Slide 9
Slide 9 text
Within
any
system,
the
more
moving
parts
there
are,
the
harder
it
is
to
understand
what’s
going
on.
Everything
is
highly
correlated
with
everything
else.
And
in
the
case
of
mongo
one
query
affects
the
performance
of
another.
I
think
the
addiKon
of
a
new
query
pa]ern
to
mongo
actually
has
a
nonlinear
detrimental
effect
on
a
human’s
ability
to
understand
what
the
hell
is
going
on.
low
cost/high
frequency
queries
will
execute
on
data
that’s
fully
paged
into
memory,
but
if
you’re
also
high
cost
queries
that
execute
less
frequently,
that
can
cause
faults
and
that
could
cause
the
next
batch
of
high
rate
queries
to
fault
because
the
data
was
flushed
out
of
memory.
Or
even
worse,
if
the
low
frequency
operaKon
is
a
write
it
can
cause
those
high
frequency
queries
to
queue
up.
The
more
collecKons
and
the
more
query
pa]erns
you
have
on
a
single
cluster,
the
harder
it
is
to
figure
out
exactly
what’s
going
on
when
there
is
a
problem.
If
you
look
in
the
profling
logs
someKmes
it’s
unclear
if
things
are
there
because
they
are
actually
slow,
or
general
compeKKon
for
resources
made
them
slow.
It
can
be
9