►
From YouTube: Lightning Talk: The Adaptive Edge: Balancing Cost Efficiency, Usability, and Perfo... Tom McCollough
Description
Lightning Talk: The Adaptive Edge: Balancing Cost Efficiency, Usability, and Performance Tradeoffs - Tom McCollough, Section
The network edge presents new challenges in optimizing application deployments, as the deeper an application is pushed into the Internet, the greater the associated costs to run those workloads. This talk will cover a set of optimization considerations and strategies that allow workloads to be as close to users as possible, bounded by a business value construct. At the same time, we will discuss the importance of maintaining simplicity in the usability of the system, so that developers can specify how deeply they want their application to run without overallocation of resources. Using CNCF core technologies, including Kubernetes and Prometheus as a base, this talk establishes a framework for evaluating projects within the CNCF landscape and their suitability for edge use cases. We will examine Kubernetes Federation techniques, Multi-cluster orchestrator systems, and Traffic Direction and Service Discovery strategies against a selection criteria to assist architects in making “good-fit” decisions.
A
A
A
I'm
not
sure
exactly
what
I
was
thinking.
I
guess
I
hadn't
thought
it
through.
I
went
ahead
and
pressed
ok
on
that
on
that
question
and
thereby
accepted
a
little
bit
of
technical
debt
that
as
an
industry,
it
would
take
us
years
to
address
effectively
well.
I've
got
great
news
collectively
as
an
industry
we're
working
to
resolve
that
technical
debt
revisit
that
we're
revisiting
it
with
a
new
software
paradigm,
a
software
paradigm
that
respects
the
physics
of
software
engineering
and
recognizes
that
compute
and
data
need
to
be
strategically
placed
in
our
physical
world.
A
I'm
tom
mccullough,
I'm
a
product
manager
at
section
where
we
work
on
problems
like
these
addressing
problems
at
the
edge.
Today,
I'm
going
to
be
talking
about
trade-offs
of
running
software
at
the
edge
containerized
software
at
the
edge,
specifically
the
trade-off
of
performance
to
your
end
user
versus
the
cost
of
delivering
that
performance.
A
There's
a
hint
in
the
title
to
the
solution,
I'm
going
to
outline
for
you,
the
adaptive
edge
and
what
I'm
going
to
do
is
pick
pieces
from
the
cloud
native
computing
foundation
landscape,
where
you
can
build
a
solution
that
I'm
going
to
describe.
So
I'm
going
to
be
giving
you
a
little
bit
of
road
map
on
how
you
can
balance
these
trade-offs
and
get
this
adaptive.
A
Optimized
solution
so
performance
when
I'm
talking
about
balancing
performance
against
cost,
I'm
talking
about
running
hundreds
of
your
instances
of
your
application
at
the
edge
so
that
you
can
get
on
average,
lower
distance,
lower
latency
fewer
drop
shopping,
carts
whatnot,
but
there's
a
cost
associated
with
that,
not
just
in
terms
of
cycles.
But
in
terms
of
operating
all
of
these
clusters,
I
mean
you'd
love
to
run
in
every
cluster,
can't
really
afford
that
due
to
the
cost.
A
What
if
we
could
be
adaptive?
What
if
we
could
move
the
workload
to
where
it
needs
to
be?
What
if
we
could
maintain
a
set
of
running
workloads
that
are
optimal
for
your,
your
particular
needs
the
optimal
according
to
distance
from
your
end
user,
optimal
according
to
how
many
you're
willing
to
pay
for
what?
If
we
could
continuously
adapt
that
set
of
running
workload,
locations,
adapt
it
according
to
signals,
health
signals
utilization
signals.
A
A
The
signals
I
talked
about
and
doing
that
optimization.
We
need
a
work,
a
way
to
solve
that
problem.
We'll
talk
about
pieces
for
that
and
then
finally,
we
need
to
be
able
to
direct
traffic
as
a
workload
moves
from
one
location
to
another.
Based
on
the
signals
we
will
need
to
direct
the
traffic
to
go
along
with
it,
so
the
multi
multi-cluster
problem
is
well
addressed
in
the
landscape.
