The Open-Source Monitoring Landscape

The
Open-‐Source
Monitoring
Landscape
Michael
Merideth

Sr.
Director
of
IT,
VictorOps

mike@victorops.com,
@vo_mike

My
History
and
Background
•Working
in
IT
since
the
mid
90’s

•Mostly
at
startups
in
the
Colorado
Front
Range
area

•Along
for
the
ride
during
the
“dot
com
boom”

•Build
my
career
using
open-‐source
tools
Since
the
90’s
now,
there’s
been
a
sharp
division
in
tools
and
methodology
between
the
enterprise
space
and
the
startup
and
small
business
communi;es.

Obviously,
smaller
businesses,
especially
in
the
tech
sector,
were
early
and
eager
adopters
of
open-‐
source
technology,
and
much
quicker
to
learn
to
rely
on
open-‐source
tool
chains
in
business-‐cri;cal
applica;ons.

Up
at
the
enterprise
level,
at
the
public
companies,
they’re
only
now
fully-‐embracing
Linux
as
a
business-‐cri;cal
tool,
and
I
think
that’s
largely
because
“the
enterprise”
is
star;ng
to
be
defined
by
companies
that
either
came
up
in
the
dot
com
era,
like

Google,
or
built
the
dot
com
era,
like
Sun,
or
Dell,
or
let’s
say
RedHat.

So,
the
“enterprise”
had
their
toolchain,
built
on
commercial
solu;ons
like
HPUX
and
OpenView
and
so
on,
and
the
startup
community,
the
“dot
com”
community
had
a
completely
different
toolchain,
based
on
Linux,
based
on
open
standards
and
protocols,

built
with
open-‐source
components
like
GNU,
and
Apache,
and
ISC
Bind
and
so
on.

I’m
lucky
enough
that
I’ve
been
able
to
spend
my
career
in
the
startup
sphere,
working
with
that
open-‐source
toolchain.

I
started
working
in
IT
in
the
mid
90’s
in
the

Colorado
front
range,
and
I’ve
spent
my
;me
since
then
working
for
and
consul;ng
at
early-‐stage
startups
and
other
“non
enterprise”
shops.

I’ve
been
able
to
see
first-‐hand
as
the
OSS
toolchain
has
gone
from
seat-‐of-‐the-‐pants
flying
to
mature,
stable,

mission-‐cri;cal
solu;ons
(with
seat-‐of-‐the-‐pants
mode
s;ll
available
for
the
adventurous
of
course).

I’ve
always
been
interested
in
monitoring
and
telemetry,
and
the
ways
it
can
help
make
life
easier
for
people
like
me,
so
I
really
have
my
dream
job
now,
working
for
VictorOps
in
Boulder,
Colorado.

We
spend
a
lot
of
;me
integra;ng
our
product
with

monitoring
tools,
which
gives
me
an
opportunity
to
see
a
lot
of
what’s
out
there,
and
hear
from
our
customers
about
what
they’re
using
and
how
it’s
working
for
them.

Open-‐Source
Monitoring
•Started
out
as
shell
scripts

and
“Big
Brother”

•Better
tools
emerged
in

the
late
90’s

•For
much
of
the
2000’s,
not

much
changed
in
the
space

•Now
things
are
shaking
up!
So
that’s
why
I’m
here
to
talk
to
you
about
open-‐source
monitoring
today.

When
I
started
in
IT,
the
free
monitoring
op;ons
were
a
li[le
thin
on
the
ground,
and
most
of
us
were
running
scripts
out
of
cron
and
using
other
homegrown
solu;ons
to
get
the
job

done.

Here
you
can
see
a
screenshot
of
“Big
Brother”,
which
was
a
common
OSS
tool
at
the
;me.

Isn’t
that
nice?

I
love
the
grid
view,
which
assumes
that
every
service
runs
on
every
host.

That’s
just
good
design.

In
the
late
90’s,
be[er
open-‐source
monitoring
tools
started
to
emerge
and
get
adopted.

And
for
several
years
a]er
that,
not
much
changed.

A
great
deal
has
been
happening
in
the
monitoring
space
lately
though.

I
think
it’s
in
response
to
the
rise
of

virtualiza;on
and
cloud
hos;ng,
and
in
response
to
the
rise
of
DevOps
business
methodologies.

Some
older,
more
established
solu;ons
are
making
changes,
or
ge_ng
challenged
by
new
approaches,
and
I
think
a
line
is
being
drawn
between
a
more

enterprise-‐like
all-‐in-‐one
approach
to
monitoring
solu;ons
and
a
more
modular
approach.

Monitoring
categories
Func;onal
There’s
three
main
categories
in
monitoring
where
I’m
going
to
be
concentra;ng
my
focus
here.

First
is
tradi;onal
func;onal
monitoring,
the
type
of
product
that
pages
you
when
your
website
goes
down.

Nagios
is
an
obvious
example
of
a
func;onal

monitoring
product.

Monitoring
categories
Func;onal
Time

Series

Then
we
have
;me-‐series
data
collec;on,
where
we’re
more
frequently
genera;ng
graphs
than
genera;ng
alerts,
though
there
certainly
can
be
overlap.

Cac;
and
Graphite
are
examples
here.

Monitoring
categories
Func;onal
Time

Series
Event
Finally
we
have
event
processing,
where
we’re
looking
at
log
content
and
other
near
log-‐type
events,
and
this
might
get
visualized
like
;me
series
data,
or
it
might
generate
an
alert
like
a
func;onal
monitoring
system.

Logstash
is
a
good
example
in
this

category.

So
you
can
see
there’s
a
lot
of
overlap
to
these
categories,
and
most
organiza;ons
are
using
a
combina;on
of
all
three.

I’ll
be
examining
some
projects
that
take
on
a
single
category
here,
and
a
few
that
a[empt
a
broader
scope.

Why
OSS
Monitoring?
The
practice
of
Systems
Administration
is
changing

quickly

•DevOps

•Infrastructure
as
Code

•Virtualization

•Cloud
Hosting
and
Elastic
Environments

Commercial
solutions
can’t
keep
up!
First,
though,
I
want
to
talk
about
why
open-‐source
monitoring
is
so
important.

Business
methodologies
are
changing,
from
the
ground
up.

I
don’t
know
about
you,
but
the
day-‐to-‐day
prac;ce
of
my
job
looks
radically
different
than
it
looked
five
years
ago.

DevOps,
systems
management
frameworks
and
“infrastructure
as
code”
change

the
picture
of
how
I
want
my
management
tools
to
work
together.

