Fully Automated Load Balancing, Compute resource of the cluster load gets automatically balanced. Hosts will be grouped with respect to the location of the datacenter and affinity rule will enabled onthe VM’s to run in the neededdatacenters.

1. Challenge 1
Virtual Design Master
Mohamed Ibrahim

2. Virtual Design Master 2
1. Table of Contents
1. Table of Contents ..............................................................................................................................................2
2.0. Conceptual Model ...........................................................................................................................................3
1.1 Solution Requirements................................................................................................................................3
1.2 Design Constraints......................................................................................................................................4
1.3 Design Assumptions....................................................................................................................................4
1.4 Risks.............................................................................................................................................................5
3.0. Logical Design.................................................................................................................................................6
1) Highly Available Virtualization Infrastructure..........................................................................................6
2) Dynamic Resource Control of Compute Resources.....................................................................................7
3) Separation of Clusters..................................................................................................................................8
4) Virtual Infrastructure Management. ..........................................................................................................9
5) Hypervisor Host Sizing ................................................................................................................................9
1) Storage Array..............................................................................................................................................10
2) Synchronous SAN Replication ...................................................................................................................11
3) Asynchronous SAN Replication .................................................................................................................12
1) LAN Switching............................................................................................................................................14
2) Virtual Switches .........................................................................................................................................15
3) Trunking and VLANs.................................................................................................................................16
1) Virtual Machine Size..................................................................................................................................16
2) vCPU and vRAM Management..................................................................................................................17
1) VM Image backup.......................................................................................................................................18
4.0 Document Control.........................................................................................................................................20
2.

3. Virtual Design Master 3
2.0Conceptual Model
The following sections describe the Customer Requirements, Design Constraints and Design Assumptions. The tables below
map these into the infrastructure design qualities.
1.1 Solution Requirements
Table 1: Solution Requirements
Reference Customer Requirement Design Quality
ALL.R1
Two
primary
datacenter
that
will
function
as
active/active
which
will
host
all
the
servers
in
highly
available
fashion.
Availability
ALL.R2
There
will
be
another
datacenter,
which
will
function
as
passive
and
contains
entire
copy
of
active
datacenters.
Availability
ALL.R3
Solution
should
leverage
FCoE
to
have
converged
infrastructure
for
both
Network
and
SAN
traffic
Manageability
ALL.R4
Maximum
RTO
of
business
continuity
should
be
20
mins.
BC/DR
ALL.R5
Application
to
support
99.99%
availability
with
both
local
and
global
resiliency
Availability
ALL.R6
This
design
must
use
40
GB
to
leverage
the
FCoE
infrastructure
Manageability
ALL.R7
Solution
should
support
to
functioning
of
active/active
datacenter
Availability
ALL.R8
Solution
should
support
application
deployment
Manageability
ALL.R9
Virtualization
HA
will
be
used
for
all
management
components
Manageability
ALL.R12
Virtualization
Metro
Storage
Cluster
will
be
used
between
the
active/active
datacenters
Availability
ALL.R13
Storage
shall
provide
snapshot
features
to
provide
the
snapshot
feature
to
make
the
read
only
copy
of
the
backup
images
in
the
passive
site
Recoverability
ALL.R14
Multi
Cluster
Infrastructure
should
use
to
isolate
the
workloads
Manageablity

4. Virtual Design Master 4
1.2 Design Constraints
Table 2: Design Constraints
Reference Design Constraints Design Quality
ALL.C1
Legacy
hardware
available
for
building
the
infrastructure
in
Mars
datacenters
Manageability
ALL.C2
Availability
of
cloud
Administrator
with
specialized
skills
is
very
limited
and
willingness
to
work
in
other
planet
is
challengeable.
Manageability
ALL.C3
The
adequate
link
should
be
present
between
all
the
three
datacenters.
Availability
ALL.C4
Power,
cooling
and
space
should
be
as
minimum
as
possible
Manageability
ALL.C5
Synchronous
Replication
solutions
should
able
to
provide
RTT
of
10ms
to
build
the
active/active
datacenter.
Availability
1.3 Design Assumptions
Table 3: Design Assumptions
Reference Design Assumptions Design Quality
ALL.A1
Centralized
Infrastructure
Management
and
orchestration
will
be
used
Manageability
ALL.A2
Dark
Fiber
will
be
used
between
the
Datacenter
1
to
Datacenter
2.
Performance
ALL.A3
Daily
Backup
of
VMs
will
be
scheduled
Recoverability
ALL.A4
Virtual
Machine
resource
managements
methods
will
be
adopted
Performance
ALL.A5
Network
Bandwidth
between
the
sites
are
well
adequate
to
handle
the
network
traffic
and
storage
traffic
Availability

