ITT 420

Chapter 3

Servers

Prepared by: Muhammad Azizi

([email protected])
 ITT 420

Chapter 3 Objectives
 Determine strategies for providing
server resources.
 Determine server hardware features:
enterprise and cluster.
 Discuss server hardware specifications
when purchasing an individual server.
 ITT 420

Chapter 3 Outline
 Server Hardware Strategies
 Server Hardware Features
 Server Hardware Specifications
 Server Hardware Strategies

Server Hardware Strategies

 Server machines is used to provide
application services, web services,
databases, batch computation, back-
office processing, and so on.
 A single machine might provide one
service or many.
 There are many strategies for providing
server resources. Most organizations use
a mix of these strategies.
 Server Hardware Strategies

Server Hardware Strategies

 The three most common strategies are:
 All eggs in one basket: One machine used for many
purposes
 Beautiful snowflakes: Many machines, each uniquely
configured
 Buy in bulk, allocate fractions: Large machines
partitioned into many smaller virtual machines using
virtualization or containers
 Servers variations and alternatives:
 Grid computing: Many machines managed one as unit
 Blade servers: A hardware architecture that places many
machines in one chassis
 Cloud-based compute services: Renting use of someone
else’s servers
 Software as a service (SaaS): Web-hosted applications
 Server appliances: Purpose-built devices, each providing
a different service
 Server Hardware Strategies

All Eggs in One Basket

 Purchase a single server and use it for many services.
 For example, a single machine might serve as the
department’s DNS server, DHCP server, email server, and
web server.
 In this setting it becomes critical to ensure data
integrity.
 Therefore RAID should be used so that a single disk can fail and
the system will continue to run.
 Data integrity also requires regular data backups to tape or
another system.
 If a server is expected to last a long time, it will likely
need to be expanded.
 Additional slots for memory, interfaces, and hard drive bays make
it easy to upgrade the system without replacing it.
 Downside of this strategy cause OS upgrades very
difficult and risky.
 The OS cannot be upgraded or patched until all the services are
verified to work on the new version.
 One application may hold back all the others.
 Server Hardware Strategies

Beautiful Snowflakes
A better strategy is to use a separate machine
for each service.
 Each machine is sized for the desired
application: RAM, disk, number and speeds of
NICs, and enough extra capacity, or
expansion slots, for projected growth during
the expected life of the machine.
 The benefit of this strategy is that the
machine is the best possible choice that
meets the requirements.
 The downside is that the result is a fleet of
unique machines causing variation.
 Each new system adds administrative
overhead proportionally
 Server Hardware Strategies