Virtualiza;on
and
cloud
hos;ng
mean
that
network
architectures
have
been
changing
as
well,
and
even
the
defini;on
of
“on-‐premise”
is
evolving.

But
what
hasn’t
changed
is
the
absolute

requirement
of
availability.

That
means
failure
detec;on,
that
means
capacity
planning,
that
means
predic;ve
analysis,
all
while
working
at
“elas;c”
scale.

Proprietary
tools
have
a
tough
;me
in
this
environment.

Maybe
the
licensing
model
doesn't
work
with
a
constantly
changing
network
size,
or
maybe
it
isn’t
realis;c
about
mul;-‐site
setups.

Availability
features
of
the
so]ware
itself
become
a
“premium”

feature.

Maybe
something
about
the
product
makes
network
architecture
demands
that
you
simply
cannot
meet.

And
of
course
commercial
solu;ons
are
frequently
not
excited
about
integra;ng
with
compe;tors’
products.

So,
the
commercial
tool

becomes
a
constraining
force
in
the
DevOps
workflow,
and
you
either
lose
velocity
because
of
it,
or
you
limit
your
use
of
the
tool,
and
pay
a
price
in
quality.

Open-‐source
tools
tend
to
integrate
more
easily
with
systems
management
frameworks
like
Puppet

and
Chef,
as
well
as
with
other
monitoring
tools.

They
can
be
deployed
experimentally
without
making
a
big
financial
commitment.

You
don’t
have
to
worry
about
“return
on
investment”
if
you
only
want
to
use
a
subset
of
features.

So,
open-‐source
tools

can
become
a
process
accelerator
rather
than
ge_ng
in
the
way.

What
about
SaaS?
•There
are
some
great
options
out
there

•It
can
be
a
great
way
to
get
started

•It
can
be
a
great
second-‐line
of
defense

•Cost
can
scale
very
quickly

•Even
in
a
cloud-‐hosted
environment,

you
will
still
need
“on
premise”

monitoring
Now
there’s
a
ton
of
new
SaaS
offerings
in
the
monitoring
space.

At
VictorOps
we’re
happy
to
work
with
them
all
of
course,
and
they
can
provide
great
point
solu;ons,
especially
in
the
early
product
development
stages.

I’m
also
a
big
fan
of
using
external

monitoring
services
to
provide
a
second
line
of
defense
in
case
your
primary
monitoring
solu;on
fails,
and
they
can
also
be
a
great
way
to
see
how
you’re
performing
from
different
parts
of
the
world.

However
they
tend
to
be
priced
by
the
data
point,
so
as

you
scale
up
you
tend
to
wind
up
going
from
watching
everything
to
watching
a
sta;s;cally
significant
sample,
to
watching
a
sta;s;cally
insignificant
sample,
and
it’s
the
same
problem
as
enterprise
so]ware.

You
under-‐monitor,
and
you
sacrifice
quality.

So,

if
you’re
scaling
up
at
all,
you
can
use
managed
services,
but
you’re
s;ll
going
to
want
“on-‐premise”
solu;ons
that
you
can
scale
at
will.

Functional
Monitoring
Tools
All
right,
so
let’s
start
looking
at
the
opens-‐source
tools
that
are
out
there,
star;ng
with
func;onal
monitoring.

This
isn’t
a
comprehensive
list
of
every
project
out
there,
so
your
personal
favorite
might
not
be
represented.

But
these
are
some
of
the
projects

that
I
thought
were
interes;ng
or
important
for
one
reason
or
another.

Nagios
Everybody
Loves
Nagios:

• Much
better
than
its

predecessors

• Simple
architecture

• Thriving
ecosystem
of

plugins
and
add-‐ons

• Well-‐understood
and

popular Image
Source:
linuxscrew.com
For
about
15
years
now,
the
500
pound
gorilla
of
open-‐source
monitoring
has
been
Nagios.

It
is
the
most
commonly
used
integra;on
we
see
at
VictorOps
by
a
wide
margin.

When
it
came
along,
most
of
us
were
using
“Big
Brother”
for
monitoring,
or

homegrown
scripts
running
out
of
cron.

In
that
world,
Nagios
was
a
revela;on.

It’s
strength
is
that
it’s
really
just
a
scheduler
and
event
handler,
and
all
the
complicated
check
work
is
done
by
a
highly
modular
and
portable
system
of
plugins.

It’s
easy
to

write
a
check
plugin
for
nagios,
and
given
that
the
checks
are
standalone
programs
that
are
called
with
a
fork
and
exec,
the
plugins
can
be
easily
re-‐used
by
other
monitoring
daemons
like
NRPE
or
by
completely
different
solu;ons.

Plugins
have
been
wri[en

to
do
complex,
deep
monitoring
of
most
devices
and
applica;ons
out
there,
it’s
a
thriving
ecosystem.

The
configura;on
language
for
Nagios
is
well
understood,
even
though
it
can
lend
itself
to
some
ambiguity
and
it’s
possible
for
nagios
configura;ons
to
get
impossibly
complex.

It
mainly
has
the
advantage
of
having
been
ubiquitous
for
so
long
that
virtually

everyone
in
the
industry
has
some
experience
with
Nagios
configs.

Systems
management
tools,
like
Puppet,
have
built-‐in
capabili;es
for
genera;ng
nagios
configura;ons,
so
it’s
rela;vely
easy
to
build
a
configura;on
that
automa;cally
picks
up
changes
in

your
network
architecture.

And
again,
it’s
extremely
flexible
and
can
be
adapted
for
a
lot
of
different
use
cases.

Nagios
Everybody
Hates
Nagios:

• Many
features
require
add-‐on
software

• WebUI
is
clunky
and
inefficient

• Not
scalable

• Not
highly-‐available

• Config
language
is
ambiguous
and
confusing
Nagios
is
not
without
its
cri;cs.

Again,
Nagios
core
is
rela;vely
simple
so]ware.

It
schedules
checks,
parses
the
output,
maybe
takes
ac;on,
and
has
a
basic
web
interface.

It
doesn’t
generate
graphs
or
track
metrics
without
add-‐on
so]ware.

It
doesn’t

parse
log
files
or
deal
with
event
streams.

It
doesn’t
do
automa;c
discovery.

This
is
all
true,
but
these
are
all
inten;onal
choices
that
are
built
in
to
the
architecture,
and
part
of
the
reason
why
it’s
so
modular
and
flexible,
and
rela;vely
good
at
what
it
does.