There's
a
number
of
solutions-
some
of
them
are-
are
forks
from
the
federation
special
interest.
Group
carmada
is
one
example.
A
I
just
picked
one
that
would
suit
that
allows
you
to
run
manage
multiple
clusters
without
the
cognitive
load
of
of
having
to
think
about
managing
each
cluster
individually,
that
that
solution
is
there
and
ready
to
go
with
regard
to
moving
a
workload
open.
Cluster
management
is
another
tool
that
lets
you
manage
workloads
running
on
multiple
clusters
and
the
multiple
clusters
themselves,
but
it
adds
something
different.
It
adds
a
a
placement
decision,
so
you
can
customize
your
placement
decision.
It's
a
code.
A
You
write
yourself
and
provide
it
to
open
cluster
management
that
tells
open
cluster
management
where
a
workload
needs
to
be
prioritized
or
or
not,
and
so
using
this
and
the
signals
that
we're
going
to
talk
about
in
a
second,
you
can
develop
the
ability
to
to
place
a
decision.
The
thing
about
open
cluster
management's
placement
decision,
however,
is
that
it
operates
on
the
initial
deployment
of
that
workload.
It
doesn't
revisit
it
on
a
cadence
which
we
want
and.
A
Now,
with
signals
and
optimization,
this
is
where
we're
left
empty-handed.
So
the
cncf
landscape
does
have
an
optimization
section
right
in
the
landscape
there,
but
it's
devoted
towards
optimizing
with
this
within
a
single
kubernetes
cluster
and
all
the
examples
there
are
about
optimizing
for
cost.
How
do
I
run
this
cluster
in
an
optimal
way?
You
know
minimizing
compute
resource
that
kind
of
thing
we
want
to
optimize
across
multiple
customer
clusters
and
we
also
want
to
optimize,
take
performance
into
account.
A
You
really
do
want
each
end
user
to
be
routed
to
their
closest
location,
not
to
the
closest
location
of
some
other
end
user
in
that
same
isps
network,
so
that
caching
is
an
issue
and
then,
of
course,
the
time
to
live
is
an
issue
we'd
love
to
use
the
capability
we're
developing
here
to
make
highly
available
services
and
if
a
time
to
live,
isn't
being
honored.
You've
asked
for
a
minute
and
an
isp
is
giving
you
one
hour,
then
that's
the
best
you're
going
to
be
able
to
in
the
worst
case.
A
You'd
your
switch
over
time
would
be
an
hour
for
directing
traffic,
so
dns
has
issues
kubernetes
global
load.
Balancer,
however,
brings
any
cast
into
play
for
us.
So
anycast
is
a
capability
built
into
ip
that
will
that's
purpose,
built
to
route
traffic
from
an
end
user
to
the
closest
location
where
that
workload
is
being
served
and
kubernetes
global
load.
A
Balancer
will
use
bgp
protocol
border
gateway
protocol
to
control
the
anycast
capability
of
ip
and
that's
a
perfect
solution
for
what
we
want
to
do
so
sketch,
in
my
short
little
time
here,
a
very
high
level
road
map
for
all
the
pieces
that
you
need
pulled
from
the
cncf
landscape,
to
build
an
optimized
adaptive
edge
you're,
going
to
be
able
to
balance
performance
against
cost,
get
an
optimal
solution.
You'll
get
the
effect
of
running
everywhere
without
actually
running
everywhere,
without
the
cost
of
running
everywhere.
A
So
during
the
talk,
I've
been
talking
about
performance
against
cost,
but
businesses
are
unique.
They
want
to
put
themselves
forward
in
in
a
distinctive
way.
Maybe
you
want
to
you
want
to
prioritize
something
else
at
your
business.
Maybe,
for
example,
you
want
to
prioritize
your
cloud
minimize
your
clout,
your
your
carbon
footprint.
A
You
could
build
into
your
optimization
engine
the
ability
to
prefer
green
data
centers
or
let's
say
that
compliance
is
important.
You
could
build
into
your
optimization
to
prefer
pci,
compliant
or
other
types
of
compliance,
so
your
business
can
customize
the
optimization
engine
as
needed
and
to
suit
your
purposes.