HC13081 - 01 Cloud Computing Planning Design

Cloud Computing
Planning Design
www.huawei.com
Copyright 2017 Huawei Technologies Co., Ltd. All rights reserved.

Objectives
Upon completion of this course, you will be able to:
Be familiar with the cloud computing scenarios and requirement
analysis.
Be familiar with the data center (DC) virtualization scenarios and

requirement analysis.
Analyze and write technical proposals according to project

requirements.
Copyright 2017 Huawei Technologies Co., Ltd. All rights reserved. Page 2
Contents
Design Objectives and Principles
Requirement Analysis Methods
General Solution Design
Management System Design
Storage System Design
Network Solution Design
Disaster Recovery (DR) Solution Design
Backup Solution Design
O&M Solution Design
Case Study
Cloud Computing Solution Design
Objectives and Principles
4. High performance and
reliability
1. Rational architecture
Main hardware and software have
The general solution has a clear
been deployed and verified on a
and rational architecture, and the
large-sized network. The
architecture technologies of each
redundancy deployment is provided
sub-solution are mature and easy
to ensure performance and
to deploy.
reliability.
2. E2E security
5. Centralized
Architecture security is ensured by
hierarchical security protection. management and O&M
APP security is ensured by The centralized cloud
mainstream anti-virus software. management platform supports
Data security is ensured by multiple creation and management of
data protection mechanisms. various resources on a unified
interface, which improves the
O&M efficiency.
3. Compatibility and scalability 6. Standard project implementation

The system is highly compatible, and and management
supports smooth expansion of functional Optimized project processes
modules in the future.
Comprehensive project organizational
assurance
Standard and Professional Service
Processes
Requirement survey and Deployment and
Planning and design Acceptance check
information collection implementation
5 FusionSphere
10 FusionSphere 14 Acceptance
Key component
2 Data configuration of key scheme
1 design
Information components development
Requirement
collection
survey 6
11 15 Acceptance
Computing design scheme
Data configuration of
3 servers implementation
Compatibility 7
analysis Network design
16 Submission of
12
Network data acceptance
configuration reports
8
4 Storage design
Service indicator
analysis 17
13 Service document
Storage data archiving
9 configuration
System resource
requirements
Project management Collection of comprehensive Design of complete and accurate Professional supervision of project
and professional services information solutions operation
Contents
O&M Solution Design
Case Study
Cloud Computing Requirement Survey
and Information Collection
2. Service 1. Service
migration system
requirements cloudification
requirements
1
3. New XXX Project Cloud
service
deployment Computing Requirement
requirements Requirement Survey Report
survey template Output
documents
or PoC test
2
4. Backup
requirements
XXX Project Cloud
6. Security
Computing Acceptance
5. Networking
requirements Test Report
requirements
FusionSphere Integration Design
Information Collection
System
information
Network
topology
Collection of
Collection
cloudification
of basic
requirements Customer system
information
information and
specification
Obtain the requirements
customer's - Information about
Information about
Basic project network topology common IT components,
services to be
information - Determine the position such as the domain,
cloudified of FusionSphere on
- Information about main NTP, and DNS
- Server information of the customer's
industrial APPs used by - Network device
services to be network.
the customer information
cloudified - Plan information about - Server storage device
- Server, storage, and - Service type cabinets used by information
network device information of services FusionSphere. - Reused device
information to be cloudified
information
- Whether servers are - Service association
- Device naming rules
reused between APPs
- Whether DR, backup, - Service DR and
and migration is required backup requirements
Contents
O&M Solution Design
Case Study
Functional Architecture of the Huawei
Distributed Cloud DC
Computing service Storage service Network VDC BC&DR service
DR service Operation O&M
Elastic computing Elastic block service service
management management
domain
Service
Bare metal service storage VPC Backup service

vFW platform platform
IaaS
Alarm orchestration
User management
IaaS service platform
orchestration
management
Resource
Resource
Resource Service Multi-pool and Standard
Auto scaling VDC
SLA metering multi-site API
adaptation
Data protection
Service
BCManager eBackup
management
Performance
presentation
scheduling
Resource
Topology
Service
catalog
OpenStack OpenStack OpenStack
ce pool
Resour
Virtual cluster Virtual cluster Virtual cluster

SDC SDC SDC
consumption
management
Resource
condition
Resource
SDS
metering
SDS SDS
Energy
Health
SDN SDN SDN
Infrastructure
Server Storage Network DCI Server Storage Network DCI Server Storage Network
Typical Cloud Computing Solution
Scenarios
Server virtualization NFVI Cloud DC
1 The FusionSphere server virtualization 1 The FusionSphere NFVI cloud 1 The FusionSphere infrastructure cloud
is based on the Huawei UVP protocol platform is based on the OpenStack platform is based on the OpenStack
solution. architecture. architecture.
2 Its main components include 2. Functionally, this cloud platform is 2. Functionally, this cloud platform is
FusionCompute, divided into the infrastructure, divided into OpenStack,
OpenStack, FusionManager, FusionCompute, FusionStorage, and
FusionManager, FusionSphere SOI, FusionNetwork.
eBackup, and Ultra VR. FusionCare, and standard northbound
interface.
FusionSphere virtualization solution Standard

Upper-layer application
Customers third-
interface FusionSphere system
party system
Database
APP
Video APP Web APP
Cluster
APP
Securities
APP FusionManager FusionCare
Unified northbound API
DR system Backup system

UltraVR (array-based replication and host-based
eBackup virtual backup
replication_
Performance insight system Health check and log collection

