►
From YouTube: Performance Improvements in Etcd 3.5 Release
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
A
The
second
part
is
about
inefficient
warning
log,
and
after
that
we
will
cover
the
performance
change
in
3.5
and
then,
after
that,
we
will
mention
the
new
benchmark
command
that
we
added
to
activity
repository
and
finally,
we
will
talk
about
the
ongoing
and
future
work
that
we
are
planning
to
do
in
the
future.
A
A
Its
best
fit
scenario
is
when
you
have
more
reads
than
writes
in
your
database
environment
here,
the
graph
on
the
right
side
shows
how
activity
cluster
looks
like
the
rough
particle
that
is
running
on
each
of
the
activity.
Cluster
node
behaves
like
a
state
machine,
a
changes
state
based
on
the
message
coming
from
other
peers.
A
A
A
Now,
let's
talk
about
some
of
the
important
activity,
apis,
actually,
clients
talks
to
activity
server
using
grpc.
There
are
five
important
unity.
Grpc
calls.
We
need
to
pay
attention
here,
the
first
one
input
which
basically
put
a
key
to
activity
server
database
and
the
second
one
is
range
which
can
be
used
to
read
one
or
more
keys
from
activity,
server,
database
and
so
on.
Delay
range,
which
can
be
used
to
delete
either
one
or
more
keys
from
activity
database.
A
A
A
A
A
A
A
The
active
d
rewrite
latency
can
get
significantly
reduced
by
around
90
percent.
However,
at
the
throughput
get
reduced
due
to
tx3
buffer
deep
copy,
tx
rebuffer
is
one
of
the
few.
In
current
re-transaction
data
structure,
it
holds
a
list
of
sorted
key
value
pairs
that
not
yet
committed
to
the
story.
Yet
we
observe
that
the
kubernetes
eps
server
throughput
in
activity
3.3,
is
actually
higher
than
3.4
because
of
this
tx3
buffer,
deep
copy.
A
A
A
A
A
A
A
A
A
A
The
first
solution
is
that
done
by
chao
from
aws,
and
here
is
a
link
to
his
pull
request.
A
A
A
A
A
A
A
A
A
A
Is
there
any
way
we
can
improve
the
performance
here
we
are
actually
working
on
that.
The
second
is
marshall
and
unmarshall.
Clients
send
messages
to
server,
and
the
server
will
do
a
lot
of
marshalling
and
arm
marshaling.
On
these
messages,
however,
many
times
we're
actually
doing
some
actual
marshalling
and
unmarking
operations,
these
are
not
needed,
so
we
probably
can't
provide
the
cash
so
that
so
that
we
don't
have
to
do
the
marshall
and
our
marginal
operations.