There
are
some
more
general
and
legi;mate
complaints.

The
web
interface
is
clunky
and
dated,
and
hasn’t
changed
substan;ally
in
years.

Version
3,
the
version
that’s
in
widest
use
right
now,
and
is
in
the
repos
of
the
major
Linux
distribu;ons,
is
a
bit
of
a

resource
hog
and
has
trouble
scaling
past
a
few
thousand
data
points.

It’s
not
cluster-‐aware
or
designed
for
high
availability.

It
doesn’t
lend
itself
well
to
mul;-‐datacenter
use
or
horizontal
scaling.

Nagios’s
configura;on
language
is
well-‐known,
but
widely

cri;cized
for
for
being
ambiguous
and
confusing.

Nagios
Fork’s
A-‐Comin’

• Community
dissatisfaction
with
pace
of
development

• Differing
views
on
product
scope
and
architecture

• 2009:

Icinga
fork
is
announced

• Shinken
follows
quickly

• Naemon
is
a
newer
fork

• In
2015,
things
are
getting
interesting…
The
add-‐on
ecosystem
has
addressed
a
lot
of
those
complaints,
but
there
was
growing
dissa;sfac;on
in
the
community
about
the
pace
of
development
of
the
core
Nagios
project
and
the
fact
that
Ethan
Galstad,
the
original
developer
and
the
CEO
of
Nagios

Enterprises,
was
ac;ng
as
a
sole
gatekeeper
for
changes
to
the
core
codebase.

Some
people
in
the
community
had
different
ideas
about
product
scope
and
what
features
should
be
included,
and
others
had
different
ideas
about
the
overall
product

architecture.

In
2009,
over
ten
years
a]er
Nagios’s
original
release
as
NetSaint,
a
group
of
developers
announced
a
fork
of
the
Nagios
project
called
“Icinga”,
and
it
has
been
gaining
in
popularity
ever
since.

More
recently,
the
Shinken
and
Naemon
projects

have
appeared
as
well.

All
four
of
these
projects
have
taken
radically
different
views
on
design,
architecture,
scope,
and
mone;za;on
strategy.

Now,
as
with
all
so]ware
projects
where
human
developers
and
egos
are
involved,
there
has
been
no
shortage
of
digital
ink
spilled
about
the
poli;cal
controversy
surrounding
these
forks,
and
there
have
been
cri;cal
statements
made
by
all
the
par;es
about

one
another.

That’s
well-‐covered
online,
so
I
won’t
wade
in
to
it
here.

Read
all
the
accounts
before
you
form
an
opinion,
that’s
my
advice.

The
only
opinion
I
have
is
about
whether
the
controversy
affects
me
personally,
and
for
me
that’s
a
“no”.

As
a
user
of
open-‐source
so]ware,
I
think
the
forking
of
Nagios
has
been
a
big
posi;ve.

Icinga
version
1
brought
a
lot
of
long-‐needed
improvements
to
the
web
UI,
and
added
support
for
addi;onal
database
backends,
among
other
things.

At
the
same
;me

it’s
mo;vated
Nagios
to
add
more
developers
to
the
core
project,
and
to
quicken
the
pace
of
development.

So
the
users
of
both
projects
have
benefi[ed.

I
think
we’ll
need
to
wait
and
see
what
the
ul;mate
impact
will
be
of
the
Shinken
and
Naemon

projects.

There
are
big
changes
happening
right
now
though.

Both
Nagios
and
Icinga
introduced
new
versions
in
2014
that
put
them
on
a
wildly
divergent
path.

Nagios
4
and
Icinga
2
are
both
major
rewrites,
with
Icinga
2
going
so
far
as
to
adopt
a
new
configura;on

language.

If
Icinga
was
perceived
as
“Nagios
with
nicer
UI”
before,
it
is
now
a
substan;ally
different
product
with
different
design
and
architecture
goals.

Icinga
2
•Totally
rewritten
from
the
ground
up

•New
command-‐line
interface

•Several
built-‐in
modules

•Reload
config
without
interrupting
checks

•Totally
new
configuration
language

•Includes
a
tool
for
converting
Nagios
configs
Icinga
2
is
a
complete
ground-‐up
rewrite.

It
provides
a
new
cli
management
interface
with
access
to
program
and
monitoring
configura;on
objects.

Several
common
add-‐on
features
(database
integra;on,
performance
data
collec;on)
are
now
built-‐in

modules
that
can
be
enabled
or
disabled
as
desired
from
the
CLI.

It
includes
several
performance
features,
including
the
ability
to
reload
configura;on
data
without
interrup;ng
running
checks,
and
if
you
work
in
a
shop
that
updates
Nagios
config
a
lot,
you

know
that’s
a
big
deal.

The
configura;on
language
is
completely
changed
from
the
Nagios
standard.

They
do
include
a
tool
for
conver;ng
Nagios-‐format
configs
to
the
new
Icinga
2
format.

The
config
language
is
said
to
adopt
a
“one
best
way”
design
philosophy,
with
the
idea
of

elimina;ng
the
ambiguity
that
can
arise
in
a
Nagios
configura;on.

It’s
object-‐based
and
rule
driven,
which
means
a]er
ini;al
templates
are
defined,
managing
changes
in
configura;on
should
be
simple.

I
love
the
look
of
Icinga
2.

We’re
pilo;ng
it
at

VictorOps
right
now,
and
we
may
be
adop;ng
it
for
produc;on
use.

Icinga2
Here’s
a
screenshot
of
Icinga
running
the
new
Icinga-‐web
2
webUI.

You
can
see
it’s
a
much
more
modern-‐looking
interface
than
Nagios,
but
design
is
just
part
of
it.

They’ve
made
some
rela;vely
minor
changes
to
the
way
you
interact
with
mul;ple
services

or
mul;ple
hosts
at
a
;me,
that
have
produced
big
produc;vity
gains
over
the
Nagios
UI.

Nagios
4
•Lightweight
worker
processes
to
spawn
checks

•API
improvements

•Some
features
have
been
obsoleted

•Faster
development
and
more
core
developers
Nagios
4
is
also
a
major
rewrite,
but
obviously
s;cking
much
more
closely
to
the
original
Nagios
paradigm.

Nagios
core
now
spawns
lightweight
worker
processes
that
handle
check
execu;on,
and
that
has
greatly
increased
eﬃciency.

Nagios
has
also
made

improvements
in
APIs
and
event
handling,
and
obsoleted
some
features
that
were
never
fully
implemented
in
previous
versions
of
Nagios
like
failure
predic;on,
or
that
had
become
diﬃcult
to
maintain,
like
the
embedded
Perl
interpreter.

