 switch.
DCB - DCBX
Data Center Bridging Capability Exchange
protocol is an extension of Link Layer Data
Protocol (LLDP) to discover peers and
exchange configuration information
between DCB-compliant switches.
Tolly engineers verified that the Huawei
CE6850 switch could use DCBX protocol to
Tolly.com
#214120
negotiate ETS and PFC settings. Only when
the ETS and PFC settings match between
the CE6850 switch and the connected
Spirent TestCenter, ETS and PFC could
function.
DCB - PFC and ETS
When Priority-based Flow Control (PFC) is
enabled on a switch port for inbound traffic
with certain 802.1p priorities, the port
sends back-pressure signals to reduce the
sending rate of those priorities from the
upstream device if network congestion
occurs.
Enhanced Transmission Selection (ETS)
implement QoS based on Priority Groups
(PG). In the configuration of ETS, engineers
need to match 802.1p priority 0 to 7 to PG0,
PG1 and PG15 offered by the Huawei
CE6850 switch. PG0, PG1 and PG15 use PQ
+DRR. PG15 uses Priority Queuing (PQ) to
have limitless bandwidth and host
management or IPC traffic. PG0 and PG1
use Deficit Round Robin (DRR) to have
weighted bandwidth left by PG15.
In the test, engineers enabled PFC for
priority 3 (default priority for FCoE traffic) to
make sure FCoE traffic will not have frame
loss. Then priority 0, 1, 2, 4 and 5 were
assigned to PG0; priority 3 (FCoE traffic) was
assigned to PG1; priority 6 and 7 (IPC traffic)
was assigned to PG15. The weight ratio for
PG1 (FCoE traffic) and PG0 (LAN traffic) was
set to 3:2.
There was one 10GbE link between the
CE6850 switch under test and the receiving
port of the Spirent TestCenter. Engineers
sent 8Gbps IPC traffic with 802.1p priority 7,
4Gbps FCoE traffic with priority 3 and 2Gbps
regular Ethernet traffic with priority 0 using
two 10GbE ports of Spirent TestCenter to
bypass the bandwidth of the 10GbE link at
the receiving end.
Page 8 of 13
 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

Tolly engineers then verified that the

receiving end received all 8Gbps IPC traffic Huawei CloudEngine Switches Network Policy Migration Test Bed A
which was in PG15 to have limitless
bandwidth, 1.2Gbps FCoE traffic which was
in PG1 and 0.8Gbps regular Ethernet traffic
which was in PG0. The receiving rate of
1.2Gbps FCoE traffic and 0.8Gbps regular
Ethernet traffic matched the configured 3:2
weight ratio for PG1 and PG0.

Tolly engineers also verified that the

sending rate from TestCenter’s sending
ports of the FCoE traffic reduced to be as
1.2Gbps as total (0.6Gbps from each port)
because PFC was enabled. Thus the FCoE
traffic had 0 frame loss. The sending rate of
the regular Ethernet traffic with priority 0
was still 2Gbps and had frame loss because
priority 0 was not enabled for PFC.

Network Policy Migration

Huawei provides two tools to help migrate Source: Tolly, May 2014 Figure 3
the network policy (ACL and QoS rules)
along with virtual machines’ live migration
in VMware vSphere environment.
Huawei CloudEngine Switches Network Policy Migration Test Bed B
Network Policy Migration - nCenter
The first tool is the nCenter component
under Huawei’s eSight net work
management application. Engineers first
configured nCenter to manage the
CE12800 and the CE6850 switch under test.
In nCenter, engineers then configured the
I P a d d re s s o f V M w a re v Ce n t e r
5.0.0_623373 which managed two
VMware ESXi 5.0.0_623860 hosts. So
nCenter could use vCenter’s APIs to
interoperate with vCenter and migrate
network policies on Huawei CloudEngine
switches along with the virtual machines.
See Figure 3 for the test bed.