5. Virtual Design Master 5
1.4 Risks
ID Severity Description
Risk1
Major
Most
of
the
skilled
resource
have
been
specifically
targeted
by
Zombies.
So
there
is
an
inherent
shortage
of
large
number
of
staffs
with
subject
matter
expertise.
Risk2
Critical
Since
the
legacy
infrastructure
design
is
used,
it
will
increase
the
complexity
of
the
solution.
Risk3
Major
People
and
Process
Transformation
is
key.
This
should
be
considered
as
a
separate
architecture
project
itself.
But
this
has
to
wait,
as
TOGAF
is
still
framing
up
the
people
and
process
transformation
rules
for
the
new
Zombie
era.

6. Virtual Design Master 6
3. Logical Design
This chapter focus on the key design quality attributes to satisfy the conceptual design model covered earlier in the document.
While Manageability and Availability have been listed under the design qualities, these attributes are inherent to the proposed
Infrastructure design.
Hypervizor
Host
01 Hypervizor
Host
02 Hypervizor
Host
03 Hypervizor
Host
04
Hypervisor
Cluster
Storage
DC1 Storage
DC2
Stretched
Storage
Volume
Synchronous
ReplicationVolume Volume
MARS
Datacenter
1 MARS
Datacenter
2Hypervisor
Cluster
Hypervizor
Host
01 Hypervizor
Host
02 Hypervizor
Host
n
Hypervisor
Cluster
Storage
DC1
Volume
MARS
Datacenter
3
(Passive
Datacenter)
Asynchronous
Replication
Active/Active
Datacenter
3.1- Compute
1) Highly Available Virtualization Infrastructure
Design Ref. ALL.R9
Design Choice High Availability of compute resources

7. Virtual Design Master 7
Design Quality Manageability, Availability
Requirements
Reference
ALL.R9
Requirements
Conflict
None
Justification Automatic restart of VM’s in the event of host failure. HA will
able to recover the VM’s event of 1 datacenter failure. 50% of
host capacity is reserved for the failure.
Impact Cost
Risks None
2) Dynamic Resource Control of Compute Resources.
Design Ref. ALL.R12
Design Option 1) Automated Load Balancing
2) Partial Load Balancing
3) Manual Load Balancing
Design Choice Automated Load Balancing
Design Quality Performance, Manageability
Requirements
Reference
ALL.R12
Requirements
Conflict
None
Justification Fully Automated Load Balancing, Compute resource of the
cluster load gets automatically balanced. Hosts will be grouped
with respect to the location of the datacenter and affinity rule
will enabled on the VM’s to run in the needed datacenters.
Impact None
Risks None

8. Virtual Design Master 8
3) Separation of Clusters.
Hypervizor
Host
01
Hypervizor
Host
02
Hypervizor
Host
03
Hypervizor
Host
04
Storage
A Storage
B
Stretched
Storage
LUN
Volume Volume
MARS
Datacenter
1 MARS
Datacenter
2Stretched
Management
Cluster
Hypervizor
Host
01
Hypervizor
Host
02
Hypervizor
Host
03
Hypervizor
Host
04Stretched
Cluster
Storage
A Storage
B
Stretched
Storage
LUN
Volume Volume
Production
Stretched
Cluster
Design Ref. ALL.R9
Design Option Number of Pooled, Compute, Storage and Network Resources
Design Choice 1) Management Cluster
2) Production Cluster
Design Quality Availability, Performance
Requirements
Reference
ALL.R9
Requirements
Conflict
None
Justification All the infrastructure management servers will be running in
the management clusters and critical legacy applications will be
hosted in the production metro cluster, which will be hosted
across active active datacenter.
Impact
Separate clusters. Many hosts. Increased cost. Increased
administrative overhead
Risks Increased complexity and cost.