FusionSphere SOI FusionCare FusionSphere OpenStack
Cloud management software
VSAM
VSA OpenStack management node
FusionManager (active/standby)
VSA
Virtual resource
Virtual network service Computing Storage Network
schedule
Physical cluster Physical cluster virtualization virtualization virtualization
FusionCompute FusionCompute management
system system
VRM (active/stadnby) VRM (active/stadnby)
FusionSphere FusionCare
eBackUp
Computing
subsystem
Network
subsystem
Storage
subsystem
Computing
subsystem
Network
subsystem
Storage
subsystem SOI
VM VM VM VM Infrastructure
VM VM VM VM
Computing Computing
resource resource
Network Network
FusionCompute FusionStorage FusionNetwork
resource resource
Storage
resource
Storage
resource
Server Storage Network Security
Server Virtualization Solution
System cluster planning
FusionSphere system deployment (1) Management cluster: including Dom0 and management VM
planning. Management VMs include VRM, FusionManger,
Non-functional FusionSphere SOI Network service
deployment deployment FusionCare, FusionStorage Manager, UltraVR, and eBackup.
Reliability Performance Anti-virus FusionCare
(2) User cluster: including Dom0 and user VM planning
Security Backup DR VSAM VSA
Non-functional planning
FusionManager deployment Reliability, performance, and security planning of key components
Network service planning
FusionCompute
deployment
(1) VDC planning: providing the VDC
(2) VPC planning: providing the private network service
VRM deployment
(3) VSA planning: providing the DHCP service VMs
(4) VSAM planning: VSA VM management node
Cluster deployment
Other planning
Dom0 Dom0 Dom0
(1) FusionSphere SOI: managing the performance monitoring and
Mgn_VM01 User_VM01 User_VM11
Mgn_VM02 User_VM02 User_VM12

analysis VMs
(2) FusionCare planning: providing environment health check VMs
CNA01 CNA01 CNA11
(3) Virtual anti-virus planning: providing parallel anti-virus
Dom0 Dom0 Dom0
Mgn_VM03 User_VM03 User_VM13 functions, and supporting Trend Micro and Rising anti-virus
Mgn_VM04 User_VM04 User_VM14 software of specific OSs
CNA02 CNA02 CNA12 (4) DR planning: DR solution planning for the matching UltraVR
... ... ... software
Management cluster User cluster 1 User cluster 2 (5) Backup planning: backup solution planning for the matching
eBackup software
FusionSphere NFVI Solution
Standard
Non-functional requirement design (Reliability, Security, Performance, interface
and Service cloudification) O&M/operation
system
Controller Controller Controller
FusionManager
Compute Compute
(KVM-01) ... (KVM-N)
Infrastructure (server, storage, network, storage, security, and load FusionCare

balancing)
FusionSphere Cloud DC Solution
Contents
O&M Solution Design
Case Study
OpenStack Management Node
Specification Design
< 20PM, < 200VM
Optional Software
3 controller nodes Optional node
FusionStorage Manager VM
VRM VM
FusionManager VM
2050PM, 2001000VM
3 controller nodes
VRM VM
FusionManager VM
50100PM, 10003000VM
2 database 2 VRM
3 controller nodes 2 manager VM nodes
nodes nodes
FusionManager VM
100256PM, 30005000VM
2 database 2 VRM 2 rabbitMQ

3 controller nodes 2 manager VM nodes
nodes nodes nodes
FusionManager VM
256512PM, 5000VM
2 database 2 VRM 2 rabbitMQ 4 Extend Controller

3 controller nodes 2 manager VM nodes nodes
nodes nodes nodes
FusionManager VM
OpenStack Management Node 50
PMs Controller 01
3*Controller
Controller 02 Controller 03
FusionManager FusionManager
VM VM
dhcp dhcp
router router
auth auth auth
image image image
controller controller controller
sys-server sys-server sys-server
measure measure measure
mongodb mongodb mongodb
rabbitmq rabbitmq
database database
swift swift swift
zookeeper zookeeper zookeeper
compute compute compute
sys-client sys-client sys-client
blockstorage-driver blockstorage-driver blockstorage-driver
OpenStack Management Node 50 to
100 PMs
2*Database
3*Controller
Controller 01 Controller 02 Controller 03

dhcp
VM VM
router dhcp
auth router auth
image auth image
controller image controller
sys-server controller sys-server
measure sys-server measure
mongodb measure mongodb
rabbitmq mongodb rabbitmq
swift swift swift
zookeeper zookeeper zookeeper database
compute compute compute compute
sys-client sys-client sys-client sys-client
256 PMs
3*Controller 2*RabbitMQ 2*Database
Controller 01 Controller 02 Controller 03

VM VM
dhcp dhcp
router router
auth auth auth
image image image
controller controller controller
sys-server sys-server sys-server ...
measure measure measure
mongodb mongodb mongodb
swift swift swift
zookeeper zookeeper zookeeper rabbitmq database
compute compute compute compute compute
sys-client sys-client sys-client sys-client sys-client
512 PMs
3*Controller 4*Extend Controller 2*RabbitMQ 2*Database
FusionManager
VM
dhcp
router
auth
image
controller
sys-server
measure
mongodb
swift measure
zookeeper controller rabbitmq database
compute compute compute compute
sys-client sys-client sys-client sys-client
blockstorage-driver blockstorage blockstorage blockstorage
FusionCompute Management Node
Design
FusionCompute Management Node
Design
VRM Interconnected Third Network
User Scale Data Storage Deployment
Specifications Party Adapter
CPU Memory CPU Memory
< 20 PMs, < 200 Two SAS disk groups, VM

