00:00:00
[Music]
00:00:04
so
00:00:05
why cloud why do we choose cloud
00:00:08
as the go-to network management platform
00:00:12
up for our networks well
00:00:16
the reason is it's easier to manage
00:00:20
large dynamic deployments using a
00:00:23
central
00:00:23
management system and it's easier to
00:00:26
manage
00:00:27
and deploy that management system using
00:00:28
a cloud platform it
00:00:30
it it creates operational excellence
00:00:33
operational cost savings
00:00:35
uh simplicity of deployment and also it
00:00:38
gives you the option of rolling out your
00:00:40
deployments quicker especially in a
00:00:42
distributed
00:00:43
so when you have lots of branch networks
00:00:45
branch sites
00:00:46
remote locations or when you need to
00:00:49
roll out a large number of devices in a
00:00:52
short period of time or with limited
00:00:53
staff
00:00:56
so let's take a look at how the
00:00:59
wi-fi or actually network management
00:01:01
architectures um
00:01:03
in this case wi-fi management
00:01:04
architectures evolved in the past
00:01:06
and why was that evolution necessary
00:01:10
so it all started around 20 years ago
00:01:14
in warehouses so wi-fi was first rolled
00:01:16
out to warehouses because it was
00:01:18
more convenient to have a handheld
00:01:20
scanner that's not attached to any
00:01:21
cables
00:01:22
to be able to go for that warehouse and
00:01:23
scale scan the labels
00:01:26
send them to a centralized erp system
00:01:29
and it was simply
00:01:33
easy easier and more cost effective
00:01:36
to deploy and manage networks that way
00:01:40
it provided business benefits immediate
00:01:43
business benefits to
00:01:45
the customers that were using the
00:01:47
handheld scanner type of deployments
00:01:49
now the networks at the time
00:01:52
were usually isolated autonomous access
00:01:55
point that didn't share any information
00:01:56
or any control plane with other access
00:01:58
points
00:01:59
and when the number of those access
00:02:00
points grew to a certain amount
00:02:02
usually there was a separate nms system
00:02:05
snmp based or something similar that
00:02:07
would
00:02:08
monitor those aps but those access
00:02:10
points would not
00:02:11
actively cooperate in terms of rf
00:02:15
management in terms of
00:02:16
handover in terms of security they were
00:02:18
just completely autonomous
00:02:20
isolated systems
00:02:23
now what happened with wi-fi is it
00:02:26
went beyond the warehouses so obviously
00:02:28
that was the very first use case but
00:02:30
then
00:02:31
it became apparent that having the
00:02:33
option of mobility
00:02:34
in an office environment actually proved
00:02:37
to be very beneficial as well
00:02:38
you know you're no longer bounded by the
00:02:40
wire you can move around
00:02:42
and still be able to access all of your
00:02:45
resources on your corporate network
00:02:48
so the amount of access points deployed
00:02:51
in such a scenario
00:02:52
was much larger than the amount of
00:02:55
access points
00:02:56
that you would deploy in a warehouse you
00:02:57
deploy a couple of access points you
00:02:59
need to cover the whole warehouse
00:03:00
because
00:03:00
all that was important there was
00:03:02
coverage and the application
00:03:04
that was used on those handheld scanners
00:03:06
didn't use a lot of bandwidth
00:03:08
now when you go to an enterprise
00:03:09
scenario obviously you're using
00:03:11
different kinds of application
00:03:13
email databases and then real-time
00:03:15
applications like voice and video
00:03:17
so the demand on wi-fi becomes
00:03:21
much harsher and
00:03:24
there's more access points involved and
00:03:26
more importantly
00:03:28
the environment becomes much more
00:03:30
dynamic
00:03:31
and then you need to start thinking
00:03:32
about things like rf management
00:03:35
safe and fast and secure roaming how do
00:03:37
you consolidate all those clients how do
00:03:39
you make sure that they have
00:03:41
undisturbed connectivity end to end
00:03:43
throughout your campus
00:03:45
and that's where something called a
00:03:47
controller
00:03:48
came into place now the reason for
00:03:52
deploying a controller was
00:03:53
at the time the access point themselves
00:03:55
didn't have enough resources
00:03:57
to be able to run all this intelligence
00:04:00
on top of being able to provide
00:04:02
connectivity to the client devices
00:04:04
so a physical controller a physical
00:04:05
piece of hardware a dedicated with a
00:04:07
dedicated piece of software was used
00:04:09
to manage the rf environment provide
00:04:12
fast and secure roaming capabilities
00:04:14
key os and provide also provide
00:04:17
connectivity to the distribution system
00:04:19
so it would also terminate the data
00:04:20
plane
00:04:21
on top of that you'd be running a
00:04:23
separate nms just like you would in a
00:04:26