Beautiful Snowflakes
 Asset Tracking
 When managing many unique machines it becomes
increasingly important to maintain an inventory of
machines.
 The inventory should document technical information such
as the operating system and hardware parameters such as
amount of RAM, type of CPU. Also owner, running service
and contact person
 Reducing Variations
 Always be on the lookout for opportunities to reduce the
number of variations in platforms or technologies being
supported.
 Select a default hardware vendor, model, and operating
system. Automated the configuration
 Global Optimization
 While it sounds efficient to customize each machine to the
exact needs of the service it provides, the result tends to be
an unmanageable mess.
 Avoid unnecessary local machine customization
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 The next strategy is to buy computing resources
in bulk and allocate fractions of it as needed.
 One way to do this is through virtualization.
 organization purchases large physical servers
and divides them up for use by customers by
creating individual virtual machines (VMs).
 A virtualization cluster can grow by adding more
physical hardware as more capacity is needed.
 VMs can also be resized. You can add RAM,
vCPUs, and disk space to a VM via an API call
instead of a visit to the datacenter.
 Virtualization improves computing efficiency.
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 virtualizationprovides better isolation than
simple multitasking. The benefits of
isolation include:
 Independence: Each VM can run a different
operating system.
 Resource isolation: The disk and RAM allocated
to a VM are committed to that VM and not
shared.
 Granular security: A person with root access on
one VM does not automatically have privileged
access on another VM.
 Reduced dependency hell: Each machine has
its own operating system and system libraries, so
they can be upgraded independently.
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 VM Management
 keeping a good inventory of VMs is important.
 you need to maintain an inventory of who owns
each VM and its purpose.
 Some clusters are tightly controlled, only
permitting the IT team to create VMs.
 Other clusters are general-purpose compute
farms providing the ability for customers to
request new machines on demand.
 There should be limits in place so that customers
can’t overload the system by creating too many
VMs.
 As in the other strategies, it is important to limit
the amount of variation.
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 Live Migration
 VM can be moved from one physical host to
another while it is running.
 Live migration makes management easier.
 It can be used to rebalance a cluster, moving
VMs off overloaded physical machines to others
that are less loaded.
 If a physical machine is having a hardware
problem, its VMs can be evacuated to another
physical machine.
 The architecture of a typical virtualization cluster
includes many physical machines that share a
SAN for storage of the VM’s disks.
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 Live Migration
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 VM Packing
 VMs cannot span physical machines.
 As a consequence, we often get into situations where
the remaining RAM on a physical machine is not
enough for a new VM.
 The best way to avoid this is to create VMs that are
standard sizes that pack nicely.
 Spare Capacity for Maintenance
 If a physical machine needs to be taken down for
repairs, there has to be a place where the VMs can be
migrated if you are to avoid downtime.
 One strategy is to keep one physical machine entirely
idle so that, when needed, the VMs from the machine
to be repaired can all be migrated to this machine.
 Another strategy is to distribute the spare capacity
around the cluster so that the individual VMs can
share the extra I/O bandwidth.
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 Unified VM/Non-VM Management
 Most sites end up with two entirely different ways to
request, allocate, and track VMs and non-VMs.
 It can be beneficial to have one system that
manages both.
 Containers
 Containers are another virtualization technique.
 provide isolation at the process level instead of the
machine level.
 All of the containers run under the same operating
system, but each container is self-contained as far
as the files it uses.
 Containers are much lighter weight and permit more
services to be packed on fewer machines.
 Docker, Mesos, and Kubernetes are popular
systems for managing large numbers of containers.
 Server Hardware Strategies
Buy in Bulk, Allocate
Fractions
 Containers vs. Virtualization
 Server Hardware Strategies

Grid Computing
 Grid computing takes many similar machines and manages
them as a single unit.
 Each one is configured exactly alike—same hardware and
software.
 To use the grid, a customer specifies how many machines
are needed and which software package to run.
 The grid management system allocates the right number of
machines, installs the software on them, and runs the
software.
 A big part of grid management software is the scheduling
algorithm.
 Server Hardware Strategies

Grid Computing
 Grid computing has other constraints. Often there is
more bandwidth between machines on the same
rack, and less bandwidth between racks.
 Grid computing is more efficient than virtualization
because it eliminates the virtualization overhead,
which is typically a 5 to 10 percent reduction in
performance.
 Grids are easier to manage because what is done for
one machine is done for all machines.
 Server Hardware Strategies

Blade Servers
 Each machine needs to be racked, connected to
power, networked, and so on.
 Blade servers reduce this overhead by providing
many servers in one chassis.
 A blade server has many individual slots that take
motherboards, called blades, that contain either a
computer or storage.
 Each blade can be installed quickly and easily
because you simply slide a card into a slot.
 Another benefit of blade systems is that they are
software configurable.
 Blade systems are most cost-effective when the
chassis lasts many years, enabling you to upgrade
the blades for the duration of its lifetime.
 Server Hardware Strategies

Blade Servers
 Server Hardware Strategies

Cloud-Based Compute Services

 Another strategy is to not own any machines at all,
but rather to rent capacity on someone else’s system.
 cloud-based computing lets you benefit from the
economies of scale that large warehouse-size
datacenters can provide, without the expense or
expertise to run them.
 Examples of cloud-based compute services include
Amazon AWS, Microsoft Azure, and Google
Compute Engine.
 Server Hardware Strategies

Cloud-Based Compute Services

 What Is the Cloud?
 There are three common definitions for the cloud, each
coming from different communities:
 Consumers: When a typical consumer uses the term the cloud,
they mean putting their data on a web-based platform.
 Business people: Typically business people think of the cloud
as some kind of rented computing infrastructure that is elastic.
 IT professionals: cloud computing comes down to someone
else maintaining hardware and networks so that customers can
focus on higher-level abstractions such as the operating system
and applications.
 Cloud Computing’s Cost Benefits
 Computer hardware become cheaper but the operation of
computers has become increasingly expensive. The
increased operational cost diminishes and often eliminates
the cost reductions.
 As these cost trends continue, it will become difficult for
companies to justify maintaining their own computers.
 Adoption of cloud computing is also driven by another cost:
opportunity cost. Opportunity cost is the revenue lost due
 Server Hardware Strategies