4U 5 GB 4U 5 GB 1
VMs (300 GB) RAID1 deployment
Shared storage, four

SAS disk groups, (300
2050 PMs, 200 VM
4U 5 GB 8U 8 GB 1 GB) RAID1, providing
1000 VMs deployment
independent data
storage
Local disk, two SSD Physical

50100 PMs,
8U 8 GB 12 U 16 GB 1 disk groups, (300 GB) machine
10003000 VMs
RAID1 deployment

100256 PMs,
12 U 16 GB 20 U 30 GB 1 disk groups, (300 GB) machine
30005000 VMs
RAID1 deployment

256512 PMs,
20 U 30 GB 36 U 60 GB 1 disk groups, (300 GB) machine
5000 VMs
RAID1 deployment
FusionManager Planning and Design Virtualization scenario
1. FM specification design
2. FM function design
Tenant BOSS ERP system
Cloud DC scenario
Service management system (Portal) 1. Interconnected with
ManageOne?
2. FM function design
3. FM specification
FusionManager design
Virtual host VPC Parallel Physical Authority Service

computing resource pool control provisioning
Security Elastic IP Load Multi-DC Template NFVI scenario:
DR Fault
group address balancing scheduling diagnosis 1. FM specification
design
Node type FusionManager
FusionCompute FusionStorage FusionNetwork

Fewer
Maximum 5125120 512010000
than 512
Number of Users VMs VMs
VMs
Physical hardware
CPU 4U 6U 6U
Server Storage Network Security
gateway Memory 6 GB 18 GB 24 GB
Disk 80 GB 80 GB 120 GB
Storage IOPS 50 100 150
Contents
O&M Solution Design
Case Study
Data Storage Design
Data Storage Type Data Storage Description Recommended Service Type Remarks
Local disk group RAID of the

server, or the ExtX file system Desktop cloud and virtual low Reserve 20% for
Local data storage
which functions as the data I/O test service redundancy.
storage after being started
Low I/O is needed for the
desktop cloud and
LUN of the SAN storage, which virtualization. The commercial
Reserve 20% for
Virtual data storage provides the thin allocation service that stores the
redundancy.
function advanced functions such as
thin allocation, backup, and
snapshot is needed.
1. Reserve 20% for
High I/O is needed. The virtual redundancy.
LUN of the SAN storage, which
Non-virtual data service that stores the
does not provide the thin 2. Do not support
storage advanced functions such as
allocation function the raw device direct
thin allocation is not needed.
connection feature.
High I/O is needed for the
desktop cloud and
Distributed storage, which virtualization. The commercial
Reserve 20% for
FusionStorage storage provides the data storage service that stores the
redundancy.
capability advanced functions such as
thin allocation, backup, and
snapshot is needed.
Storage Parameter Design
The V3-series storage is recommended. Only the storage capacity, storage IOPS, and RAID mode need
to be planned.
Non-virtual SAN data storage LUN capacity 2 GB
LUN capacity design
Virtual SAN data storage LUN capacity: [5 GB, 64 TB]
Effective IOPS
Effective IOPS
RAID Design in
Design in Thin Remarks
Type Common Storage
Storage Mode
Mode
1/(1 + 3x%)*(1 1. The ratio of the write I/O to the random I/O is x%. The
RAID5 1/(1 + 3x%)
17%) ratio recommended for the common office scenario is 70%.
1/(1 + 5x%)*(1 2. RAID6: As the parity bit must be calculated, four steps
RAID6 1/(1 + 5x%)
17%) are required, including reading data, reading the parity bit,
writing data, and writing the parity bit. Therefore, the write
penalty of RAID5 is 4.
3. RAID6: Two parity bits exist in RAID6. Compared with
RAID5, the parity bit must be read and written twice in RAID
1 / ( 1 + x% ) * ( 1 6. Therefore, the write penalty of RAID6 is 6.
RAID10 1/(1 + x%)
17%) 3. RAID10: Due to existence of data images, the write
operation is reduced by half. Therefore, the write penalty of
RAID10 is 2.
4. For the RAID group in thin mode, the IOPS performance
decreases by 17%.
Storage Parameter Design
Available Capacity Design in Available Capacity Design in
RAID Type Remarks
Common Storage Mode Thin Storage Mode
[Single disk capacity x (Total

Single disk capacity x (Total
number of disks Number of hot
number of disks Number of hot
RAID5 standby disks Number of RAID
standby disks Number of RAID
groups) Loss of the coffer disk]
groups) Loss of the coffer disk
x (1 + 30%)
For the RAID
Single disk capacity x (Total [Single disk capacity x (Total
group in thin
number of disks Number of hot number of disks Number of hot
mode, the storage
RAID6 standby disks Number of RAID standby disks Number of RAID
space can
groups 1) Loss of the coffer groups - 1) Loss of the coffer
increase by 30%.
disk disk] x (1 + 30%)
Single disk capacity x (Total [Single disk capacity x (Total
number of disks Number of hot number of disks Number of hot
RAID10
standby disks) x 1/2 Loss of the standby disks) x 1/2 Loss of
coffer disk the coffer disk] x 3 x (1 + 30%)
Storage Network Parameter Effective Bandwidth Formula
IPSAN in standard configuration, eight 1 GB ISCSI 8 x 1024 x (1 0.15) = 6963.2 Mbit/s