Nagios
has
also
added
new
developers
to
the
Nagios
Core
team,
and
a
succession
strategy
is
being
worked
out.

They’ve
signaled
that
they
intend
to
increase
the
pace
of
development
on
the
core
product.

Nagios
4
The
nice
thing
about
Nagios
4
is
that
it
accomplishes
all
this
while
maintaining
configura;on
fidelity,
and
in
fact
the
day-‐to-‐day
opera;on
of
Nagios
4
is
essen;ally
unchanged
from
earlier
versions.

This
means
that
if
you’re
using
Puppet
to
manage
your

Nagios
configura;on,
you
can
upgrade
to
Nagios
4
quite
easily
and
enjoy
the
performance
benefits
without
a
ton
of
work.

On
the
downside,
the
webUI
s;ll
hasn’t
evolved
significantly
from
Nagios
3,
and
is
star;ng
to
feel
really
dated.

Lots
of
common
tasks

are
tedious
and
difficult
from
the
UI.

They
have
changed
the
default
look-‐and-‐feel
from
“black”
to
“white”,
so
it’s
a
more
readable
interface,
but
it’s
s;ll
frames
and
tables.

Shinken
•Complete
reimplilention
in

Python

•Modular
architecture
with

multiple
processes

•Config-‐compatible
with
Nagios
Shinken
aims
to
have
it
both
ways,
being
a
complete
reimplementa;on
of
Nagios
core
as
a
series
of
modular
Python-‐based
processes,
but
that
maintains
legacy
Nagios
conﬁgura;on
compa;bility.

They
boast
very
impressive
performance
numbers,
and
an

architecture
that
provides
for
high-‐availability
and
high-‐scale.

That
comes
at
the
cost
of
having
a
lot
of
moving
parts
and
a
lot
of
external
dependencies.

It’s
also
distributed
using
the
PIP
packaging
system,
and
then
it
has
it’s
own
internal
package
repository-‐
like
func;onality
for
adding
modules
to
the
so]ware.

So
if
you
like
the
idea
of
layers
upon
layers
of
package
managers
and
dependencies,
then
Shinken
may
be
for
you.

There’s
also
a
commercial
Shinken
Enterprise
product,
which
includes
some
addi;onal
features
not
included
in
the
community
edi;on.

So
if
you
like
the
idea
of
PAYING
for
external
dependencies
and
extra
package
managers,
be
sure
to
check
that
out.

Naemon
•Forked
from
Nagios
4

•Aims
to
add
frequently
requested
features

•Non-‐commercial
entity
in
control

•Explicit
goal
of
a
more
open
development
process
Finally
I’ll
men;on
Naemon,
which
is
a
more
recent
fork,
this
one
based
on
the
Nagios
4
codebase.

The
goal
of
the
project
is
to
introduce
a
lot
of
the
same
community-‐driven
feature
requests
that
Icinga
and
Shinken
do,
but
also
with
an
explicit
goal
to
have

a
non-‐commercial
en;ty
managing
the
project
and
ensuring
an
inclusive
development
process.

So
if
you
ﬁnd
that
the
poli;cs
around
using
Nagios
or
Icinga
or
Shinken
really
bother
you,
then
Naemon
might
be
a
good
choice.

They
just
hit
their
1.0
release
a

bit
over
a
week
ago
so
that
project
is
really
just
ge_ng
going.

Commercial
Nagios
Derivatives
•Groundwork

•Opsview

•Others,
I’m
sure

•If
you
want
to
pay
for

Nagios,
pay
Nagios
Image
source:
Cafepress
There
are
some
explicitly
commercial
products
out
there
that
are
based
on
Nagios
as
well;
Groundwork
and
Opsview
are
two
examples.

I
don’t
see
a
lot
of
value
in
them,
they
don’t
really
oﬀer
anything
over
Nagios
core
except
visual
cru]
and
a
licensing
fee.

If
you
want
to
pay
to
run
Nagios,
then
you
can
always
pay
Nagios.

Sensu
•Similar
scope
to
Nagios

•Architected
as
a

composable
framework

•Several
external

dependencies
Image
Source:
florin.myip.org
There
are
a
couple
other
projects
I’ll
men;on
that
are
focused
on
func;onal
monitoring:

Sensu
has
a
fairly
similar
scope
and
feature
set
to
Nagios,
but
takes
the
form
of
a
composable
framework
of
components,
including
a
web
interface
that
can
serve
as
a
front-‐end
for
several
Sensu
servers.

Like
Shinken,
it
has
a
lot
of
external
dependencies,

and
ge_ng
it
up
and
running
is
no
joke,
but
if
you
need
extreme
flexibility
or
have
very
specific
architectural
requirements,
it
might
be
for
you.

At
VictorOps
we’re
seeing
a
growing
number
of
our
customers
using
Sensu

Monit
•Runs
locally
on
each
server

•Thresholds
can
trigger
scripted
actions
or
alerts

•Lightweight
web
interface
shows
status
and
stats

•M/Monit
is
a
commercial
product
that
offers
a

central
dashboard
Monit
runs
as
an
agent
on
each
monitored
host.

It
can
be
conﬁgured
to
watch
processes,
ﬁles,
or
other
metrics,
and
take
scripted
ac;ons
when
thresholds
get
crossed.

Monit
provides
a
lightweight
web
interface
for
looking
at
status
and
sta;s;cs.

A

commercial
product
m/monit,
can
provide
a
central
management
point
for
several
hosts
running
monit,
but
it’s
not
a
required
component.

In
a
managed
environment
you
might
have
it
send
events
to
a
Nagios
host
and
in
that
way
decrease
poller
load,
or

avoid
some
of
the
security
concerns
with
running
NRPE.

Managed
vs.
Discovery
•Managed
configs
work
great
in
an
infrastructure-‐as-‐
code
environment

•Also
fine
for
small
environments
that
don’t
change

•Auto-‐discovery
solves
problems
for
some
larger

shops
with
heterogenous
networks

•Network
tripwire
All
of
those
solu;ons
are
great
if
you
are
in
a
DevOps
workflow,
and
you
have
a
managed
infrastructure,
using
something
like
Puppet
or
Chef.

The
hosts
and
resources
that
get
monitored
are
either
defined
in
well-‐known
standard
text
configura;on
files,
or

can
be
managed
programma;cally.