Cloud-Based Compute Services

 Software as a Service
 Software as a service (SaaS) means using web-based
applications.
 Many small companies have no servers at all.
 They are able to meet all their application needs with
web-based offerings.
 For example, they host their web site using a hosting
service, they use a web-based payroll provider, they
share files using Dropbox, and they use Google Apps for
Work.
 When a company adopts such a strategy, the role of the
IT department becomes that of an IT coordinator and
integrator.
 Server Hardware Strategies

Server Appliances
 An appliance is a device designed specifically for a
particular task.
 The computer world also has appliances:
 file server appliances,
 web server appliances,
 email appliances,
 DNS/DHCP appliances
 Although a senior SA can engineer a system dedicated to
file service or email out of a general-purpose server.
 Purchasing an appliance can free the SA to focus on other
tasks.
 The other benefit of appliances is that they often have
features that can’t be found elsewhere.
 Server Hardware Strategies

Hybrid Strategies
 Most organizations actually employ a number of
different strategies.
 Small organization has a few snowflakes and
possibly a few eggs in one basket.
 Medium-size organizations usually have a
virtualization cluster plus a few snowflakes for
situations where virtualization would not work.
 Large organization have a little of everything.
 The only strategy we recommend against is an
organization trying to deploy all of these strategies
simultaneously.
 Server Hardware Features

Server Hardware Features

 Server hardware is designed with a bias toward
performance and remote manageability.
 Server hardware differs from workstation hardware
both in form factor and in the options available at
time of purchase.
 It is also different in how we configure it and manage
it.
 Server hardware generally falls into one of two major
categories:
 Enterprise: server hardware prioritizes the needs of
commercial applications that enterprises tend to use
 Business App
 Email Server
 File & Web Server
 Cluster: server hardware prioritizes the needs of cluster or
distributed computing applications such as Hadoop,
Cassandra, web server farms
 Server Hardware Features

Server Hardware Features

 Server Hardware Features

Workstations Versus Servers

 Workstations and servers are fundamentally different.
 servers have higher uptime requirements than
workstations, because many people rely on them.
 Server operating systems are different from workstation
operating systems. (e.g. Window Server, Ubuntu Server)
 Server Hardware Design Differences
 Characteristics that differ for servers versus workstations :
 More CPU performance: Servers tend to have faster CPUs and more
of them.
 High-performance I/O: Servers tend to have more I/O capacity.
 Expandability: Servers usually have more physical space inside for
hard drives and more slots for cards and CPUs.
 Upgrade options: Servers are designed for growth.
 Rack mountable: Generic servers and server appliances should be
rack mountable.
 High-availability options: Server hardware should include various
high availability options, such as dual power supplies, RAID.
 Remote management: Server hardware should be capable of being
remotely managed.
 Server Hardware Features

Workstations Versus Servers

 Server OS and Management Differences
 Servers use a different OS and their configurations are
managed differently.
 Server hardware runs a server OS. Microsoft Windows
Server Edition includes additional software for providing
services such as Active Directory.
 In the Linux world, and more recently with Windows
Server Core, the server edition is stripped down to the
bare essentials.
 A system is more reliable and secure when it is
smaller.
 Maintenance is easier because there are fewer software
packages to be upgraded.
 Security is improved because a package that hasn’t
been installed is one fewer source of vulnerabilities.
 Server OSs are often patched on a different schedule.
 Network configuration is generally done by
hardcoding the configuration
 Server Hardware Features

Server Reliability
 Because servers need to be more reliable, and have
higher uptime than workstations.
 one of the ways we prepare for equipment failure is to
buy server hardware with additional features for
reliability and data integrity.
 servers should be housed in a restricted-access,
climate controlled environment, with protected power
—in other words, a computer room or data center.
 Levels of Redundancy
 N+1 redundancy is used when we wish to indicate that
there is enough spare capacity for one failure.
 N+2 redundancy would mean there is enough spare
capacity for two failed power supplies.
 For example, there may be two power supplies in the
system. Either can fail and the system keeps running.
 Server Hardware Features