storage network adapters, 0.15 bandwidth loss
FusionStorage Planning Architecture
Strong consistency data control protocol: a variant of Multi-PAXOS, with optimized high-efficiency I/O
I/O request sequence based on the primary OSD node for consistency
Support for F redundancy; Fail Stop (F+1 replication tolerance)
After the OSD node is added to the cluster, the lost I/O must be retrieved from the primary OSD node again
(State transfer to catch up before rejoin)
Software functional module:
ZK&MDC: ZookeeperMetadata Controller: Be responsible for components within the cluster (OSD and status
view control).
OSD: Object Storage Device: Be responsible for data distribution, replication, and rebuild.
VBS: Virtual Block System: Be responsible for receiving the I/O requests from VMs.
Control flow
FusionStorage Planning Deployment
Data flow
PAXOS status Status Driver SCSI

view
layer VBS VBS VBS
VBS
ZK
MDC
ZK MDC
Status view Data Stateless interface layer
MDC flow
ZK
OSD OSD OSD OSD OSD OSD OSD OSD OSD

Data replication
RSM (Replicated State Machine) data layer
Two deployment modes are
available: integrated
deployment and separated
deployment. Separated
deployment is
recommended.
(2)
FusionStorage Planning Networking
Core Core
Customer Side
2*10 GE 2*GE
2*10 GE
GE management
Huawei Device Side Switch switch: 48 ports
Trunk
2*10 GE 2*10 GE
GE GE GE GE GE
Storage Node Storage Node Storage Node
...
40 GE
RH2288H RH2288H RH2288H
2*56GE IB 2*56GE IB 2*56GE IB

Storage Node
Service Node
Server Plane
Storage Plane
Management Plane
Principles
MDC deployment principles
Each storage host reserves 5 GB memory and CPU for one MDC process.
One storage host has a maximum of one MDC process, and every environment has at least
three MDC processes.
One MDC can process at most two resource pools, and at most 2000 OSDs concurrently.
In the RAC security level, the MDC must be deployed across cabinets.
ZK deployment principles
Three or five ZKs can be started in an environment.
ZK disk deployment:
In the standard integrated USB flash drive scenario, the ZK disk exclusively occupies one disk. By default, slot
12 of the E9000 blade is used.
The ZK disk exclusively occupies one disk and one slot. As a result, the FusionStorage capacity is affected.
ZK disk deployment:
In the cloud platform scenario, the ZK disk and OS occupy the RAID1 disk group and partition 10 of the OS.
The total storage space is 50 GB.
The ZK disk cannot be deployed on the host where the management VM is located.
The ZK disk shares the system disk and node storage space with the OS.
Principles
Resource pool deployment principles
Two copies support at most 96 disks, and three copies support at most 2000 disks. At most 50,000 disks are
supported.
The capacity and slot ID of every OSD in the resource pool must be the same.
The OSD may consist of SAS, SATA, and SSD card.
Multiple resource pools
At most 128 resource pools are supported.
A host can belong to at most three resource pools. A single OSD can belong only to a specific resource pool.
The NVDIMM cache can belong only to a resource pool. The SSD cache can belong to multiple resource pools.
VBS deployment principles

The VBS provides the SCSI interface and the iSCSI interface to external entities.
For the OLTP or OLAP, it is recommended that two VBSs be configured for every host.
In the VMWare interworking scenario, one VBS is deployed on every ESXi host, and multiple paths are
configured.
FusionStorage Planning Host
Resource
Type Requirement
CPU 8-core
Memory VBS: 4 GB
OSD size: 2.5 GB (when the disk capacity is equal to or less than 2
TB)
3.5 GB (when the disk capacity ranges from 2 TB to 4 TB)
3.5 GB (when the IB network adapter and SSD are used)
MDC: 5 GB
Storage ZK disk: a standalone disk
ZK partition: 50 GB
Network Storage plane: 10GE network or 56GE IB network
Management platform: GE network
iSCSI plane: 10GE network or 56GE IB network
Contents
O&M Solution Design
Case Study
Logical Networking in the Virtualization
Scenario
Terminal Terminal
Service network Internet Intranet

Management network
Storage network
Network connection
Firewall Firewall
Storage heartbeat
Load Load
Core layer balancer balancer
SSL VPN SSL VPN

Convergence
layer
Access
layer
Virtual network planning FusionSphere+SAN FusionSphere+SAN

Virtual networking design networking design
Virtual network service
layer planning
FusionStorage networking
design
Management Service Management IPSAN storage Management IPSAN storage

BMC plane
Network plane plane plane 22 plane plane plane
plane FusionStorage Storage Storage

storage plane heartbeat plane heartbeat plane
Computing resource pool Primary storage resource pool Backup resource
Logical Networking in the Cloud DC or
NFVI Scenario
Terminal Terminal
Internet Intranet
Firewall Firewall
Load Load
Core layer balancer balancer
SSL VPN SSL VPN
Convergence
layer
Access layer
OpenStack management node

network
Computing node network
Storage network
Primary storage resource pool
Typical SAN Storage Networking
FusionSphere+ SAN
IP SAN storage networking FC SAN storage networking

SAN storage networking
Server1 Server2
Vlan20 Vlan40
VRM Vlan20 Vlan40
Vlan30 Vlan50 CNA02
Vlan30 Vlan50 CNA01
eth0
HBA0 HBA1 HBA0 HBA1
Management
Service plane
plane
eth0 Lanswitch1 L2 network Lanswitch2 1 3 5 7 1 3 5 7