So
it’s
possible
to
wire
them
into
your
systems
management
framework
and
get
your
configura;on
auto-‐generated.

Of
course,
not
everybody
lives
in
that
world.

Auto
Discovery
is
an
important
feature
for
a
lot
of
shops.

If
you
work
for
a
larger
company
that’s
making
acquisi;ons,
you
might
find
yourself
managing
an
inherited
infrastructure
with
li[le
prior
knowledge
and

poor
documenta;on.

In
this
case,
a
tool
that
can
discover
the
current
state
of
a
network
and
alert
you
to
changes
from
that
state
can
be
an
invaluable
learning
and
troubleshoo;ng
tool
while
you’re
building
documenta;on.

Auto-‐discovery
can
also
act
as
a
network
tripwire,
detec;ng
devices
that
may
have
been
placed
on
your
network
without
your
knowledge.

Most
monitoring
systems
with
auto-‐discovery
can
send
a
no;fica;on
when
a
new
device
is
detected.

OpenNMS
•Released
in
1999

•JVM-‐based

•Efficient
ping-‐based
auto-‐discovery

•Native
support
for
SNMP
OpenNMS
is
an
example
of
a
discovery-‐based
monitoring
solu;on
that’s
been
around
almost
as
long
as
Nagios.

They’re
here
at
the
conference
so
I’m
sure
many
of
you
have
spoken
to
them.

It’s
JVM-‐based
so
it’s
pre[y
easy
to
get
it
deployed
and
running.

In
my
experience
it
runs
fine
with
OpenJDK.

Once
you’ve
tuned
the
discovery
se_ngs
it
can
pick
up
the
state
of
a
network
quickly
and
provide
a
picture
of
where
everything
is
running.

One
thing
to
note
with
OpenNMS,
and
a
lot
of
discovery-‐based
tools,
is

that
the
hosts
being
discovered
have
to
respond
to
ICMP
to
be
discovered.

So
it’s
less
effec;ve
for
finding
“rogue”
equipment
on
the
network,
and
switch
or
router
ACLs
may
limit
the
effec;veness
of
discovery
in
distributed
networks.

OpenNMS
uses
SNMP
to
gather
metrics
and
do
local
checks
on
monitored
systems.

This
makes
it
especially
handy
if
you’re
managing
a
lot
of
network
gear
or
appliance-‐type
devices,
like
video
conferencing
systems
and
so
on.

The
Assimilation
Project
•“Nanoprobe”
agents
do
discovery
and
monitoring

•Zero
network
footprint
discovery
listens
to
ARP
traffic,

so
nodes
don’t
need
to
respond
to
ping
(or
anything)
to

be
discovered

•Agents
form
a
self-‐organizing
mesh
of
redundant

monitors

•http://assimilationsystems.com
There’s
a
brand-‐new
project
in
the
early
stages
of
development
called
“The
Assimila;on
Project”.

It
was
started
by
Alan
Robertson,
who
you
may
know
as
the
founder
of
the
Linux
HA
project
and
long
;me
maintainer
of
Heartbeat.

It
uses
“nanoprobes”,
or

basically
li[le
agents,
that
communicate
state
to
a
central
data
store,
and
these
nanoprobes
do
distributed
“zero
network
footprint”
discovery
by
listening
to
arp
traﬃc.

This
means
that
it
can
quickly
discover
your
network
topology
even
if
your
hosts
don’t

respond
to
pings,
and
since
the
process
is
distributed
and
agent-‐based,
it
can
discover
complex
and
heterogenous
networks
(as
long
as
the
agents
can
connect
back
to
the
master).

Monitoring
is
also
highly
distributed
and
redundant,
so
the
loss
of
any
one

node
will
not
result
in
the
loss
of
monitoring
for
any
other
node.

In
a
lot
of
ways
it
represents
a
new
paradigm
in
systems
monitoring,
and
I’m
probably
not
doing
it
jus;ce
here.

I
really
encourage
you
to
check
out
the
project
website,
there
are
some
good
illustra;ons
of
the
core
concepts
at
play,
that
give
you
an
inkling
of

the
poten;al
power
of
this
so]ware.

If
total
informa;on
awareness
is
your
goal,
or
you
just
like
messing
around
with
bleeding-‐edge
technology,
Assimila;on
should
be
on
your
list.

Zenoss
•Similar
scope
to
OpenNMS

•Zenoss
sells
an
enterprise
product,
but
the
free

product
is
community
supported

•“Zen
Pack”
plugins
provide
specialized
monitoring

and
alerting
capabilities
Zenoss
is
another
tool
that
does
discovery.

Like
OpenNMS,
it
also
does
;me-‐series
monitoring,
and
also
has
event
handling
capabili;es.

Hopefully
you
had
a
chance
to
stop
and
talk
to
them
at
their
booth
during
the
show;
I
get
the
impression
that
their

community
outreach
team
takes
their
jobs
very
seriously.

So
even
though
their
main
focus
is
a
commercial
enterprise
product,
the
community
version
is
OSS,
and
it’s
viable.

Zabbix
•Similar
scope
to
OpenNMS

•Includes
an
agent
to
perform
local
checks
and
report

system
metrics

•Zabbix
is
a
commercial
company,
but
their
software

is
all
OSS
Zabbix
is
another
product
that
does
network
discovery,
similar
to
OpenNMS.

Zabbix
also
includes
an
agent
that
can
be
installed
on
monitored
hosts
and
provides
system-‐level
metrics
(again,
OpenNMS
achieves
this
through
SNMP).

Zabbix
is
distributed
by
a

commercial
en;ty
who
oﬀer
support
and
consul;ng,
but
the
so]ware
is
all
distributed
under
a
GPL
license.

Time-‐Series
Monitoring
All
right,
now
I
want
to
highlight
some
projects
that
are
focused
on
;me-‐series
data.

The
early
Players
•MRTG

•RRDtool
This
category
really
started
with
MRTG,
and
later
rrdtool.

They
produced
useful
graphs,
but
were
diﬃcult
to
manage,
and
didn’t
scale
well.

Cacti
•RRDtool
with
a

nice
web-‐based

management

wrapper
and

dashboard

builder
Cac;
came
along
in
the
early
2000’s
as
a
web
conﬁgura;on
wrapper
around
RRDtool.

It
solved
a
lot
of
the
ease
of
use
problems,
but
didn’t
address
the
scalability
of
the
RRDtool
backend,
or
the
inherent
scalability
challenges
of
an
SNMP
poller-‐based

graphing
tool.