9. Virtual Design Master 9
4) Virtual Infrastructure Management.
Design Ref. ALL.R9
Design Choice Centralized virtualized infrastructure management with
orchestration
Design Quality Manageability
Requirements
Reference
ALL.R14
Requirements
Conflict
None
Justification Reduce the number of integration point with infrastructure
component.
Impact
Simplifies the use and configuration
Risks None
5) Hypervisor Host Sizing
Design Ref. ALL.C4
Design Option Host Sizing “scal-up” or “scale-out”
Design Choice 1) Scale Up
Design Quality Availability, Performance
Requirements
Reference
ALL.C4
Requirements
Conflict
None

10. Virtual Design Master 10
Justification Since space is a concern in thte datacenter we will go with
scale up plan. All hosts will be of identical model and
purchased with maximum CPU (latest generation,
maximum cores per socket). This will create a
standardized environment where host hardware can be
reused anywhere within the Corporation. Scale Up is the
preferred host configuration for Data Center VMs that
have heavy RAM workloads.
Impact
Predictable performance and increased cost
Risks None
3.2- Storage
1) Storage Array
Design Ref. ALL.R13
Design Choice 1) Storage Array with FCoE front end ports
Design Quality Availability, Performance and management
Requirements
Reference
ALL.R13
Requirements
Conflict
None
Justification Storage Array with FcoE ports will leverage SAN traffic
into the network traffic. Array should support inline de-
dup and flash caching tier to provide the performance to
the I/O workload.
Impact
Performance and increased cost

11. Virtual Design Master 11
Risks None
2) Synchronous SAN Replication
Hypervizor
Host
01 Hypervizor
Host
02 Hypervizor
Host
03 Hypervizor
Host
04
MARS
Datacenter
1 MARS
Datacenter
2
Stretched
Cluster
Distributed
File
System
Storage
Cluster-­‐A
Physical
Storage-­‐A
Storage
Cluster-­‐B
Physical
Storage-­‐B
SAN
Replication
LUN
(Read-­‐Write)
LUN
(Read-­‐Write)
Design Ref. ALL.R13
Design Choice Storage replication device that support synchronous replication
and asynchronous replication.
Design Quality Availability, Performance and management
Requirements
Reference
ALL.R13
Requirements
Conflict
None

12. Virtual Design Master 12
Justification SAN Synchronous replication supports both R1 and R2
LUN should be R/W on both the datacenter that leads to
build Metro Storage Cluster. Synchronous replication will
be used between the datacenter 1 to datacenter 2. Dark
fiber will be used between the datacenter 1 to Datacenter
2.
Impact
Performance and increased cost
Risks None
3) Asynchronous SAN Replication
Hypervizor
Host
01 Hypervizor
Host
n Hypervizor
Host
01 Hypervizor
Host
n
MARS
Datacenter
1 MARS
Datacenter
3
Storage
Cluster-­‐A
Physical
Storage-­‐A
Storage
Cluster-­‐B
Physical
Storage-­‐B
Asynchronous
Replication
LUN
(Read-­‐Write)
LUN
(Read-­‐Only)

13. Virtual Design Master 13
Design Ref. ALL.R13
Design Option 1) Hypervisor based replication
2) SAN Based Replication
Design Choice SAN based replication will be used to have the passive copy of the
data to maintain the good RPO requirement.
Design Quality Availability, management
Requirements
Reference
ALL.R13
Requirements
Conflict
None
Justification SAN Replication will be used to have the R/O copy of the
R2 Device. In the event of failure in the active datacenter,
passive copy will brought up with in the defined SLA.
Asynchronous replication will be used between the
datacenter 1 to datacenter 3.
Impact
Performance and increased cost
Risks None

14. Virtual Design Master 14
3.3- Network
Hypervizor
Chassis
FcoE
Sw1
FcoE
SW2
10Gb
Capable
Switch
1 10Gb
Capable
Switch
2
40Gb
Capable
Switch
1 40Gb
Capable
Switch
2
1) LAN Switching
Design Ref. ALL.R6
Design Choice Physical LAN Switches and Host NIC will be redundant and 10 Gb
capable.
Design Quality Availability, management
Requirements
Reference
ALL.R6
Requirements
Conflict
None
Justification Redundant high speed LAN will be used for VM traffic
and Hypervisor service