Server Reliability
 Data Integrity
 Hard disks and SSDs eventually wear out and die. we must
have a plan to deal with this eventuality.
 RAID
 By using RAID (Redundant Array of Independent Disks) levels
1 and higher. When a disk fails, the system keeps running and
we have time to replace the disk.
 RAID 1 stores all data twice, once on each disk of a two-disk
mirror. If one disk fails, the system can simply rely on the
remaining disk. If the second disk fails before restoration is
complete, we lose all data and must restore data from
backups.
 RAID levels 2 and higher are similar but create redundancy
in ways that are more efficient, have better performance, or
can survive two disk failures.
 Server Hardware Features

Server Reliability
 Non-RAID Approaches
 The alternative to RAID is to assume data integrity is
handled elsewhere.
 For example, suppose a group of redundant web servers
all contain the same data, which they receive from a
master.
 The web servers themselves do not need RAID because if
a disk fails, the web server is simply shut down for
repairs and the traffic is divided among the remaining
web servers.
 Another example of handling data integrity elsewhere is
a distributed storage system such as Google’s GFS,
Hadoop’s HDFS, Cassandra, and others.
 These systems store copies of data on many hosts,
essentially creating RAID-like storage that is distributed
among many machines, not just many disks.
 Server Hardware Features

Server Reliability
 Hot-Swap Components
 Hot-swap refers to the ability to add, remove, and
replace a component while the system is running.
 Hot-swappable components increase the cost of a
system.
 This additional cost is justified when it eliminates
downtimes for expansion or repairs.
 The more quickly a failed component can be replaced,
the better.
 RAID systems usually run more slowly until a failed
component has been replaced and the RAID set has been
rebuilt.
 Server Hardware Features

Server Reliability
 Servers Should Be in Computer Rooms
 Servers should be installed in an environment with
proper power, fire protection, networking, temperature
and humidity control, and physical security.
 The two most important points are power and cooling.
 For reliable operations, servers need power that is
reliable and clean.
 Servers are designed to run at a particular operating
temperature, usually around 10 to 35◦C (50 to 95◦F).
 it is best to mount it in the rack immediately before
installing the OS and other software.
 A telecom closet with good cooling and a door that can
be locked is better than having your company’s payroll
installed on a server sitting under someone’s desk.
 Server Hardware Features

Remotely Managing Servers

 Remote management means that it should be possible to
do all system administration tasks involving the machine
from a remote location.
 Remote management systems have the benefit of
permitting you to operate a system’s console when the
machine is in a bad state or otherwise disconnected from
the network.
 Remote access is important because servers are often
housed in machine rooms located around the world.
 Security implications must be considered when you have a
remote console.
 console systems should have properly considered
authentication and privacy systems.
 For example, you might permit access to the console
system only via an encrypted channel, such as SSH, and
insist on authentication by a one-time password
system, such as a handheld authenticator.
 Server Hardware Features

Remotely Managing Servers

 Integrated Out-of-Band Management
 Modern servers have remote management capabilities
built-in.
 Such systems are generically called out-of-band (OOB)
management, and have an Ethernet port as an interface.
 Once an IP address is assigned to the OOB interface, one
can connect over the network to a remote management
subsystem that provides access to the console.
 These remote management systems are isolated enough
from the main system that they are accessible even
when the main system is down or powered off.
 Server Hardware Features

Remotely Managing Servers

 Non-integrated Out-of-Band Management
 Third-party products can add some OOB functionality to
systems that lack built-in remote management.
 Remote Power Cycle: One can gain the ability to
remotely power a machine off and on using a switched
power distribution unit (PDU).
 Remote Console with IP-KVM: A KVM switch is a
device that lets many machines share a single keyboard,
video screen, and mouse (KVM).
 Remote Console with Serial Consoles: Network
equipment, appliances, and older servers have serial
consoles. For sites with many servers, it is common to
have console server software, such as Conserver to
manage the serial consoles.
 Server Hardware Features
Separate Administrative
Networks
 It is common for servers to have a separate NIC that is
connected to a dedicated administrative network.
 Separating administrative traffic from normal service traffic
has a number of benefits.
 the primary benefit is often to isolate disruptive traffic.
Backups, for example, consume a large amount of
bandwidth and can interfere with normal service traffic.
 In addition, the administrative network is often more stable
and static, while the service network is more dynamic.
 This separate network should be engineered to use
separate equipment so that it will not be affected by
outages in the main network.
 Server Hardware Features