Munin
•Local
agents
with

a
central
poller

•Goes
beyond

SNMP
monitoring

•Can
produce

alerts
based
on

thresholds
Munin
breaks
through
some
of
the
poller-‐based
scale
challenges
of
an
RRDtool-‐based
solu;on,
by
using
an
agent
that
gathers
data
locally
and
transmits
it
in
batches.

The
master
can
also
pass
alerts
to
Nagios
if
the
agents
report
threshold
viola;ons.

Munin

also
breaks
out
of
SNMP-‐based
checks
by
making
process
and
ﬁlesystem
checks
simple
in
the
agent.

Out
of
the
box,
the
agent
collects
a
ton
of
metrics,
so
it’s
a
good
way
to
get
a
comprehensive
look
at
a
system.

On
the
downsite,
it’s
not
known
for
great

performance,
and
the
agent
may
be
a
bit
heavy
for
some
people’s
tastes.

Graphite
•Whisper
storage
engine
is
similar
to
RRDtool,
but

can
be
clustered

•Modular
framework
includes
listeners,
pollers,
and
a

visualization
layer

•New
data
points
automatically
created
in
the

database
More
recently,
new
projects
have
taken
a
more
modularized
approach,
which
has
allowed
for
more
mixing
between
visualiza;on
layers,
storage
layers,
and
data
sources.

The
storage
back-‐ends
have
become
very
free
form,
allowing
for
the
crea;on
of

arbitrary
objects
and
key-‐value
pairs.

Graphite
is
a
popular
example.

It’s
a
framework
that
includes
a
storage
engine
similar
in
many
ways
to
RRDtool,
a
daemon
that
listens
for
;me
series
data,
and
a
web-‐based
visualiza;on
layer.

It’s
notably
diﬀerent
than
Cac;
and
Munin
because
it
doesn’t

include
a
poller,
it
listens
for
connec;ons
and
passively
accepts
data,
automa;cally
crea;ng
new
tables
when
new
data
sources
connect.

So
provisioning
new
hosts
is
eﬀortless.

Lots
of
compa;ble
tools
have
been
created
to
feed
it
data,
including
SNMP

pollers,
so
more
tradi;onal
setups
and
data
sources
can
also
be
accommodated.

Graphite
-‐
Tree
View
Here’s

a
look
at
the
graph
composer,
which
allows
you
to
browse
through
all
of
your
data
points,
combine
them
in
a
single
graph,
and
apply
ﬁltering,
aggrega;on,
transforma;ons
and
so
forth
to
the
data.

Very
powerful.

Graphite
-‐
Dashboard
And
here’s
an
example
of
a
graphite
dashboard.

It’s
super
easy
to
compose
dashboards
of
several
graphs
though
the
web
interface.

InﬂuxDB
•Similar
scope
to
Graphite

•Designed
for
horizontal
scale

•Easy
web
API
for
the
listener

•SQL-‐like
query
language
for
building
graphs

•No
external
dependencies
InﬂuxDB
is
a
brand-‐new
project
with
similar
scope
to
Graphite.

It’s
designed
to
be
easily
horizontally
scalable
and
its
listener
presents
a
standard
h[p
API,
making
it
easier
to
write
data.

It
also
has
a
SQL-‐like
query
language
with
func;ons
you
can
use
to

build
complex
graphs
or
new
data
sources.

It’s
simple
to
install
with
no
external
dependencies.

The
project
isn’t
calling
itself
produc;on-‐ready
yet,
but
it
should
be
soon,
and
it’s
worth
keeping
an
eye
on.

Prometheus
•Just
open-‐sourced
by

SoundCloud

•Similar
to
Graphite,
but

adds
alerting

functionality

•Includes
a
data
poller,

but
can
passively

receive
data
too
Image
Source:
SoundCloud
Prometheus
is
an
internal
project
at
SoundCloud
that
they
open-‐sourced
in
January.

It
is
another
;me-‐series
database
similar
to
InﬂuxDB,
but
that
adds
aler;ng
func;onality.

It
includes
its
own
poller,
but
can
also
passively
receive
data
like
Graphite
and
InﬂuxDB.

No;ce,
this
new
genera;on
of
tools
is
all
highly
modular
and
composable.

To
some
extent
you
can
mix
the
front-‐ends
and
back-‐ends,
and
some
collec;on
tools
can
be
used
with
just
about
all
of
them:

Collectd
•Gathers
metrics
and
emits
them
to
a
configurable

destination

•Runs
as
an
agent
on
a
variety
of
architectures

•Easy
to
manage
configuration
for
example,
collectd
runs
as
a
lightweight
daemon
on
monitored
hosts,
gathering
metrics
based
on
a
built-‐in
plugin
library
and
emi_ng
them
at
regular
intervals
to
a
des;na;on.

It’s
a
great
way
to
feed
system
sta;s;cs
like
CPU
u;liza;on
and
disk
IO
to
a

;me-‐series
database,
and
can
be
managed
easily
by
Puppet
or
Chef.

Statsd
•Middleware
to
make
it
easy
to
get
data
into
Graphite

•TCP
and
UDP
listener
receives
metrics,
forwards

them
to
the
Graphite
listener

•Open-‐source
project
from
Etsy,
inspired
by
a
similar

project
from
Flickr
statsd
is
another
example.

It
provides
a
TCP
and
UPD
listener
that
can
receive
metrics
in
a
simple
format
and
forward
them
on
to
a
backend
like
Graphite
or
InﬂuxDB.

It’s
an
easy
way
to
add
instrumenta;on
to
in-‐house
so]ware.

Etsy
released
it,
inspired
by

a
similar
project
from
Flickr.

Grafana
•Visualization
interface

•Dashboard
composer

•Can
use
multiple

backends
simultaneously

•Mix
and
match
graphs

from
different
tools
on
a

single
dashboard
Grafana
provides
a
visualiza;on
interface
and
dashboard
composer
that
can
be
used
with
mul;ple
backends,
including
graphite
and
InfluxDB.

In
can
embed
annota;ons
into
;me-‐series
graphs,
and
can
build
dashboards
that
include
data
from
different
back-‐
end
data
stores.

As
you
can
see,
there’s
a
lot
of
ac;vity
happening
in
this
category,
and
these
tools
work
great
in
elas;c
environments
and
DevOps
workflows
because
the
storage
engines
are
so
free-‐form.

I
expect
a
lot
of
changes
and
a
bit
of
consolida;on
in
this
space
over

the
next
few
years,
but
it’s
so
easy
to
get
value
out
of
these
projects
quickly
that
it’s
worth
doing
some
experimen;ng
and
trying
some
new
things
out.