autonomous ap scenario but the
00:04:28
controller made it possible to
00:04:30
to manage the dynamics of the wi-fi
00:04:33
network so that's how controllers
00:04:35
came into place with the evolution
00:04:38
of well computing and with the
00:04:40
introduction of
00:04:42
virtualization where we'd be able to use
00:04:46
the same physical resources for multiple
00:04:49
applications or
00:04:50
multiple operating systems the
00:04:52
controller the physical control came
00:04:54
from a
00:04:55
dedicated server to a virtual machine
00:04:57
that's what we call
00:04:58
a virtual control deployment
00:05:02
obviously that provides cost benefits
00:05:04
because a
00:05:05
you can run that controller on the same
00:05:07
hardware
00:05:08
that you're using for your email
00:05:11
exchange server or
00:05:12
other types of applications
00:05:16
and at the same point the harder on the
00:05:18
access points evolve
00:05:20
so the price of the cpu and memory
00:05:23
started becoming more affordable so we
00:05:25
got more and more resources in the
00:05:27
access points and then what also
00:05:28
happened at the same time as
00:05:30
some of the controller functionality was
00:05:33
actually implemented on the software
00:05:34
running on top of the aps and that
00:05:36
became a hybrid ap so
00:05:38
for some environments like remote
00:05:40
branches you would run the hybrid ap to
00:05:42
manage the dynamics of the environment
00:05:44
within the branch whereas the central
00:05:46
virtualized controller would handle
00:05:48
all the campus-based access points
00:05:52
with the evolution of cloud networking
00:05:55
that controller function was taken
00:05:57
from your data center from your
00:06:00
hypervisor from your virtualization
00:06:02
solution to
00:06:03
a public cloud and that controller
00:06:07
function would be implemented off-prem
00:06:09
or in a public cloud environment and the
00:06:11
access point
00:06:13
that same access point would then talk
00:06:14
to the cloud to the cloud controller in
00:06:17
order to provide
00:06:18
rf management and fast and secure
00:06:20
roaming so
00:06:21
that's how a cloud controller was born
00:06:24
the
00:06:25
final in the fifth step of the evolution
00:06:27
is a distributed control now
00:06:28
the problem the problem when you take a
00:06:31
controller and put it in the public
00:06:33
cloud is
00:06:33
you're creating a dependency on
00:06:36
your wan connections for in order for rf
00:06:40
management infested secure roaming
00:06:43
qos application visibility and control
00:06:45
all these
00:06:46
features that are needed to manage the
00:06:48
dynamics of the wi-fi network to work
00:06:51
are dependent on your wan connections if
00:06:53
your wing goes down
00:06:54
or if somewhere in the transit between
00:06:57
your ap
00:06:58
and your virtual cloud controller
00:07:01
there's a disruption of service
00:07:03
you have a disruption of service on your
00:07:05
wi-fi which is not ideal
00:07:07
so that's the reason why distributed
00:07:08
control was created
00:07:10
and distributed control keeps all the
00:07:12
control plane local
00:07:14
and the only the only part of the system
00:07:18
implemented in the cloud is the network
00:07:20
management system so the control place
00:07:22
stays local there's no centralized
00:07:24
control there's no controller
00:07:26
the access points implement protocols
00:07:29
that provide control plane functionality
00:07:33
without having a central
00:07:37
point of decision making so they make
00:07:39
all those decisions
00:07:40
by communicating through those protocols
00:07:42
between one another and the cloud is
00:07:44
only used to manage and monitor so to
00:07:46
configure
00:07:47
monitor gather information and if the
00:07:49
when
00:07:50
connection goes down between your site
00:07:52
and the cloud
00:07:54
your network is still functional the
00:07:55
only thing you don't have
00:07:57
is immediate visibility into what's
00:08:00
going on in a network for a single pane
00:08:01
of glass so for that cloud platform
00:08:03
because there's no connection to the
00:08:04
site
00:08:06
however the access points will cache all
00:08:08
the monitoring data and once the one
00:08:10
connectivity is established that data is
00:08:12
going to come back and it's going to be
00:08:13
pushed
00:08:14
into your management system so while
00:08:18
you're not able to actively change any
00:08:20
configuration
00:08:22
during the time that you don't have one
00:08:24
connectivity between the side and the
00:08:25
cloud
00:08:26
everything else is still up and running
00:08:28
and that's the major benefit
00:08:30
over all controller based deployments so
00:08:33
you don't
00:08:34
have any central point of failure you
00:08:36
don't have any central bottlenecks
00:08:38
everything is done locally when it comes
00:08:41
to the data in the control plane