Maintenance Contracts and

Spare Parts
 A maintenance contract is an agreement with a vendor that
it will fix or replace hardware and provide other support for
a certain duration of time.
 Vendor SLA
 Vendors tend to have a variety of maintenance contract options,
with different service level agreements (SLAs).
 For example, maintenance contracts may offer to ship a
replacement for a bad part with a 4-hour response time, a 12-
hour response time, or next-day options.
 Spare Parts
 Some vendors have a self-support plan where the customer
purchases a spares kit and the maintenance contract is simply
an expedited replacement plan for any spares that are
consumed.
 The spares kit is less expensive than buying a second machine,
often called a cold spare because it is left powered off (cold) until
needed.
 Server Hardware Features

Maintenance Contracts and

Spare Parts
 Tracking Service Contracts
 Sometimes during an outage we discover that the machine is
not on the service contract.
 It is good practice to write purchase orders for service contracts
for 10 percent more than the quoted price of the contract.
 Put machines on the contract at the time of purchase.
 Track the service contracts as part of the machine inventory.
 Cross-Shipping
 Some vendors will not ship the replacement part until they
receive the broken part.
 This practice significantly increases the time to repair.
 Better vendors will ship the replacement immediately and
expect you to return the broken part within a certain amount of
time.
 This practice is called cross-shipping; the parts cross paths
as they are delivered.
 Server Hardware Specifications

Server Hardware Specifications

 Designing hardware is all about making different tradeoffs.
 Vendors have different product lines.
 Within a product line there are many hardware choices:
CPUs, RAM, storage, and so on.
 Fundamentally we should begin with the application
requirements in mind and use those requirements to guide
each decision, including the amount of RAM, the amount of
storage, the number and types of CPUs, and so on.
 Server performance is very complex, but in general an
application is usually waiting for one of four things: disk,
CPU, memory, or network.
 If we had perfect knowledge of our system, it would have
just enough of all four without any wasted capacity.
 Server Hardware Specifications

Models and Product Lines

 Most vendors have multiple product lines.
 The line that emphasizes initial purchase price is often
called the “value line” and trades expandability to achieve
a lower initial purchase price.
 The line that emphasizes TCO has models that may cost
more initially, but save money in the long term. These
machines have higher performance and additional
expansion capabilities that prevent obsolescence.
 The line that emphasizes performance provides access to
the latest technologies, such as next-generation CPUs, or
amounts of RAM that seem ludicrous by today’s standard
but will be normal expectations in a few years.
 Server Hardware Specifications

Server Hardware Details

 CPUs
 The most fundamental decision in selecting a server is the CPU—
in particular, the choice of a few high-speed cores versus many
slow cores.
 CPU manufacturers started putting multiple CPU cores on the
same chip.
 Instead of a CPU that is twice as fast, you can get a single CPU
with two CPU cores.
 the computing power has doubled, but any single-threaded
program will run more slowly.
 Multiple CPUs and Cores
 Your purchasing choice is often between a few fast cores or many
slower cores.
 If an application could utilize all of the cores, the fastest result would
be achieved with the processor that has the most aggregate
horsepower.
 In contrast, if the application was single-threaded, the fastest result
would
 come from using the model with the fastest core—in this example, the
3.5 GHz cores on the E5-2637 v3
 Server Hardware Specifications

Server Hardware Details

 Multiple CPUs and Cores
 Server Hardware Specifications

Server Hardware Details

 Example Applications
 A single-threaded legacy application would run fastest on a
single fast core. Any additional CPUs would go mostly unused.
 High-throughput web server software systems usually
manage a pool of threads, assigning each incoming HTTP
request to a different thread.
 Graphics computation may or may not benefit from many
CPUs.
 A virtualization cluster generally runs better with many
cores. VMs can be allocated dedicated cores or they can
share cores.
 Other CPU Types
 The most commonly used family of CPUs is the x86-64
architecture, but other systems do exist.
 The ARM architecture is particularly interesting because it has
extremely good power efficiency but slower than x86-64
 ARM is incompatible with x86, so it doesn’t run operating
systems or software that were not specifically ported to the ARM
architecture.
 Server Hardware Specifications