Sidebar:

Poller
vs.
Listener
•Polling
over
a
network
is
inefficient

•TCP
is
more
expensive
to
the
originator

•Poller
connections
are
longer-‐lived

•Listeners
can
be
used
for
functional
monitoring

•Scale
demands
it,
we’re
going
to
see
more
of
it
You
may
have
no;ced
that
a
lot
of
the
more
recent
tools
have
switched
from
a
poller
model
to
a
listener
model.

This
makes
so
much
sense
with
;me-‐series
data,
but
I
also
think
the
model
has
applica;on
in
func;onal
monitoring.

Poller-‐based
monitoring
is

just
inherently
ineﬃcient,
since
so
much
of
the
TCP
heavy
li]ing
is
concentrated
in
one
place,
and
frequently
the
connec;on
has
to
be
held
open
while
the
check
is
run
or
the
metric
sampled.

Emi_ng
events
to
a
passive
listener
means
more
of
the
TCP
work

is
distributed,
and
the
check
or
sample
can
run
before
the
connec;on
is
opened,
so
you
only
need
to
communicate
long
enough
to
send
the
data.

In
a
managed
environment,
a
lot
of
health
checks
can
be
run
locally
and
problems
emi[ed
to
Nagios
or

another
event
handler.

This
scales
be[er
than
poller-‐based
solu;ons
in
so
many
environment,
I
have
to
expect
we’ll
start
seeing
more
of
it
on
the
func;onal
side.

The
Assimila;on
project
has
a
fascina;ng
approach
to
this
as
well;
again,
I
urge
you
to
check
it

out
on
their
website.

Event
Monitoring
Okay,
on
to
our
third
category,
Event
monitoring

The
Bad
Old
Days
•Logwatch

•Logmon

•Php-‐syslog-‐ng
Event
monitoring
and
visualiza;on
is
a
fairly
undeveloped
area
by
contrast,
but
one
that’s
star;ng
to
heat
up.

For
a
long
;me,
if
you
had
limited
resources
and
limited
money,
you
had
to
rely
on
hourly
or
daily
runs
of
logwatch
to
look
for
errors
in
system

logs,
or
run
programs
like
“logmon”
locally,
basically
watching
a
file
with
a
tail
-‐f
and
emailing
alerts.

Resource
intensive
and
clunky.

Splunk
came
along
in
the
early
2000’s
offering
an
a[rac;ve
centralized
log
analysis
tool,
but
with
a
steep
price
that
put
it

out
of
reach
for
a
lot
of
small
shops
that
produce
a
lot
of
log
data.

Other
open
source
log
management
tools
have
been
produced,
like
php-‐syslog-‐ng,
but
rela;onal
data
stores
like
MySQL
struggle
with
log
data
and
the
results
offered
poor
search
performance
and
poor
scalability.

Php-‐syslog-‐ng
has
been
rewri[en
as

“Logzilla”,
by
the
way,
and
is
now
an
exclusively
commercial
product.

ELK
-‐
Logstash
•The
“ELK”
stack
is
comprised
of

Elasticsearch,
Logstash,
and
Kibana

•Clusterable,
horizontally
scalable
data

store

•Lots
of
listeners,
filters,
event
handlers

•Beautiful
easy
to
use
visualization

layer
Recent
advances
in
open-‐source
data
stores
like
Elas;csearch
has
given
rise
to
Logstash,
which
provides
very
similar
func;onality
to
Splunk
with
a
completely
open-‐source
stack.

This
is
commonly
referred
to
as
the
“ELK”
stack,
comprised
of
Elas;csearch
as

the
back-‐end
data
store,
Logstash
to
receive
and
process
log
events,
and
then
pass
them
to
the
storage
backend,
and
Kibana,
the
web-‐based
visualiza;on
layer
that
makes
it
easy
to
create
dashboards
of
your
log
data
including
histograms
and
pie
charts.

As

of
2013,
Elas;csearch
is
the
commercial
keeper
and
distributer
of
the
ELK
stack
components.

Logstash
-‐
Kibana
Image
source:
ifconﬁg-‐a.com
Here’s
a
screenshot
of
the
Kibana
part
of
the
ELK
stack.

As
you
can
see,
it
provides
not
only
ways
to
examine,
search
and
ﬁlter
your
logs,
but
also
to
visualize
them
in
interes;ng
ways.

Nagios
Log
Server
•Nagios-‐branded
ELK
stack

•Includes
user
management
and
security
features

•Promises
easy
clustering

•No
free
version
Last
fall,
Nagios
annouced
the
Nagios
Log
Server,
which
is
a
commercialized
version
of
the
ELK
stack
with
mul;-‐user
security
features
and
integra;on
with
Nagios
built-‐in,
and
promises
easy-‐to-‐deploy
clustering,
something
that
can
be
challenging
with
a

standard
Elas;csearch
deployment.

There’s
no
free
version,
but
it
might
be
a
way
to
get
ELK
in
the
door
if
you
have
a
manager
that
frets
about
so]ware
without
support,
or
you
simply
don’t
have
;me
to
invest
in
ge_ng
an
elas;csearch
cluster
running.

Graylog
2
•Similar
scope
to
ELK
stack

•Commercial
company
releases
software
as
OSS,
and

sells
support
and
consulting

•External
dependencies
on
MongoDB
and

Elasticsearch
graylog2
is
a
product
very
similar
to
the
ELK
stack,
and
it’s
ge_ng
close
to
a
1.0
release.

Similar
to
Elas;csearch,
Graylog
Inc.
releases
all
of
their
code
as
open-‐source,
and
oﬀers
produc;on
and
development
support
contracts
for
enterprise
users.

It
has

external
dependencies
on
MongoDB
and
Elas;csearch.

Event
Monitoring
•A
lot
of
room
for
growth

•A
lot
of
room
for
consolidation
with
time-‐series

projects

•It’s
still
a
great
time
to
be
alive
and
generating
log

data
I
expect
to
see
the
event
management
tools
space
heat
up
like
the
;me-‐series
space
has,
even
as
the
two
categories
start
to
see
more
overlap.

Again,
I
think
it
will
be
a
while
before
a
clear
overall
winner
emerges
in
terms
of
storage
backends,
visualiza;on

layers,
and
event
handlers,
but
what’s
out
there
today
has
plenty
of
value
to
oﬀer.

If
you’re
not
running
Logstash
or
graylog2
today
you
owe
it
to
yourself
to
get
it
going.