One ACL policy (denying destination IP as

an outbound policy) were assigned to the 5000V refers to IBM 5000V virtual distributed switch
VM group which contains one VM on the
first ESXi host. Engineers live migrated the Source: Tolly, May 2014 Figure 4

© 2014 Tolly Enterprises, LLC Tolly.com Page 9 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

VM to another ESXi host and used ping to
check the connectivity.
The VM was accessible all the time on the
network with the traffic not matching the
ACL (deny policy). Tolly engineers verified
that the ACL policy were migrated and
enforced to the proper different switch
with the VM’s migration.
Network Policy Migration - VSI
Manager with VEPA
The method above can control the ACL and
QoS policies between different hosts but
cannot control the traffic between VMs on
the same host and in the same VLAN
because the traffic only goes into the
vSwitch on the host without passing to the
physical Huawei CE6850 switch.
Virtual Edge Port Aggregator (VEPA)
standard is developed to direct all network
traffic of any virtual machine to the physical
switch. So the ACL and QoS policies on the
physical switch can control all network
traffic of any virtual machine. The built-in
vSwitch in VMware ESXi host does not
support VEPA natively. So, engineers
installed the IBM distributed virtual switch
5000V on the VMware ESXi hosts and
enabled VEPA for the 5000V. Then
engineers configured Huawei VSI Manager
to work with IBM 5000V, VMware vCenter
and the CE6850 switch under test. See
Figure 4 for the test bed.
Overlay Network Gateway - VXLAN
As shown in Figure 5, one Huawei CE7850
switch and one Huawei CE6850HI switch
(CE7850-2 and CE6850-2 in the test bed)
acted as the VXLAN Tunnel End Points
(VTEP). The CE7850 or CE6850HI switch at
top acted as the VXLAN overlay network
gateway.
Engineers first verified the Layer 2 and
Layer 3 connectivity in the VXLAN network.
When VTEP1 and VTEP2 were with the
same network segment header VNI and the
two Spirent TestCenter ports were in the
same subnet, the TestCenter ports were
able to communicate with each other
using the VXLAN network. When VTEP1
and VTEP2 were with different network
segment header VNIs (in two different
VXLAN network segments) and the two
Spirent TestCenter ports were in different
subnets, the CE7850 or CE6850 switch at
top acted as the gateway and provided
VXLAN network Layer 3 connectivity
Huawei CloudEngine Switches VXLAN Gateway Test Bed
between VTEP-1 and VTEP-2 so the two
TestCenter ports could always communicate.
Engineers then verified the Layer 2 and
Layer 3 connectivity between the VXLAN
overlay network and the traditional
network. CE6850HI-1 in the test bed
simulated the traditional network out of
the VXLAN overlay network. The CE7850 or
the CE6850HI switch at the top could use
its port connected to CE6850-1 as a VTEP
and provide Layer 2 and Layer 3
connectivity between the VXLAN overlay
network and the traditional network.
OpenFlow SDN
SDN - Topology Discovery
Two Huawei CE6850 switches and one
Huawei CE7850 switch were configured to
connect to the Huawei Agile Controller.
Wireshark was used to capture the traffic
between the controller and the switches.

One ACL policy (denying destination IP as

an outbound policy) were assigned to the
VM group which contains one VM on ESXi
Host B. Engineers live migrated the VM
from ESXi Host B to ESXi Host C and using
ping to check the connectivities.

Tolly engineers verified that the ACL policy

were migrated and enforced to the proper
different port of the CE6850 switch with
Source: Tolly, May 2014 Figure 5
the VM’s migration.

