Manual Chapter :
Plan BIG-IQ Interface Requirements
Applies To:
Show Versions
BIG-IQ Centralized Management
8.3.0,
8.2.0,
8.1.0,
8.0.0
Plan BIG-IQ Interface Requirements
How many interfaces are required for a BIG-IQ deployment
BIG-IQ v8.0.0 can operate in a single network card configuration.
Previous versions required that you have at least two, although only one had to be
configured. Consider the following best practice recommendations for field use.
Because a single network interface card (NIC) is
not supported in versions prior to 8.0.0, you cannot upgrade to BIG-IQ version 8.0.0 if you
are running BIG-IQ in a single-NIC configuration in a previous version.
For lab or proof of concept demonstrations, BIG-IQ can be configured to
use only one interface. In this configuration all data flows through the
out-of-band
interface. Also referred to as eth0, or OOB,
this 10Mb interface can only handle so much.
For production use, F5 recommends a minimum of two interfaces. In this configuration only BIG-IQ management functions use the OOB interface. The rest of the data uses one of the
in-band
interfaces. Because the eth1, eth2, and eth3 interfaces are 100 times faster (10 Gb vs. 100 Mb), they are obviously much better at handling the data requirements of a production environment.
For better isolation, security, and performance F5 recommends adding
additional in-band interfaces. The number of additional interfaces needed for optimal
performance depends on your business needs.
A good understanding of the data flow requirements between each
component in your BIG-IQ solution helps a lot when you consider how many interfaces and
network subnets best suit your network environment. For a discussion of this data flow,
refer to
How does data flow in a BIG-IQ solution
on
support.f5.com
Determine the network configuration needed for your
deployment
The deployment requirements for the BIG-IQ solution you install depends
on which functions you need to perform. Generally, as you add interfaces and subnets to
your BIG-IQ solution, you can expect increases in performance level, security, and
dependability.
BIG-IQ deployment options
What functions does your deployment need to
perform?
Which hardware components do you need?
Optimal number of interfaces and subnets
Manage and configure BIG-IP devices. For example,
you might want to just manage virtual edition licenses or configure local
traffic and security policy objects.
2 BIG-IQ CM systems (1 active, 1 standby)
BIG-IP devices
eth0 to access the BIG-IQ CM.
eth1 and a single (Discovery) subnet to communicate to
the BIG-IQ devices.
Manage and configure BIG-IP devices.
Collect and view Local Traffic, DNS, and Device
statistical data from the BIG-IP devices.
Collect, manage, and view events and alerts from BIG-IP services (like
LTM, APM, or FPS).
2 BIG-IQ CM systems (1 active, 1 standby)
BIG-IP devices
3 or more
DCDs.
An external storage device (if you choose to back up your DCD
content).
eth0 to access the BIG-IQ CM
eth1 and a (Discovery) subnet to communicate to the BIG-IQ devices.
eth2 (see note below the table) and a (Cluster) subnet to communicate to
handle data management and storage on the DCDs.
Manage and configure BIG-IP devices.
Collect and view Local Traffic, DNS, and Device
statistical data from the BIG-IP devices.
Collect, manage, and view events and alerts from BIG-IP services (like
LTM, APM, or FPS).
Separate network traffic to
support large, distributed deployments of the F5 BIG-IQ solution for
improved performance, security, and interactions in multiple data center
environments.
Or, for disaster recovery capability, you could operate
multiple data centers, each with its own set of BIG-IQ systems. (For
additional detail, refer to
Managing Disaster
Recovery Scenarios
in the
Setting up
and Configuring a BIG-IQ Centralized Management Solutio
n article
on s
upport.f5.com
)
2 BIG-IQ CM systems (1 active, 1 standby
3 or more DCDs
BIG-IP
devices
An external storage device (if you choose to back up your DCD
content)
eth0 to access the BIG-IQ CM
eth1 and a (Discovery) subnet to communicate to the BIG-IQ
devices.
eth2 (see note below the table) and a
(Cluster) subnet to communicate to handle data management and storage on
the DCDs.
eth3 (see note below the table) and a
(Listener) subnet to improve performance
If you plan to use the eth2 and eth3 interfaces, but
your VE has only 2 interfaces, use the software application for the hypervisor used to
create your VE instances to add two additional interfaces.
How
does data flow in a BIG-IQ
solution
To design the network structure that your BIG-IQ solution requires, it's
helpful to understand a little bit about how the BIG-IQ operates.
There
are four functions essential to BIG-IQ
operations.
Management (out of band)
Discovery
Listener
Cluster
Each of these functions is enabled by a specific flow of data. Not all BIG-IQ
solutions require all four functions, so if your solution doesn't use one of those
functions, you probably won't need the network infrastructure that supports that function.
Additionally, understanding the data flow requirements for each function helps you
understand how decisions you make regarding the number of network interfaces and the number
of network subnets you use can impact the reliability, performance, and security of your
BIG-IQ solution.