0 2 4 6 0 2 4 6
Bond0
SNS2124-1 SNS2124-2
eth1
BMC
eth4
Storage heartbeat
Bond1
plane
Controller A Controller B
eth5 SAN Vlan20 Vlan30 Vlan40 Vlan50 Vlan20 Vlan30 Vlan40 Vlan50 FC SAN
A1 A4 Controller A Controller B
eth2 Storage plane B1 - B4
eth3 Controll subrack
FusionStorage Networking
FusionSphere+
FusionStorage
10GE network + 2 network 10GE/IB network + 4 network 10GE/IB network + 6 network
ports ports ports
FSM VRM eth0

eth0 Service plane
FSM VRM eth0
eth0 Service plane
FSM VRM eth0
eth0
Service plane Management
plane
Management
Management plane
plane
eth4 eth4 eth4
eth0 eth0 eth0
eth0 Bond2 Bond2 Bond2
Host eth0 eth0
Host Host Bond0 Bond0 Bond0 eth5 eth5 eth5
Bond0 Bond0 Bond0 Host Host Host
eth1 eth1 eth1
eth1 eth1 eth1 BMC
Host Host Host
BMC BMC eth0 eth0 eth0
BMC BMC BMC eth2/ib0 eth2/ib0 eth2/ib0 Bond0 Bond0 Bond0
Bond1 Bond1 Bond1 eth1 eth1 eth1
BMC BMC BMC
eth3/ib1 eth3/ib1 eth3/ib1 eth2/ib0 eth2/ib0 eth2/ib0
Storage plane Bond1 Bond1 Bond1
eth3/ib1 eth3/ib1 eth3/ib1
Storage plane
Storage plane
Virtual Network Service Deployment
O&M Load
Firewall Firewall
network balancer
(active) (standby)
(F5)
VDC
Network Network
egress egress
Convergence
Convergence
layer
layer
(management
Stacking (service plane)
plane) Stacking
Access layer
Access layer
(management
(service plane)
plane) Stacking Stacking
VPC
VSA
Software Software Software Software

Router Gateway Firewall DHCP server
Routing Internal Direct

network network network
Resource
cluster
Service VM
Management network VSA management network Service network
Contents
O&M Solution Design
Case Study
General DR Design Solution
FC link IP link DWDM link
WAN
Primary Secondary
Cloud DC DC Egress DC
Egress router Cloud DC
egress router egress
Cloud network WDM device WDM device

GSLB
domain GSLB
SLB
Core
SLB
Core
SLB
SLB switch
switch
Access Access
switch switch
OpenStack BCManager
OpenStack VM VM VM VM
BCManager VM VM VM VM
4*controller eReplication
4*controller eReplication VM VM VM VM
VM VM VM VM
Computing domain Computing Storage

Cloud platform Storage data
domain
access data access Cloud platform
management
management
domain domain
Storage
Storage domain
domain
FucionCloud Active-Passive DR for
Private Cloud
ManageOne (SC) DR ManageOne (SC) DR Architecture characteristics
service provisioning service provisioning
component (active) component (passive)
Built based on the OpenStack
BCManager eReplication BCManager eReplication open architecture; supporting
cloud DR management cloud DR management
component component evolution; preventing lockout
Supporting tenant DR service
VM 1 VM 2 VM 3 VM 1' VM 2' VM 3'
application; reducing the
FusionSphere 6.0 FusionSphere 6.0 service provisioning period from
cloud platform cloud platform
one week to less than half an
hour
Based on the storage
replication technology; not
Replication
Storage Storage
occupying VM resources;
Storage Storage
replication
replication supporting RPO = 0; zero
Primary Cloud DC Secondary Cloud DC impact on VM performance
DR Network Design
FusionCloud private cloud DR requires interworking of three layers (external_api, external_om, and storage
replication networks) of the production DC and disaster recovery DC.
external_api
OpenStacks of the product DC and disaster recovery DC share keystone authorization. Therefore, OpenStacks of the
disaster recovery DC must access the external_api network where the keystone of the product DC is located through the
Layer 3 network. ManageOne SCs of the product DC and disaster recovery DC use the external_api network address for
DR replication. Therefore, Layer 3 interworking is required for this network address. BCManager eReplication is deployed
in the disaster recovery DC, and needs to use the external_api network to access resources of the production DC through
interworking of three layers.
external_om
Keystone_databases of the product DC and disaster recovery DC use the external_om network address for DR replication.
Therefore, interworking of three layers is required for this network address.
Storage replication network

If storage copies of the product DC and disaster recovery DC are located on the IP network, interworking of three layers is
required between two DCs.
DR Network Plane Planning
VLAN IP address
Network plane Notes
planning planning
Manage SC DR network plane Four SC of production center
ment and DR center needs
plane two replication IPs
respectively.
Keystone DR network Four Keystone_database of

production center and
plane DR center needs two
replication IPs
respectively.
BCManager Network One

Plane
IO plane Storage DR network 2 x Storage Each storage needs

two replication IP
plane quantity addresses.
DR Network Bandwidth Planning
High Level
Subtype Notes
Design
DR link type FC/iSCSI Fibre Channel links are recommended for synchronous
replication, and iSCSI links are recommended for
asynchronous remote replication.
DR network FC networking: when it is set as synchronous replication (the