© 2014 Tolly Enterprises, LLC Tolly.com Page 10 of 13

 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches
Tolly engineers verified that the Hello,
Features_Request, Features_Reply,
Set_Config, Get_Config_Request,
Get_Config_Reply, Multipart_Request,
Multipart_Reply, Packet_In, Flow_Mode
and FLow_Removed packets with the
OpenFlow 1.3 protocol were all captured.
OpenFlow protocol header for the packets
all included version 0x04 which is defined
as the OpenFlow 1.3 protocol.
The Huawei Agile Controller also used LLDP
to discover the topology of the switches.
Tolly engineers verified that the topology
map in the Huawei Agile Controller shows
the network topology and topology
change accurately.
SDN - Multiple Flow Tables with
Third Party SDN Controllers
Tolly engineers verified that the Huawei
CE6850 and CE7850 could be controlled by
the third party SDN controller - Ryu SDN
Framework (http://osrg.github.io/ryu/) 3.8.
Engineers used Wireshark to capture traffic.
Hello, Features_Request, Features_Reply,
Set_Config, Get_Config_Request,
Get_Config_Reply, Multipart_Request,
Multipart_Reply, Packet_In, Flow_Mode
and FLow_Removed packets with the
OpenFlow 1.3 protocol were all verified
between the Ryu and the CloudEngine
switches under test.
After traffic was passed to the CE6850 and
CE7850 switches, the Ryu controller
advertised multiple flow tables to the
Huawei CE6850 and CE7850 switches with
Flow_Mod packets. Engineers verified that
the flow tables were successfully applied to
the switches.
© 2014 Tolly Enterprises, LLC
SDN - Layer 2 and Layer 3 Line-rate
Forwarding
Traditional Layer 2 and Layer 3 forwarding
are based on MAC and FIB tables. When
managed the Huawei Agile Controller, the
CloudEngine switches can forward traffic
by using the flow tables applied from the
SDN controller instead of using MAC and
FIB tables.
When a switch receives traffic, it passes the
traffic to the controller. The controller learns
the MAC and IP addresses of the traffic and
uses a shortest path algorithm to calculate
the Layer 2 and Layer 3 forwarding paths
with the network topology it discovered.
For Layer 2 forwarding, the controller then
applies flow tables with the proper Output
command to each switch. So the switches
know how to forward the traffic. For Layer 3
forwarding, the controller applies flow
tables with the Output command as well as
Huawei CloudEngine Switches SDN Test Bed
Source: Tolly, May 2014
Tolly.com
#214120
as Set filed and Decrease TTL commands to
let the switches know the forwarding path
and change the MAC addresses and the
TTL value of the traffic.
Tolly engineers verified the procedure
above using two Huawei CE6850 and one
Huawei CE7850 switches with the Huawei
Agile Controller.
Line-rate traffic was used with 10GbE ports
on the switches to test. In both the Layer 2
forwarding test and Layer 3 forwarding
test, there was no frameloss with 10Gbps
128-byte frames.
SDN - Policy-based Routing
Tolly engineers verified that along with the
shortest path algorithm used by the
controller, certain policies can be defined.
In the test, engineers defined that traffic
from and to certain IPs must go through
one specific switch. Then all the matched
Figure 6
Page 11 of 13
 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches
traffic went through that switch even
though the path was not the shortest one.
Other traffic still followed the shortest
paths.
SDN - Dynamic Traffic Engineering
The Huawei Agile Controller can control the
switches to adjust the forwarding paths
dynamically according to the traffic load.
Network topology is shown in Figure 6. All
hosts were simulated by Spirent TestCenter.
The traffic between the two upper hosts
(traffic in purple and brown) had higher
priority. The traffic between the two lower
hosts (traffic in blue and green) had lower
priority.
The rate threshold was set to be 8Gbps to
fall back to the backup path. The traffic
CE7850-32Q-EI
CE6850-48S6Q-HI
C6850-48S4Q-EI
CE6810-48S4Q-EI
CE5850-48T4S2Q-HI
CE5810-48T4S-EI
CE5810-24T4S-EI
Source: Tolly, May 2014
© 2014 Tolly Enterprises, LLC
#214120
between the upper two hosts was with
bidirectional 4Gbps. The traffic between
the lower two hosts was increased from
0Gbps to 10Gbps and then fell back to
3Gbps granularly.
Tolly engineers verified that, when traffic
between the lower two hosts reached
4Gbps, the traffic path changed to the
backup path calculated by the controller as
shown in Figure 6. When the traffic
dropped to less than 4Gbps, the traffic path
went back to be the shortest path.
Devices Under Test
CloudEngine 7850EI V100R003C00SPC600
CloudEngine 6850HI V100R003C00SPC600
CloudEngine 6850EI V100R003C00SPC600
CloudEngine 6810EI V100R003C00SPC600
CloudEngine 5850HI V100R003C00SPC600
CloudEngine 5810EI V100R003C00SPC600
CloudEngine 5810EI V100R003C00SPC600
Table 5
Tolly.com Page 12 of 13
 Huawei CloudEngine 7800/6800/5800 Series Data Center Switches #214120

