This document outlines a design scenario for virtualizing the infrastructure of a digital media company called eRaw-mv. It discusses examining the design process, extracting key design factors like requirements, constraints, assumptions, and risks. It then focuses on three main requirements: virtualizing remaining tier 1 applications, implementing disaster recovery, and modernizing remote/branch offices. For each requirement, it explores conceptual and logical design options to consider at the physical implementation level, highlighting specific areas to address like performance, licensing, and monitoring. The presenters aim to demonstrate how to holistically analyze a design scenario by tying business needs to technical solutions.
4. Before we start
• Get involved!
• If you use Twitter, feel free to tweet about this
session (use hashtags #VSVC4995 &
#vmworld)
• We encourage you to take photos or videos
of today’s session and share them online
• This presentation will be made available
online after the event
5. Agenda
• Review key design concepts
• Examine the design process
• Outline a scenario
• Extract design factors
• Focus on a few design areas
• Discuss the impact of our decisions
6. Review key design concepts
• You must view the design
holistically (“intimately
interconnected and
interdependent”), not as a
collection of parts
• Tying everything together
are the functional
requirements
• Decisions are driven by
functional requirements as
well as assumptions and
constraints
7. Review key design concepts (cont’d)
Constraints and assumptions = non-functional
requirements or qualities
Constraints Assumptions
Restrictions (or limitations) placed
on you which affect the design
Expectations that you can’t
confirm, so you explicitly exclude
them
Examples:
• Must use their existing SAN
• Servers you recommend must
come from vendor X
Examples:
• Sufficient IP addresses are
available
• Windows licenses for vCenter
Update Manager are covered
by company’s agreement
8. Review key design concepts (cont’d)
• A vSphere design can be measured or
evaluated according to five key dimensions:
• Availability
• Manageability
• Performance
• Recoverability
• Security
• Every design decision has an impact and
should be justifiable
9. Examine the design process
Stages come from Zachman EA taxonomy:
- de facto standard for classifying EA artifacts
1. Contextual Scope (Planner/Strategist)
2. Conceptual Requirements (Owner)
3. Logical System model (Designer)
4. Physical Specifications (Builders)
5. Detailed Configuration (Implementer)
6. Functional Operation (User)
10. Examine the design process (cont’d)
• The logical design defines the
attributes required and their
relationships (the how)
• The physical design considers
the constraints and states the
reality of the solution—not
always physical equipment, at
least tangible specifics (the
what)
Logical versus physical is an essential
part of the design process
11. Outline a scenario
• You have just been appointed as the new infrastructure
architect of eRaw-mv, a distributor/reseller of digital
images and videos.
• This is not a greenfield deployment, but you have buy-in
from the CIO to make changes and modernize the
infrastructure (but without much budget).
• The CIO enthusiastically hired an external consultancy
company to virtualize his base servers about 4 years
ago, but nothing has really changed since then – it is
running vSphere 4.0
12. Outline a scenario: company
• eRaw-mv has around 450
employees.
• Headquarters are in San
Francisco with 3 satellite
offices:
o Engineering offices in
Salt Lake City & Denver
o A sales office in LA
• The CIO recently signed a 3
year co-lo agreement with
plans to use this for DR.
13. Outline a scenario: workloads
• eRaw-mv has 120 VMs and 9 non-virtualized servers in the
headquarters (HQ).
• The CIO is keen to fully virtualize all his servers, but there
has been concerns about tier 1 performance.
• Existing non-virtualized tier 1 servers are vCenter,
Exchange, MS SQL cluster (for internal apps), and Oracle
DB on Linux (backend for mission critical customer portal)
• Application and infrastructure performance stats have been
gathered
• Each remote office has 5-10 VMs. They lack
standardization, but are running well and the CIO sees this
as low-priority task.
14. Outline a scenario: hardware
• The eRaw-mv HQ has 1 year-old rack servers with
sufficient capacity for current growth on existing VMs.
• The head office SAN is a FC traditional array with no real
caching or tiering options.
• The CIO doesn’t trust SAN performance for his tier 1
apps. Tier 1 non-virtualized servers each use several
trays of DAS.
• The servers in the remote offices are out of warranty.
They rely on DAS, and partly because of this the ESX
hosts have never been patched.
15. Outline a scenario: next steps?
• This is just a bare bones scenario
• What additional steps might be needed in real
world?
o Collate more information – current state analysis
o Meet with technical teams
o Check colo agreement
o Site visits
o Identify stakeholders
• Audience feedback: what else?
16. Extract design factors: requirements
• Virtualize remaining servers
• DR design and implementation
• Deal with out of warranty servers in remote
offices
• “Modernize” the infrastructure
• Audience feedback: what other requirements?
17. Extract design factors: constraints
• DR site is fixed (facilities, WAN, services)
• DR will be done internally (no DRaaS)
• Traditional FC SAN for existing VMs
• CAPEX budget for 2013/2014 is very limited
• WAN links cannot be upgraded
• Audience feedback: what other constraints?
18. Extract design factors: assumptions
• There are no performance issues with the
existing setup (servers/storage).
• No server/storage capacity exists for new
workloads in HQ
• The neworking hardware will profide sufficient
ports, redundancy and there are no bottlenecks
• Audience feedback: what other assumptions?
19. Extract design factors: risks
• Bandwidth may be insufficient for DR RPOs
• WAN bandwidth during failover event to DR
• Existing FC SAN isn’t good enough for tier 1
apps
• Very old version of vSphere
• No ESXi security patches applied to remote
office servers
• Audience feedback: what other risks?
20. Focus on a few requirements
• Let’s group this by the 3 main requirements:
o Virtualize tier 1 servers
o Disaster recovery (DR)
o Remote office/branch office (RO/BO)
• We’ll follow this general framework:
o Conceptual – what do we want to do
o Logical design options > Physical design
21. Focus on requirements: Tier 1 apps
P2V or rebuild?
Use of VM
reservations/shares
DRS rules
Fault Tolerance vApps
New hosts or re-use
existing?
vSphere licensing
impacts
vFlash to improve
existing SAN?
Re-use existing tier 1
DAS
Second SAN Cluster design Alarms/monitoring
22. Focus on requirements: RO/BO
Standardized VM
builds
Standardized
hardware
Centralized or
distributed workloads
App delivery
methodologies
Feasible to extend
warranty?
HCL
Enhanced vMotion
(better use of DAS)
Engineering/Sales
differences-similarities
23. Focus on requirements: DR
Which VMs
need DR?
RPO/RTO
Rate of change vs.
WAN capacity
Active DR site vs.
passive DR site
Re-use older RO/BO
servers in DR?
DR capacity
requirements
Replication
techniques
Licensing Linked mode?
Base infrastructure
required (tier 0) SSO design impacts
24. Summary
• Must view the design holistically
• The design is driven by both functional &
non-functional requirements (constraints)
• Consider the design principles (AMPRS)
• Follow “Conceptual > Logical > Physical”
process
• Understand design decision impacts and
justify choices