100 mbps
overhead of private information needs to be considered), the
bandwidth network bandwidth must be no less than the peak write
bandwidth/(1-11%)/ 70%.
IP networking: when it is set as asynchronous remote
replication (the overhead of private information can be
ignored), the network bandwidth must be no less than the
peak write bandwidth/ 70%.
DR network 50 ms
latency
DR network 0.05%
packet loss rate
DR network jitter 0%
DR Storage System Design
Production Storage Storage Pool
Production Storage Disk Domain Disk Type
Pool Properties
5600 V3 Diskgroup-001, xx disks StoragePool_001 Capacity: XXX GB CD_56V3_DR
RAID: RAID5-9
StoragePool_002 Capacity: XXX GB CD_56V3_DR
RAID: RAID5-9
5800 V3 Diskgroup-002, xx disks StoragePool_001 Capacity: XXX GB CD_58V3_DR
Raid: Raid5-9
Storage Pool
DR Storage Disk Domain DR StoragePool Disk Type
Properties
5600 V3 Diskgroup-001, xx disks StoragePool_001 Capacity: XXX GB SZ_56V3_DR
RAID: RAID5-9
StoragePool_002 Capacity: XXX GB SZ_56V3_DR
RAID: RAID5-9
5800 V3 Diskgroup-002, xx disks StoragePool_001 Capacity: XXX GB SZ_58V3_DR
RAID: RAID5-9
Production Storage DR Storage
Production StoragePool Available zones DR storage StoragePool Available zones

storage
5600 V3 Pool_001 az1.dc1 5600 V3 Pool_001 az1.dc1
Pool_002 az2.dc1 Pool_002 az2.dc1

Management Domain Design
Keystone DR Design
Cloud DC A Cloud DC B
4-node OpenStack Controller 4-node OpenStack Controller

Local HA Local HA
Keystone database replication
Nova/CinderDB KeyStone (primary) KeyStone (secondary) Nova/Cinder DB
Management
Management network
Management network
network
In the DR scenario, the keystone database must be backed up to the DR site, whereas the system
databases (such as cinder and nova) do not need to be backed up to the DR site. Therefore,
databases are split. Considering deployment of the local HA and remote DR, four physical nodes are
needed for every node, among which two nodes are used to deploy the keystone database, and the
other two nodes are used to deploy other system databases.
Active/standby redundancy of the keystone database is implemented through database replication
between the production DC and the disaster recovery DC.
The external_om network plane must be used for replication of the keystone database.
The OpenStack management planes of the production DC and the disaster recovery DC share the
keystone of the production DC.
Management Domain Design SC DR
Design
Cloud DC A Cloud DC B
ManageOne SC (primary) ManageOne SC (secondary)

Local HA SC database replication Local HA
Management
Management network Management
network network
The SC adopts the VM deployment mode, and the local HA is used.

The ManageOne SC uses the external_api network address for DR replication.
Contents
O&M Solution Design
Case Study
Architecture of the Cloud Backup
Solution
Architecture characteristics
Backup service
management platform
Management
Open architecture
ManageOne SC flow
Data flow Built based on the OpenStack open
architecture
Data protection service
Backup storage on the extensible source
platform
DJ-DPS end
Simple and self-service
Self-backup/self-recovery service of
DR management
Cloud platform tenants
software
FusionSphere OpenStack Backup service application and visual
BCManager eBackup
operation management
Reliable and efficient
Supporting deduplication and compression,
saving the storage space by 20%

Supporting permanent incremental backup
and recovery
Automatic switchover of backup nodes
upon failures
Elastic scaling
Elastic scaling of the capacity, reducing
Production storage Backup storage investment risks at the early stage

OceanStor V3/FusionStorage OceanStor V3 Linear scaling of backup performance,
(NAS)/OceanStor 9000 supporting a maximum of 64 backup nodes
Introduction to eBackup
Load sharing Performance linear
scaling
Adopt the cluster structure to automatically Expand the backup performance in a linear
balance the load of backup tasks and evenly manner by adding backup nodes.
allocate backup tasks to another server.
Backup
bandwidth
Cluster-based
scheduling of tasks
1600 MB
Task
distribution
1200 MB
800 MB
Node Node Node Node

A B C D
400 MB
Number of
backup nodes
Backup nodes
0 1 2 3 4
Introduction to eBackup
Disk data IMG1_BitMapFile IMG1_ADDR1 IMG1_ADDR2 IMG1_ADDR3 IMG1_ADDR4
IMG1_ADDR1
A B C D A B C D
IMG1_ADDR2
IMG1_ADDR3
IMG1_ADDR4
Disk data IMG2_BitMapFile IMG2_ADDR1
IMG2_ADDR1
E B C D IMG1_ADDR2
E
IMG1_ADDR3
IMG1_ADDR4
Disk data IMG3_BitMapFile IMG3_ADDR2 IMG3_ADDR4
IMG2_ADDR1
E F C G F G
IMG3_ADDR2
IMG1_ADDR3
IMG3_ADDR4
General Design of Cloud Backup
DC
egress Egress
router
Cloud network
domain GSLB
Core switch SLB SLB
Access switch
OpenStack eBackup V
Computing
DJ-DPS VM VM VM
3*controller M
V
VM VM
M
VM domain
Backup execution node
Cloud platform Storage
management data
access
domain
Backup Product
Storage zone zone
domain
DJ-DPS design Backup execution design Backup storage system design Network design
DJ-DPS Design of the Backup
Management Domain
DJ Network Plane VLAN Planning Interface Rate

Management network plane 413 GE
Tenant network plane 401 GE
Internal network plane 417 GE
Backup Storage Design
The backup storage can be the V3 NAS or OceanStor 9000.
Disk domain
Number of
Disk Domain Disk Type Storage Pool Raid Level
Disks
Domain_01 SAS 48 Pool01 RAID5
File system
Shared File System Capacity Notes