Test Equipment Summary

About Tolly The Tolly Group gratefully acknowledges the providers
of test equipment/software used in this project.
The Tolly Group companies have been
delivering world-class IT services for more Vendor Product Web
than 25 years. Tolly is a leading global
provider of third-party validation services
for vendors of IT products, components XM12 Chassis
Ixia
and services. IxNetwork 7.22.9.9.9EA
http://www.ixiacom.com
You can reach the company by E-mail at
sales@tolly.com, or by telephone at
+1 561.391.5610. HWS-11U-KIT Chassis
Spirent TestCenter v3.95
Visit Tolly on the Internet at: TestCenter v4.20.0576.0000 http://www.spirent.com
http://www.tolly.com

Terms of Usage
This document is provided, free-of-charge, to help you understand whether a given product, technology or service merits additional
investigation for your particular needs. Any decision to purchase a product must be based on your own assessment of suitability
based on your needs. The document should never be used as a substitute for advice from a qualified IT or business professional. This
evaluation was focused on illustrating specific features and/or performance of the product(s) and was conducted under controlled,
laboratory conditions. Certain tests may have been tailored to reflect performance under ideal conditions; performance may vary
under real-world conditions. Users should run tests based on their own real-world scenarios to validate performance for their own
networks.
Reasonable efforts were made to ensure the accuracy of the data contained herein but errors and/or oversights can occur. The test/
audit documented herein may also rely on various test tools the accuracy of which is beyond our control. Furthermore, the
document relies on certain representations by the sponsor that are beyond our control to verify. Among these is that the software/
hardware tested is production or production track and is, or will be, available in equivalent or better form to commercial customers.
Accordingly, this document is provided "as is," and Tolly Enterprises, LLC (Tolly) gives no warranty, representation or undertaking,
whether express or implied, and accepts no legal responsibility, whether direct or indirect, for the accuracy, completeness, usefulness
or suitability of any information contained herein. By reviewing this document, you agree that your use of any information contained
herein is at your own risk, and you accept all risks and responsibility for losses, damages, costs and other consequences resulting
directly or indirectly from any information or material available on it. Tolly is not responsible for, and you agree to hold Tolly and its
related affiliates harmless from any loss, harm, injury or damage resulting from or arising out of your use of or reliance on any of the
information provided herein.
Tolly makes no claim as to whether any product or company described herein is suitable for investment. You should obtain your own
independent professional advice, whether legal, accounting or otherwise, before proceeding with any investment or project related
to any information, products or companies described herein. When foreign translations exist, the English document is considered
authoritative. To assure accuracy, only use documents downloaded directly from Tolly.com.  No part of any document may be
reproduced, in whole or in part, without the specific written permission of Tolly. All trademarks used in the document are owned by
their respective owners. You agree not to use any trademark in or as the whole or part of your own trademarks in connection with
any activities, products or services which are not ours, or in a manner which may be confusing, misleading or deceptive or in a
manner that disparages us or our information, projects or developments.

214120 ivcofs3 yx-2015-02-02-VerK

© 2014 Tolly Enterprises, LLC Tolly.com Page 13 of 13