Knowing what data must flow between BIG-IQ components for each function
helps you determine the ideal number of network interfaces needed for your BIG-IQ solution.
The solution you choose and the performance levels you need determine the optimal number of
interfaces you need. For example, if you use BIG-IQ just to manage your BIG-IP licenses,
your setup will be very different than if you use BIG-IQ to centrally manage a set of
applications across multiple data centers.
To get started, consider the generic BIG-IQ setup illustrated in this
diagram.
This illustration shows a BIG-IQ system performing centralized
management functions for four BIG-IP devices. Four data collection devices (DCDs), managing
the data generated by the applications running on the BIG-IP devices. One way to understand
the data flow requirements of this system is to think of the flow hierarchy.
The BIG-IQ uses an out of band (OOB) subnet to perform internal
management functions and to facilitate autofailover high availability.
The BIG-IQ manages your BIG-IP devices, so direct communication
to each of those devices is an obvious necessity. However, to give you the insight
you need for that management, the BIG-IQ needs the analytics, events, and alert data
that those BIG-IP devices store on your DCDs. BIG-IQ uses the Discovery subnet to
communicate with the BIG-IPs and the DCDs.
Data generated by the traffic running on your BIG-IP
applications routes to the DCDs that store and manage this data. This data ranges
from metrics you use to analyze device and application performance to events and
alerts generated by the traffic services running on those BIG-IP devices. This data
flows on the Listener subnet.
The DCD cluster manages your data using an Elasticsearch
database. The database makes replicas of your data and distributes those replicas
throughout the cluster so that no single DCD failure can put your data at risk. This
data flows on the Cluster subnet.
Best practices for which interface to route to which subnet is spelled out in this
table.
Interface
Name
Primary Functions
eth0
Management out-of-band
Best practice is to use the eth0, out-of-band (OOB)
interface to
communicate
with the BIG-IQ CM. The
100 Mb interface speed is not a problem for the modest amounts of data required
for this function.
eth1
Discovery
BIG-IQ uses the Discovery function to communicate
with the BIG-IP devices it manages and the BIG-IQ DCDs that manage the data.
Best practice is to use a dedicated, Discovery self IP interface and subnet for
this communication. Depending on your interface configuration, the eth1
interface has 10-100x more bandwidth than the OOB, management (mgmt)
interface.
eth2
Cluster
BIG-IQ CM and DCDs use the
Cluster
function to route and
manage the Elasticsearch (ES) data replicas that store your data. BIG-IQ CM and
all of the DCDs in the ES cluster use this communication channel. Best practice
is to use the eth2 interface for this communication.
eth3
Listener
BIG-IQ DCDs use the Listener function to receive
data from application traffic event logging and alerts; as well as performance
analytics from the BIG-IP devices. Best practice is to use the eth3 interface
for this communication.
The interfaces on a BIG-IQ Virtual Edition (VE) are
commonly referred to as eth0-ethn, while the interfaces on hardware platforms are referred
to as 1.0-1.n. Because
BIG-IQ
VE is the more common use case, in
this article the term eth0 will be used to describe either eth0 on a VE and interface or
1.0 on a hardware platform.
Do I need multiple networks?
The simple answer is, yes; unless you are deploying BIG-IQ for a lab or
proof of concept demonstration, best practice is to use at least two interfaces. The ideal
number of networks for you depends on which functions your BIG-IQ solution requires and
what type of environment you deploy it on.
Although BIG-IQ can deploy using a single network subnet and interface,
best practice is best to use the following networks and interfaces:
On eth0, an out-of-band network for internal BIG-IQ management,
and autofailover high availability (HA).
On eth1, an in-band network for BIG-IPdevice management (BIG-IQ
CM and DCD inter-node communication) and Cluster (Elasticsearch configuration and
status operations) functions.
On eth2, an in-band network for Listener/Discovery that connects
the BIG-IQ CM to the DCDs and to the managed BIG-IP devices. BIG-IQ uses this network
to receive events and analytics from the BIG-IP devices.
Interface
Recommended Function(s)
Speed
eth0
Management (internal), Autofailover HA
100 Mb
eth1
Management, Cluster, & Listener
10 Gb
eth2
Management, Cluster, & Listener
10 Gb
For lab or proof of concept demonstrations, BIG-IQ can be configured to
use only one interface. In this configuration all data flows through the
out-of-band
(OOB) interface. Also referred to as eth0, this
100 Mb interface can only handle so much.
For production use, F5 recommends a minimum of two interfaces. In this configuration only BIG-IQ management functions use the OOB interface. The rest of the data uses one of the
in-band
interfaces. Because the eth1, eth2, and eth3 interfaces are 100 times faster (10 Gb vs. 100 Mb), they are obviously much better at handling the data requirements of a production environment.