FileSystem01 65 GB eBackup system database
backup
FileSystem02 2 TB Production data backup
eBackup Networking Design
Network Plane Notes
Backup management plane It is a communication plane for eBackup backup software to provide services. The maintenance terminal or
third-party system uses the IP address of this plane to log in to the eBackup backup management system to
configure and manage backup and recovery services.
Production management A communication plane between backup servers and backup proxies or backup servers and management
plane planes at the production end:
When the production storage type is FusionStorage: The plane is used to transmit management messages between
eBackup (including the backup server and backup proxies) and a VRM or between eBackup and FSM.
When the production type is SAN: The plane is used to transmit management messages between backup server and a
VRM or between backup server and a CNA.
Storage plane A communication plane between backup servers/proxies and the production/backup storage plane:
When the production storage type is FusionStorage: The plane is used to transmit backup messages between backup
proxies and FusionStorage or backup proxies between backup storage.
When the production type is SAN: The plane is used to transmit backup messages between backup proxies and backup
storage.
Internal communication It is a communication plane between backup servers and backup proxies.
plane
Network Plane VLAN Planning Interface Rate
Backup/Production management plane 413 GE
Internal communication plane 401 GE
Storage plane 418 10GE or 16 Gbit/s Fibre Channel
eBackup Networking Design
HA design of the eBackup management server
Server Design Description HA Design Description Notes
Server 2
Network adapter Six NICs for each server Every two NICs are bonded.
Power Dual power supplies
Disk Disk group RAID1
Host eBackup backup server
eBackup backup
management plane eBackup internal eBackup production eBackup production
storage plane port
port communication plane management plane
port network port
Access switch Access switch
Aggregation switches Aggregation switches

or core switches or core switches
External share storage
Contents
O&M Solution Design
Case Study
ManageOne Management System
Organization/Service Organization/Service Administrator
Service management
VDC 1 VDC 2 SLA management
vAPP VPC vAPP VPC Metering management
Self operation and Self operation and Resource pool management
Service catalog Service catalog
maintenance maintenance
Unified service lifecycle management
Unified cloud and non-cloud

Unified management of management
heterogeneous resource pools Physical Virtual
resource resource pool
Server, storage, and network

Unified multi-DC management
DC 1 DC 2 DC n
ManageOne System Architecture
ManageOne
SC OC
Service catalog Topo management Alarm management

Tenant management
management
Process approval Performance analysis

Service metering
& capacity prediction Report management
Service gateway Management plug-ins
IaaS resource pool Big data BC&DR

Computing Storage Network Big data BC&DR
service service service service service
Cloud OS FusionInsight DJ
Logical Deployment of ManageOne
End user Operation O&M
personnel personnel
Public network/enterprise O&M network

office network
1. ManageOne supports unified service
provisioning and monitoring of multiple DCs.
2. ManageOne is deployed in headquarters
Headquarters DC
DC, and is connected to the OpenStack and

3rd ITIL eSight of each branch DC through the
private line or VPN.
SC 3. When the SC provisions services, the user
OC
specifies the provisioned resource pool (that
OpenStack is, OpenStack).
4. The OC can grant only the local DC
management authority for the DC
FC vCenter eSight/3rd ITOM
administrator by means of authority and
Managed objects (computing, storage, and network) domain based management.
VPN or
VPN or private private line
line
OpenStack Branch DC OpenStack Branch DC

Managed objects (computing, storage, and
Managed objects (computing, storage, and
network) network)
Cloud Computing O&M Tool
FusionCare Execute immediately.
FusionCare has two main Specify a time. View fault details.

functions: health check and
Add a node.
information collection. By week View node details.
Modify a node. By month Export reports.

The health check process is shown
For the first
in the figure on the right. The login, change Select check Handle unqualified
Delete a node. Sending
process includes the following the password. items of each by email
items by referring
to recommended
node.
steps: selecting a node to be actions.
checked, selecting check items,

Log in to Select a Perform the Handle the
performing the health check, and FusionCare. Select a node.
check item. health check. check results.
handling the check results.
Add a node.
The information collection process

is shown in the figure on the right. Modify a node. Download logs.
The process includes the following For the first Select collection Select
collection
steps: selecting a node to be login, change Delete a node. items of each
node. time.
Delete collection
tasks.
the password.
checked, selecting collection
items, performing information
Log in to Select a Handle the
collection, and handling the FusionCare. Select a node.
collection item. Collect logs. collection results.
collection results.
Contents
O&M Solution Design
Case Study
Desktop Cloud Deployed in Shenzhen
Stock Exchange
Challenges
Data security is the top priority of a securities company. Currently, the PC office
system has the greatest security risks.
Data is scattered and difficult to search and manage. It is slow to deploy services,
and system maintenance is complicated.
A lot of resources are wasted. Multiple hosts are configured for multiple networks.
The overall resource utilization rate is less than 25%. The annual IT investment
exceeds 1,000,000 RMB.
Solution
Cloud DC + thin client, E2E security desktop cloud solution
Unified scheduling of resources and adoption of the clone technology to improve the
resource utilization rate
Automatic O&M and unified software release and update
Customer Benefits
"Since the Huawei desktop cloud system is
deployed, information security of Shenzhen Data is placed in the DC in a unified manner and is not accessible by local
operators. Behavior can be audited, operations can be properly tracked, and the
S t o c k E xc h a n g e h a s b e e n g r e a t l y number of information security events decreases by more than 60%.
enhanced. In addition, centralized O&M of Desktop VM release is implemented at several minutes, and the O&M efficiency
the desktop cloud saves a lot of costs for increases 10 times.
us, which is unimaginable in the PC era." IT resources are elastically allocated. Resource requirements for scenarios (such as
outsourced development) are flexibly handled. The resource utilization rate
by Shenzhen Stock Exchange increases by 20% to 30%.
Huawei Virtualization Solution Helps
the State Grid
Challenges
There are more than 62,000 devices, and the hardware resource utilization rate is
only more than 20%.
The equipment room expansion costs and power consumption continuously
increase, and the TCO is constantly high.
The intensive IT infrastructure service must be built to facilitate investment
decision making of the group and realize fast and unified deployment of services.
Solution
Virtualization of computing, storage, and network; vertical optimization of E2E
software and hardware performance; winning the first place in the comprehensive
test (function, performance, pressure, and service stability)
Compatible with the mainstream X86 servers, storage, and network devices, and
operating systems
Improving service reliability by means of resource virtualization and elastic
scheduling; building DR centers in three places
"In the two rounds of tests conducted by State Grid

