planning and designing server virtualisation
TRANSCRIPT
ICANWK525B Configure an Enterprise Virtual Computing Environment
ICTSUS5187A Implement Server Virtualisation for a Sustainable ICT system
Planning and Designing Server Virtualisation
An organisation has existing servers Issues: lots of power, low utilisation of
resources, cost of maintenance They are considering the feasibility of
virtualising these servers They have asked you to assess this
feasibility You need to write them a feasibility report
Planning to Virtualise Servers
Virtual servers need the same disk space as real ones
Virtual servers need roughly the same RAM (memory) as real ones
Virtual servers need the same bandwidth as real ones
Virtual servers, on average, only utilise about 15% to 20% of their CPU capacity
Things to Remember
You will either use the existing hardware, buy completely new hardware, or have a mix of old and new hardware
Given the 15-20% rule, plan for about 1/5 the number of equivalent CPUs◦ Less physical servers
However, same RAM requirements, so more RAM per physical server than before
Same disk requirements◦ Either more disks in each server, or a SAN
Same bandwidth, less servers => more or faster NICs per server. Even more so, if using SANs
Plan for Replacement Hardware
Ensure you can migrate VMs between physical servers, to allow you to have physical server downtime
Ensure flexibility to map drive space to all the VMs across all the physical servers
Ensure some spare capacity in case one physical server dies
Implication: some spare server CPU/RAM capacity
Implication: a SAN for the physical drives◦ Implication: more bandwidth and physical
infrastructure for the SAN
Plan for Redundancy & Flexibility
Baseline: measurement of the current system
I would measure over a week: Peak/average/minimum CPU usage per day Peak/average/minimum RAM usage per day Peak/average/minimum network bandwidth
per day Peak/average/minimum disk bandwidth per
day This will inform you as to the hardware and
networking capabilities of the new system
Take a Baseline
Ten existing servers, each with two quad-core Xeon CPUs, 8GB RAM, 2TB disk space
Connected to a switch with 10Gbps Ethernet ports
Baseline measurements:◦ 15% CPU utilisation, peaks at 100% on some servers for
a few minutes◦ Average RAM use: 6GB per server, peaks at 8GB but not
at same time across all servers◦ Network usage: two servers average 6 Gbps, the rest
average 1 Gbps, with occasional peaks of 5Gbps◦ Disk usage: at present, 14TB of the 20TB used◦ Disk bandwidth: average 3Gbps total, peaks to 5 Gbps
Example Scenario
Scale CPUs to 1/5: have only four quad-core Xeon processors
RAM: still need 10x6GB = 60GB, perhaps a bit more
Network: need at least 20Gbps Disk usage: need at least 14TB Disk bandwidth: need at least 5Gbps
Now, we have to spread this across the new physical servers
We want redundancy and flexibility
Example Scenario
2 or 3 servers: I would go for 3 servers◦ 2 can carry the load while one is down◦ Two quad-core Xeons per server => 6 Xeons not 4
RAM: 24GB per server times 3 = 72GB total
Bandwidth to the users: we need a 10Gbps multiport switch plus 7Gbps per server to the switch => go for 10Gbps NICs per server
Gives 30Gbps overall bandwidth
New Scenario
Storage Area Network (SAN):◦ RAID 6, we need 2 drives spare in an 8-drive array◦ Eight 3TB drives = 24TB, but only six 3TB = 18TB
usable
Bandwidth to the SAN: 2Gbps per server, so go for 10Gbps NICs◦ 3 NICs at 10Gbps is overkill
SAN device needs a 10Gbps NIC, plus a 10Gbps multiport switch to the servers
New Scenario
New Scenario
10 servers, now 3: power savings Redundancy with 3 servers: migrate VMs
if/when one server is down Need SAN device & associated
switch/cables May need new client-side switch Can we re-use some of the existing servers?
◦ Hard to tell without knowing the exact servers
Scenario Results & Questions
Feasibility report: is it feasible to virtualise the servers. Sections of the report:
Overview of the problem Analysis of existing hardware Baseline of existing system Recommendation of new design
◦ Advantages, disadvantages of new design◦ Ability to re-use existing infrastructure
Risk analysis: ◦ What could go wrong with the new design◦ What could go wrong during the transition
Will be a multi-page document, very formal
Feasibility Report
Feasibility given constraints: budget, time to complete, existing space etc.
You will need to give multiple options for new systems, plus the “null” option: do nothing◦ You can still mark one as the preferred option
Compare the advantages & disadvantages of each: cost, performance, risks.
Gives customer the ability to choose something that meets their organisational requirements and meets their constraints
Feasibility Report
The organisation has chosen one of the solution plans
Now it is design time Choose specific hardware to match the plan Choose specific virtualisation software Ensure existing customer server software
can be virtualised and migrated into VMs on the new system
Designing the Solution
VMware vSphere with ESXi Microsoft Hyper-V running Win Server 2012 Links:
◦ VMware: Comarison Against Hyper-V◦ InfoWorld: vSphere versus Hyper-V◦ Pay per Cloud: vSphere versus Hyper-V
Not just technical comparison, think of training, support, ease of management, ability to work with existing infrastructure◦ i.e the total cost of ownership (TCO)
The Main Server Virtualisation Products
Need to be able to migrate software on existing servers into VMs:◦ Operating system, applications, data stores
Both VMware and Hyper-V come with tools to do this:
VMware vCenter Converter Microsoft System Center Virtual Machine Ma
nager
Migrating Existing Servers to VMs
During your baseline, you should have done an audit of existing server operating systems and the application they are running◦ e.g. Windows Server 2008, Linux, web services
etc. Make sure any solution you choose can
virtualise these systems.◦ Both vSphere and Hyper-V can do Windows and
Linux vSphere has minimum
hardware requirements If chosen, make sure your design meets
these
Software & Hardware Requirements
Physical space, power, cooling If replacing an existing server system, you
will most likely need less space, power, cooling
However, during the transition, you may need more space, power, cooling◦ And the transition may take weeks
For both a new and a replacement build, you must determine the space, power and cooling requirement of the servers
Environmental Considerations
Server room design is out of scope for this unit. Some useful links:
M.Moser: Server Room Design UC Davis: Server Room Best Practices Requirements for the Design of ICT Rooms
Environmental Considerations