15. Virtual Design Master 15
Impact
High Speed, Highly Reliable, Reduced cabling and Redundant
uplinks
Risks None
2) Virtual Switches
Hypervizor
Host
Uplink
1
Uplink
2
Uplink
3
Uplink
4
Centralized
Virtual
Switch
1 Centralized
Virtual
Switch
2
Host
Mgmt
VM
Migration
VM
Traffic
Backup
Traffic
Design Ref. ALL.R6
Design Choice Virtual Switches will be used to isolate the functional traffic groups
with each 2 uplinks attached to it..
Design Quality Availability, management
Requirements
Reference
ALL.R6
Requirements
Conflict
None
Justification One Virtual Switches will be used for vmotion and host
management and another Virtual Switches will be used
for VM traffic and backup network.

16. Virtual Design Master 16
Impact
CNA Required, Reduced chance of congestion and Redundant
uplinks
Risks None
3) Trunking and VLANs
Design Ref. ALL.R6
Design Choice Trunking and VLANs
Design Quality Availability, manageability
Requirements
Reference
ALL.R6
Requirements
Conflict
None
Justification Trunking VLAN’s will be used to isolate the functional
sub traffic types on the virtual switches.
Impact
Scalable Network and reduced network cabling
Risks None
3.4- Virtual Machine
1) Virtual Machine Size
Design Ref. ALL.A4
Design Choice Standard Virtual Machine sizes will be deployed
Design Quality Manageability
Requirements
Reference
ALL.A4

17. Virtual Design Master 17
Requirements
Conflict
None
Justification Protects the Non-NUMA Application from NUMA limits
being exceeded and experiencing performance
degradation. After Deployment VM Performance can be
easily analyzed and sized.
Impact
Scalability and Reduced Administration overhead.
Risks None
2) vCPU and vRAM Management
Design Ref. ALL.A4
Design Choice Virtual Machine CPU and RAM Management mechanism will be
used.
Design Quality Manageability
Requirements
Reference
ALL.A4
Requirements
Conflict
None
Justification Must able to provide Tier-1 VM’s priority access to
resources.
Impact
Improved Performance Prioritization of resources by policy.
Risks None

18. Virtual Design Master 18
3.5- Backup and Recovery
1) VM Image backup
Design Ref. ALL.A3
1) VM Image based backup
2) Agent based backup
Design Choice VM Image backup
Design Quality Recoverability
Requirements
Reference
ALL.A3
Requirements
Conflict
None
Justification Daily Consistent Copy of All VM’s at Datacenter 1 and
Datacenter 3.
Impact
Protection against corruption, crashes and deletion of VM.
Risks None

19. Virtual Design Master 19
3.6- Business Continuity/Disaster Recovery (BC/DR)
Hypervizor
Host
01 Hypervizor
Host
02 Hypervizor
Host
03 Hypervizor
Host
04
Hypervisor
Cluster
Storage
DC1 Storage
DC2
Stretched
Storage
Volume
Synchronous
ReplicationVolume Volume
MARS
Datacenter
1 MARS
Datacenter
2Hypervisor
Cluster
Hypervizor
Host
01 Hypervizor
Host
02 Hypervizor
Host
n
Hypervisor
Cluster
Storage
DC1
Volume
MARS
Datacenter
3
(Passive
Datacenter)
Asynchronous
Replication
Active/Active
Datacenter
Design Ref. ALL.A4
Design Choice Datacentre 1 and Datacentre 2 will work as Active/Active
datacenter and Datacenter 3 will be configured as Passive
datacenter
Design Quality Manageability
Requirements
Reference
ALL.A4

20. Virtual Design Master 20
Requirements
Conflict
None
Justification Datacenter 1 & 2 will work as active/active datacenter,
which will be logically one datacenter to avoid the planned
downtime and business continuity. Datacenter 3 will be
passive datacenter, which will ready to switch over all the
service in the event of disaster.
Impact
Improved Performance Prioritization of resources by policy.
Risks None
4. Document Control
Version Reason for Change Author Reviewer Date
1.0 Initial Version Mohamed Ibrahim 06-July 2015