CHAPTER 13
High Availability and
Disaster Recovery
High availability concepts
• High availability (HA) is simply the ability of a system or a network to
continuously operate without failure.
• A common strategy for setting up HA within an organization is to
implement fault tolerance in the form of redundancy in hardware
components on devices and network infrastructure.
Terminology on high availability
• Mean Time To Repair (MTTR): This is the time required/needed to resolve an
issue.
For instance, if an IT professional spends a total of 60 hours per year repairing a
server during an unplanned maintenance window and the server was repaired 8
times during that same year, then MTTR = Total repair time/number of repair =
60/8 = 7.5 hours.
• Mean Time Between Failure (MTBF): This is the predicted time between the
outages of a system.
For instance, if a critical server operates for 8,745 hours per year and experienced
10 failures within the same year, then MTBF = Total uptime/number of failures =
8745/10 = 874.5 hours.
• Recovery Time Objective (RTO): This is the goal of getting the system up and
running back to a specific service level after an outage has occurred.
• Recovery Point Objective (RPO): This is determined by how data loss is
considered to be acceptable or how far back the data goes to bring the system
back online.
Diverse paths
• if the router between the organization’s corporate network and the internet
goes down, users and client devices will not be able to access the resources
on the internet.
• Having multiple/diverse paths between your organization and the data
center is an important factor to consider when implementing HA concepts.
Diverse paths (contd.)
• Multiple/diverse paths focus on ensuring an organization has more than one available
path to and from a data center or the internet.
• Since data centers are hosting resources, servers, and devices for many customers,
the data center also needs redundancy internet connections from various ISPs within
the region.
• Using multiple ISPs for a data center provides greater redundancy, ensuring access to
the data center resources is always available when needed.
Infrastructure redundancy
• Fault tolerance refers to the ability of a system to continue to operate normally, despite the failure
of one or more of its constituent parts.
• One commonly implemented configuration for fault-tolerant system is load balancing.
• Load balancing is a configuration technique that aims to disseminate workloads among all of the
available resources.
• Incoming traffic from clients is initially directed at the load balancer, which then utilizes its
preconfigured balancing algorithm to determine which of its backend servers will receive the traffic.
• Common load balancing/scheduling algorithms include
• round-robin (a simple algorithm where requests are sequentially distributed to servers as they
arrive),
• weighted round-robin (as with round-robin, but servers are assigned different weightings, and
the ones with higher weightings receive larger shares of incoming requests), and
• least connection (servers with smaller numbers of client connections are preferred over
saturated servers).
Load Balancer
• Clustering refers to the aggregation of several nodes into
a group, such that the group of nodes behaves as though
it were a single node. For example, a server cluster, is
where each server delivers content to clients in the same
manner as a single server would.
• Clustering adds a degree of fault tolerance to a system,
so long as the cluster is configured correctly.
• Another technique commonly used to provide fault
tolerance is network interface card (NIC) teaming.
• NIC teaming refers to a technique in which several NICs on a server are combined into a group to provide higher
capacity or improved fault tolerance to the server.
• When configured to provide increased fault tolerance, NIC teaming balances traffic across all of the NICs and links
in the group, allowing traffic to continue flowing if any of the individual NICs in the group fails.
• This concept of combining several links into one highly available link can also be implemented on network
equipment (such as switches) through the concept of port aggregation.
• Port aggregation allows several physical ports on devices to be combined into one logical port on the device.
• This process can be performed through particular protocols on devices such as the link aggregation control
protocol (LACP).
Active-active versus active-passive configurations
• The concept of active-passive configurations allows network professionals to
install and configure two devices of the same type and function on the network,
allowing only one device to operate at a time.
• If one of the two devices fails on the network, the secondary device can take over
and become the new primary device.
• Within the active-passive configuration, there’s always constant communication
between both devices as they are configured as a pair.
• The configurations and real-time session information between devices in an
active-passive state need to be constantly synchronized with each other as
failover may happen at any time within an organization.
• In an active-active state, two devices of the same type are configured and
operating at the same time.
• This type of configuration is usually more complex to design and operate
compared to the active passive configuration.
• Since both devices are active and forwarding traffic at the same time, the packet
can flow in many different directions.
active-passive mode
• Two routers have been configured to operate in
an active-passive mode,
• where R1 is configured to operate as the primary
router for forwarding packets between the
internet and the internal network and
• R2 becomes the standby router.
• While these two routers are online, they both
exchange keep-alive messages with each other.
• If R2 does not receive the keep-alive messages
from R1 after a specific time, R2 will
automatically assume the role of the primary
router for forwarding packets to and from the
internet for the internal clients.
active-active mode
• R1 and R2 are both
operating in an active-
active state.
• Therefore, traffic from one
computer may take the
outbound path through R1
to access the internet, and
returning traffic may not
take the same path but use
the path through R2 and
back to the computer.