Planning For High Availability: Lesson 7
Planning For High Availability: Lesson 7
Lesson 7
Skills Matrix
Technology Skill
Objective Domain
Objective #
Plan application
servers and services
1.4
Provision data
4.2
5.2
Backing Up Data
The simplest and most common type
of data availability mechanism is the
disk backup, that is, a copy of a
disks data stored on another
medium.
The traditional medium for network
backups is magnetic tape, although
other options are now becoming
more prevalent, including online
backups.
Shadow Copies
Backups are primarily designed to protect
against major losses, such as drive
failures, computer thefts, and natural
disasters.
The loss of individual data files is a fairly
common occurrence on most networks,
typically due to accidental deletion or user
mishandling.
For backup administrators, the need to
locate, mount, and search backup media
just to restore a single file can be a regular
annoyance.
Windows Server 2008 includes a feature
Shadow Copies
Shadow Copies is a mechanism that
automatically retains copies of files on a server
volume in multiple versions from specific points
in time.
When users accidentally overwrite or delete
files, they can access the shadow copies to
restore earlier versions.
This feature is specifically designed to prevent
administrators from having to load backup
media to restore individual files for users.
Shadow Copies is a file-based fault tolerance
mechanism that does not provide protection
against disk failures, but it does protect against
the minor disasters that inconvenience users
and administrators on a regular basis.
Offline Files
Offline Files works by copying
server-based folders that users select
for offline use to a workstations local
drive.
The users then work with the copies,
which remain accessible whether the
workstation can access the server or
not.
No matter what the cause, be it a
drive malfunction, a server failure, or
a network outage, the users can
Offline Files
When the workstation is able to reconnect
to he server drive, a synchronization
procedure.
replicates the files between server and
workstation in whichever direction is
necessary.
If the user on the workstation has modified
the file, the system overwrites the server
copy with the workstation copy.
If another user has modified the copy of
the file on the server, the workstation
updates its local copy.
If there is a version conflict, such as when
users have modified both copies of a file,
Offline Files
To use Offline Files, the user of the
client computer must first activate
the feature, using one of the
following procedures:
Windows XP and Windows Server
2003 Open the Folder Options
control panel, click the Offline Files
tab, and select the Enable Offline
Files checkbox.
Windows Vista and Windows Server
2008 Open the Offline Files control
Disk Redundancy
Disk redundancy is the most common type
of high availability technology currently in
use.
Even organizations with small servers and
modest budgets can benefit from
redundant disks, installing two or more
physical disk drives in a server and using
the disk mirroring and
RAID-1 and RAID-5 capabilities are built
into Windows Server 2008.
For larger servers, external disk arrays and
dedicated RAID hardware products can
provide more scalability, better
Application Resilience
Refers to the ability of an application
to maintain its own availability by
detecting outdated, corrupted, or
missing files and automatically
correcting the problem.
Server Clustering
Server clustering can provide two
forms of high availability on an
enterprise network.
In addition to providing fault
tolerance in the event of a server
failure, it can provide network load
balancing for busy applications.
Failover Cluster
Servers themselves can suffer failures that
render them unavailable.
Hard disks are not the only computer
components that can fail, and one way of
keeping servers available is to equip them
with redundant components other than
hard drives.
The ultimate in fault tolerance, however, is
to have entire servers that are redundant
so that if anything goes wrong with one
computer, another one can take its place
almost immediately.
In Windows Server 2008, this is known as
Heartbeats
The servers in an NLB cluster
continually exchange status
messages with each other, known as
heartbeats.
The heartbeats enable the cluster to
check the availability of each server.
Convergence
When a server fails to generate five
consecutive heartbeats, the cluster
initiates a process called convergence,
which stops it from sending clients to the
missing server.
When the offending server is operational
again, the cluster detects the resumed
heartbeats and again performs a
convergence, this time to add the server
back into the cluster.
These convergence processes are entirely
automatic, so administrators can take a
server offline at any time, for maintenance
TS Session Broker
One problem inherent in the load
balancing of terminal servers is that a
client can disconnect from a session
(without terminating it) and be assigned
to a different terminal server when he or
she attempts to reconnect later.
To address this problem, the Terminal
Services role includes the TS Session
Broker role service, which maintains a
database of client sessions and enables a
disconnected client to reconnect to the
same terminal server.
Summary
In computer networking, high availability
refers to technologies that enable users to
continue accessing a resource despite the
occurrence of a disastrous hardware or
software failure.
Shadow Copies is a mechanism that
automatically retains copies of files on a
server volume in multiple versions from
specific points in time.
When users accidentally overwrite or
delete files, they can access the shadow
copies to restore earlier versions.
Summary
Offline Files works by copying serverbased folders that users select for
offline use to a workstations local
drive.
The users then work with the copies,
which remain accessible whether the
workstation can access the server or
not.
Summary
When you plan for high availability,
you must balance three factors: fault
tolerance, performance, and
expense.
The more fault tolerance you require
for your data, the more you must
spend to achieve it, and the more
likely you are to suffer degraded
performance as a result of it.
Summary
Disk mirroring is the simplest form of
disk redundancy and typically does
not have a negative effect on
performance as long as you use a
disk technology, such as SCSI (Small
Computer System Interface) or serial
ATA (SATA), that enables the
computer to write to both disks at
the same time.
Summary
Parity-based RAID is the most
commonly used high-availability
solution for data storage, primarily
because it is far more scalable than
disk mirroring and enables you to
realize more storage space from your
hard disks.
One way of protecting workstation
applications and ensuring their
continued availability is to run them
using Terminal Services.
Summary
Windows Installer 4.0 is the
component in Windows Server 2008
that enables the system to install
software packaged as files with a
.msi extension.
One of the advantages of deploying
software in this manner is the built-in
resiliency that Windows Installer
provides to the applications.
Summary
A failover cluster is a collection of
two or more servers that perform the
same role or run the same
application and appear on the
network as a single entity.
Summary
The NLB cluster itself, like a failover
cluster, is a logical entity with its own
name and IP address.
Clients connect to the cluster rather
than to the individual computers, and
the cluster distributes the incoming
requests evenly among its
component servers.
Summary
The Terminal Services role includes
the TS Session Broker role service,
which maintains a database of client
sessions and enables a disconnected
client to reconnect to the same
terminal server.
Summary
In the DNS Round Robin technique,
you create multiple resource records
using the same name, with a
different server IP address in each
record.
When clients attempt to resolve the
name, the DNS server supplies them
with each of the IP addresses in turn.
As a result, the clients are evenly
distributed among the servers.