that last nearly half a year, the Huawei cloud platform Customer Benefits
has won the first place in the comprehensive test by
The virtualization technology improves the server utilization rate by more than
passing all function tests and achieving excellent 30%, improves the resource reuse capabilities, and reduces the hardware
results in half of the performance and application test procurement in the future by 30% to 40%.
items. Huawei was one of the two vendors that finally The annual power consumption expenditure is cut by more than RMB 27,000,000,
won the bid of State Grid, and has obtained the and the O&M efficiency increases10 times.
bigger share of the bid." The cloud hardware resource pool is established, the new service provisioning
efficiency increases by 50%, and the non-planned service interruption time
by China Electric Power Research Institute decreases by 60%.
Converged Resource Pool Deployed by
Inner Mongolia Mobile Management Service
Network
management
supporting VDC supporting VDC supporting VDC
Customer requirements
Internet surfing
Operation and
SMS gateway
platform
analysis
Build a unified resource pool for the company to

log
provide resources for three supporting domains,

Service Service Service
system system system including the service support, management
supporting, and network management.
All resources are distributed through the cloud
management platform, and resource distribution
is completed using the resource application and
approval process.
Software, hardware, and heterogeneous devices
Big data resource
ManageOne unified management in the resource pool are managed in a unified
pool
manner.
Physical Virtual resource pool

Big data resource pool
analysis Five vSphere FusionSphere
system Physical
vSphere resource resource pool Solution
resource pool
resource pool (new) (new) Implement resource pooling and unified
pools (old) management to share resources and improve the
resource utilization rate.
The new resource pool is used to manage the
existing heterogeneous resource pool and the new
physical and virtual resource pool.
VDCs are divided based on departments so that
multiple departments can share the unified resource
Physical Server Network Backup Security Storage
pool.
resource
Summary
Desk cloud application scenarios and requirement analysis
DC virtualization scenarios and requirement analysis
Can write technical proposals based on the project requirement

analysis.
Can write each chapter of the technical proposal independently.
Quiz
1. What is the composition of a cloud computing project?
2. Which aspects should be considered in the cloud computing project

design?
Quiz
1. True or False
1. A cloud computing project generally consists of the following processes:

requirement analysis, capacity planning, solution design, installation and
deployment, and configuration and commissioning. ( )
2. Multiple-Choice Question
1. The cloud computing project design consists of: ( )
A. Network design
B. Security design
C. Disaster recovery design
D. O&M design
Thank You
www.huawei.com

HC13081 - 01 Cloud Computing Planning Design

Uploaded by

Copyright:

Available Formats

HC13081 - 01 Cloud Computing Planning Design

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HC13081 - 01 Cloud Computing Planning Design

Uploaded by

Copyright:

Available Formats

Cloud Computing

Copyright 2017 Huawei Technologies Co., Ltd. All rights reserved.

Be familiar with the data center (DC) virtualization scenarios and

Analyze and write technical proposals according to project

Requirement Analysis Methods

General Solution Design

Management System Design

Storage System Design

Network Solution Design

Disaster Recovery (DR) Solution Design

Backup Solution Design

O&M Solution Design

3. Compatibility and scalability 6. Standard project implementation

Requirement Analysis Methods

General Solution Design

Management System Design

Storage System Design

Network Solution Design

Disaster Recovery (DR) Solution Design

Backup Solution Design

O&M Solution Design

Requirement Analysis Methods

General Solution Design

Management System Design

Storage System Design

Network Solution Design

Disaster Recovery (DR) Solution Design

Backup Solution Design

O&M Solution Design

Bare metal service storage VPC Backup service

Virtual cluster Virtual cluster Virtual cluster

FusionSphere virtualization solution Standard

DR system Backup system

Performance insight system Health check and log collection

Mgn_VM02 User_VM02 User_VM12

Controller Controller Controller

Infrastructure (server, storage, network, storage, security, and load FusionCare

Requirement Analysis Methods

General Solution Design

Management System Design

Storage System Design

Network Solution Design

Disaster Recovery (DR) Solution Design

Backup Solution Design

O&M Solution Design

2 database 2 VRM 2 rabbitMQ

2 database 2 VRM 2 rabbitMQ 4 Extend Controller

auth auth auth

image image image

controller controller controller

sys-server sys-server sys-server

measure measure measure

mongodb mongodb mongodb

swift swift swift

zookeeper zookeeper zookeeper

compute compute compute

sys-client sys-client sys-client

blockstorage-driver blockstorage-driver blockstorage-driver

Controller 01 Controller 02 Controller 03

256GE IB 256GE IB 2*56GE IB