For better isolation, security, and performance F5 recommends adding
additional in-band interfaces. The number of additional interfaces needed for optimal
performance depends on your business needs.
A good understanding of the data flow requirements between each
component in your BIG-IQ solution helps a lot when you consider how many interfaces and
network subnets best suit you. For a discussion of this data flow, refer to How does
data flow in a BIG-IQ solution on
support.f5.com
.
Network environment for proof of concept demonstration
To
deploy BIG-IQ for a proof of concept (POC)
demonstration, you can use a single
network interface and subnet. The available bandwidth and performance level are not
generally suitable for production, but more than sufficient for seeing how the BIG-IQ
Centralized Management works for you.
This figure illustrates the network topology required to deploy BIG-IQ
for a POC demonstration.
Single NIC network topology
Network environment for two subnets
You can deploy BIG-IQ to a production environment with just two network
interfaces and network subnets. Although this option requires fewer resources and
configuration effort, you might need to consider some trade-offs if you choose this
configuration:
Might not provide the bandwidth and performance levels your
solution requires as Elasticsearch queries and replication traffic competes for
bandwidth on a shared network/interface.
Elasticsearch traffic between cluster members (some of which is
not encrypted) is open to the outside world. You can mitigate this issue by using
firewalls to restrict access to this subnet.
This figure illustrates the network topology required to deploy BIG-IQ
with just two subnets.
Dual NIC network topology
When your network topology combines cluster management and listener traffic on the same subnet you need to perform some additional routing work. For detail, see
Setting up routing for an in-band subnet
on
askf5.com
.
For optimum performance, F5 makes the following maximum round trip latency recommendations:
For connections between these components
Round trip latency cannot exceed
between any two DCD or BIG-IQ devices in a DCD cluster
75 ms.
between the BIG-IQ CM and the BIG-IP devices it manages
250 ms.
between the managed BIG-IP devices and the DCDs that collect their data
250 ms.
Routing set up for an in-band subnet
If you configure the networks that support your BIG-IQ solution so that
the discovery and listener functions use the same subnet, F5 recommends that you configure
the routing so that the listener and cluster functions use a different network interface.
When you discover a BIG-IP from the BIG-IQ, you can either use the
BIG-IP device's out-of-band eth0 address or you can use one of the device's self IP
addresses. Using the eth0 address for the discovery function is fine, but you want to make
certain that traffic for the listener functions flows over an in-band, self IP address.
When you configure a BIG-IQ or BIG-IP device, you specify whether to use
the management (mgmt) address (eth0) to discover that device or one of the device's self IP
addresses, which always use another network interface. Once you set the discovery address
for a device, the process for changing it, requires that you reconfigure any DCD cluster
setup you have already completed. Consequently, it's best to get this right the first time.
The volume of traffic that the BIG-IP needs to send to the DCD for
statistics/analytics as well as logging events and alerts is referred to as listener
traffic. This listener traffic can overwhelm the BIG-IP device's eth0 interface. To avoid
performance degradation triggered by excessively high traffic levels, F5 recommends that
you route this traffic over one of the device's self-IP addresses. These self IP addresses
use one of the (eth1-3) interfaces instead of the slower eth0 interface. To ensure that the
listener traffic uses one of the higher speed interfaces, you can configure a routing
scheme.
If you don’t create a route to the DCD listener addresses that prefers
the BIG-IP device's default gateway, it is quite possible that the listener traffic will
use the (10-100 times slower), OOB network.
A simpler way to make sure that traffic for these two functions use a
separate interface and subnet is to add additional subnets.
Network environment for three subnets
F5 recommends that you use at least three network
interfaces and network subnets for the environment into which you
deploy your BIG-IQ Centralized Management solution. This configuration
provides the minimum bandwidth and performance levels recommended for a
production environment.
Isolated, in-band interfaces and VLANs are
available for specific functions.
Better isolation for troubleshooting.
More secure design: Elasticsearch traffic
between cluster members stays within your firewalls.
More bandwidth between nodes
Dual stack IPv4/v6 is supported on the in-band
VLANs. The out-of-band eth0 interface does not support dual stack
IPv4/v6.
The out-of-band management (mgmt) network
supports a floating IP address. The floating IP address is required
for BIG-IQ systems in an automatic failover HA configuration. The
in-band interfaces do not support the floating IP address.
This figure illustrates the network topology
required to deploy BIG-IQ with three subnets.
Three NIC network topology
When your network topology combines cluster management and listener traffic on the same subnet you need to perform some additional routing work. For detail, see
Setting up routing for an in-band subnet
on
askf5.com
.
Network environment for four subnets
Adding a fourth network interface and subnet provides all of the benefits that you get when you use 3 network subnets, but allows you to separate the Listener traffic from the cluster management traffic. This configuration provides greater bandwidth for both of these key functions.
This figure illustrates the network topology required to deploy BIG-IQ with four subnets.