Server Hardware Details

 Memory
 Random access memory (RAM) is where running programs and their data
are stored.
 CPU speed outpaced RAM speed. CPUs were starved for data because the
RAM subsystem could not keep up.
 At the same time the entire system got so fast that the latency (the time
it takes information to travel from one place to another) between the CPU
and the RAM started to dominate system performance.
 To improve this situation, CPU designers created a number of solutions
such as various caching architectures and NUMA.
 Caches
 A cache stores frequently used or soon-to-be used data close to the CPU
for faster access. Caches improve performance but add complexity.
 Modern processors have multiple levels of caches, called Levels 1, 2, 3,
and 4.
 L1 cache is on the core itself.
 L2 cache is near the core, often shared among cores that are on the
same chip. The L1 cache feeds from the L2 cache.
 L3 cache is shared among all CPUs of the machine
 The L4 cache, for CPUs that have it, is also incorporated into the CPU
package. It feeds the L3 cache.
 Server Hardware Specifications

Server Hardware Details

 NUMA
 Another way that CPU designers have improved RAM latency is by
introducing Non-Uniform Memory Architecture (NUMA) RAM systems.
 To improve RAM latency, chip designers created a new RAM architecture.
Each partition, or bank, of RAM would be very close to a particular CPU.
 The other CPUs could access the bank of another CPU, but such access
would be slower or at higher latency
 Designers modified operating systems to take advantage of this improved
access speed.
 Swap Space
 Swap space is a feature of an OS’s virtual memory system
whereby if the system runs out of RAM, the OS can overflow to
disk.
 Disks are approximately 100 times slower than RAM, so the
process of pausing a program to bring a swapped page in from
disk is a performance nightmare.
 Most servers don’t use virtual memory because performance
while swap space is used is terrible.
 However, an application crashing because the system ran out of
memory is even worse.
 Server Hardware Specifications

Network Interfaces
 Servers often have multiple network interface cards (NICs)
because they are connected to different networks,
 Or because they are ganged or grouped together for
increased bandwidth or resiliency.
 A server with many clients needs more bandwidth than one
with fewer clients, assuming all clients generate the same
amount of bandwidth.
 Sometimes multiple NICs are grouped together to provide
more bandwidth or more resiliency.
 For example, two 10 Gbps NICs might connect to the same
switch, providing 20 Gbps aggregate bandwidth. This
practice is often called bonding, or teaming, the available
bandwidth.
 Server Hardware Specifications

Disks: Hardware Versus Software

RAID
 Disks, particularly those with moving parts, are the most
failure-prone components in most systems.
 RAID can be achieved through hardware or software.
 Software RAID
 This setup provides RAID via device drivers, part of OS.
 this approach is that it is inexpensive, because it is included in
the price of the operating system.
 The downside is that it is usually not as fast because it is running
on the same CPUs as the rest of the system.
 Software RAID often has fewer features. For example, it may
support only RAID 1 mirrors.
 Software RAID on the boot disk brings a certain amount of
complexity when a disk fails.
 If that is the disk that failed, the BIOS may not be smart enough
to try the second disk.
 Typically you have to manually reconfigure the boot drive to be
the remaining good disk.
 Server Hardware Specifications

Disks: Hardware Versus Software

RAID
 Hardware RAID
 RAID is implemented in a hardware device, usually a card that is
plugged into one of the expansion slots of the server.
 The server sees each RAID group as a hard disk.
 Hardware RAID usually provides higher performance and more
features than software RAID.
 They usually have dedicated hardware that performs parity
calculations.
 The downside of this approach is the higher cost. RAID
controllers can be very expensive.
 Some motherboards have integrated RAID controllers. These are
often simple RAID controllers that provide RAID 1 mirroring only.
 Server Hardware Specifications

Power Supplies
 The second-most failure-prone component in a server is the
power supply
 It is very common to have N+1 redundant power supplies
so that if one fails, the system can keep running.
 Each power supply should also have a separate power cord.
 Each power supply should draw power from a different
source: a separate circuit or uninterruptible power supply
(UPS).
 Generally each power distribution unit (PDU) in a data
centre is its own circuit, so plugging each power cord into a
different PDU assures two power sources.
 Server Hardware Specifications

Things to Leave Out

 What you don’t want on a server is anything you won’t use.
 This includes fancy video cards, keyboards, monitors, and
mice.
 Servers do not display graphics locally, so fancy video
graphics cards are a waste.
 Servers don’t need beautifully designed front panels and
cases.
 Cases should be designed to provide proper airflow at the
lowest price possible.
 If you look at pictures of Google data centres, you’ll see
rack after rack of servers that are cost reduced to an
extreme.
End of chapter 3