Chances
are
you
won’t
have
to
wait
long
to
see
a
beneﬁt.

Our
Categories
Again
Func;onal
Time

Series
Event
Okay,
that’s
a
very
quick
survey
of
our
three
major
categories.

If
I
le]
your
favorite
project
out,
let
me
know,
I’ll
probably
want
to
hear
about
it!

Now,
as
I
said
at
the
start,
there’s
a
lot
of
overlap
between
the
func;onality
of
the
tools
in
these
three

categories,
and
some
of
the
tools
I
men;oned
could
ﬁt
in
mul;ple
categories.

It
can
be
temp;ng
to
look
for
a
single
solu;on
that
captures
all
of
this
func;onality,
and
several
projects
and
companies
are
aiming
at
a
uniﬁed
solu;on.

Do-‐It-‐All
Products
•OpenNMS

•Zenoss

•Zabbix

•Nedi

•Nagios
Enterprises

OpenNMS,
Zabbix,
and
Zenoss
don't
just
do
discovery
and
monitoring,
they
also
graph
;me-‐series
data,
and
have
asset
management
and
network
management
func;onality
built-‐in.

“Nedi”
is
a
similar
project
that’s
s;ll
defining
it’s
commercial
and
release
model,
but
for
now
is
mainly
defined
by
a
long
list
of
external
dependencies.

Nagios
Enterprises
is
in
the
process
of
building
a
suite
of
commercial
products
around
Nagios
that’s
comprised
of
a
lot
of
different
components
(like
Nagios
Log
server),
but
at
least
gives
you
a
single
neck
to
choke
for
support
I
guess

Monolithic
vs.
Modular
Monolithic:

•If
all-‐in-‐one
gets
the
job
done,
then
great

•Good
for
smaller
scale,
non-‐tech-‐focused
companies

Modular:

• Agile
and
DevOps
shops
require
flexibility
and

innovation

• Good
for
tech
driven
and
ops-‐focused
companies
Some
of
these
all-‐in-‐one
products
and
suites
of
products
can
be
very
a[rac;ve,
and
if
you
find
one
that
meets
all
of
your
needs,
then
it
can
save
you
a
lot
of
;me
you
might
otherwise
spend
composing
a
monitoring
infrastructure
out
of
individual
parts.

For

a
lot
of
shops
that
operate
at
a
smaller
scale
but
otherwise
have
more
enterprise-‐like
requirements
(i.e.
the
primary
business
isn’t
technology),
it
may
be
a
wise
approach
to
look
for
a
single
product
to
do
everything,
and
adapt
your
workflow
to
the
tool
if

you
need
to.

For
shops
adop;ng
DevOps
methodologies,
and
managing
infrastructure
as
code,
especially
those
working
in
elas;c
environments
where
architectures
and
requirements
can
change
over
;me,
there’s
tremendous
opportunity
to
combine
the
latest

genera;on
tools
in
ways
that
were
difficult
or
impossible
just
a
few
years
ago,
and
produce
business
value
quickly.

If
you
find
that
one
of
the
monolithic
product
works
for
you
in
that
sort
of
environment,
then
great,
but
it’s
probable
that
a
mixture
of
smaller

components
will
address
your
specific
needs
more
easily.

The
“Event
Stream”
Concept
Source:
@obfuscurity
For
a
few
years
now,
Jason
Dixon,
who
organizes
the
“Monitorama”
conference
which
is
going
to
be
in
Portland
this
year,
has
described
the
“event
stream”
paradigm
in
monitoring.

Rather
than
try
to
reproduce
his
diagram
I’ve
just
swiped
it
from
one
of
his

slide
decks.

It
goes
like
this:
events
enter
the
toolchain.

Depending
on
event
type
and
content,
it
might
get
sent
to
a
storage
backend
for
visualiza;on,
or
aggregated
with
other
events
before
ge_ng
stored.

Depending
on
content,
it
might
change
state
in
a

state
engine,
trigger
an
event
handler
or
aler;ng
process,
ﬁre
oﬀ
an
email
or
a
VictorOps
alert.

By
using
small,
composable
tools
in
combina;on,
you
create
the
event
stream
that
suits
your
technology
stack
and
environment.

Nagios-‐Based
Toolchain
Collectd
NRPE
Monit
Nagios
ELK
Graphite
Syslog
This
doesn’t
mean
necessarily
turning
your
back
on
ins;tu;onal
knowledge
and
established
tools.

Nagios
can
sit
comfortably
right
in
the
middle
of
the
event
stream.

Icinga
2-‐Based
Toolchain
Collectd
NRPE
Monit
Icinga2
ELK
Graphite
Syslog

Icinga
2
was
designed
with
modularity
in
mind,
and
has
a
built-‐in
support
for
sending
metric
data
to
Graphite.

You
can
do
this
with
Nagios
as
well,
it’s
just
not
quite
as
easy.

Sensu,
Shinken,
Monit,
Collectd,
they’re
all
poten;al
components
in
a
modular

monitoring
toolchain.

You
can
pick
the
pieces
that
best
meet
your
par;cular
needs,
and
combine
them
in
a
way
that
they
lend
value
to
each
other
rather
than
crea;ng
islands
of
data.

The
current
genera;on
of
open-‐source
monitoring
tools
is
giving
us
easier
access
to
deep
insight,
be[er
visibility
into
faults,
and
amazing
interoperability.

These
solu;ons
are
quickly
becoming
cri;cal
sources
of
not
just
opera;onal
and
developmental

intelligence,
but
of
business
intelligence
that
has
value
outside
of
technical
departments.

This
helps
Dev,
DevOps
and
IT
teams
more
eﬀec;vely
communicate
their
overall
value,
and
makes
life
a
lot
easer
managing
high-‐demand
24/7
plauorms.

The

technology
is
moving
way
too
fast
for
“Enterprise”
solu;ons
to
keep
pace,
meaning
that
as
a
user
of
open-‐source
monitoring
solu;ons,
you
don’t
have
to
wonder
what
you’re
missing
out
on
from
commercial
products.

Open-‐source
is
now
at
the
vanguard
of

monitoring
and
telemetry
so]ware
development,
and
it’s
a
great
;me
to
be
involved.

Questions?
mike@victorops.com

@vo_mike
If
you
have
any
ques;ons,
I’d
love
to
answer
them
now.

The Open-Source Monitoring Landscape

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to The Open-Source Monitoring Landscape (20)

Recently uploaded (20)

The Open-Source Monitoring Landscape