►
From YouTube: The Architecture of Microservice Transaction Based on Distributed Consensus with... Jumpei Nishitani
Description
Don't miss out! Join us at our upcoming event: KubeCon + CloudNativeCon Europe in Amsterdam, The Netherlands from 18 - 21 April, 2023. Learn more at https://kubecon.io The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.
The Architecture of Microservice Transaction Based on Distributed Consensus with Kubernetes - Jumpei Nishitani, Hitachi, Ltd.
When constructing an enterprise system by Microservice Architecture, the engineers faced various problems. We think the most difficult problem is how to update data in the Database consistently. Most engineers will choose a Microservices design pattern called “DB per Service” when we construct microservices with a Database. The engineers must consider Transaction Management. If you construct Transaction Management like the “Saga Design pattern”, your application will be complex, and you must work a lot to maintain data consistency in system runtime as an operator. That’s why Saga does not have sufficient functionality to construct a transactional system that supplies consistency and scalability. To resolve this problem, Microservice Transaction Management System that balanced strong consistency, loosely coupled, and high productivity is necessary. We will introduce the concept of Microservice Transaction management based on distributed consensus.
A
My
name
is
John
penistein
I'm
working
for
the
Hitachi
in
Japan,
and
this
is
my
first
time
expressed
and
joined,
International
Conference
in
English,
so
I'm,
better
Navas,
but
I'll
do
my
best.
In
my
experience,
I
have
a
developed,
a
web
server
and
a
patch
and
your
applications.
How
about
like
a
glass
fridge
for
10
years
and
after
that,
I
could
do
the
modern
Edition
plan
for
our
customer.
For
example,
infrastruct
change
like
the
on-premise
system
to
the
public
cloud
and
now
I'm
researching
the
developing
microservice
transaction
system.
A
Today
I
will
be
present
consensus
and
transaction
coming
stock
called
paxos
committee.
This
paper
was
written
by
two
famous
people
and
first
is
a
doctor.
Jim
Gray,
who
is
famous
for
the
relational
database
technology
and
second,
is
the
Dr
Leslie
Lambert,
who
is
a
famous
for
the
distributed
constant
sexual
algorithm
Texas.
A
A
What
is
the
Microsoft
excuse?
This
is
a
way
to
developers
system
for
the
volume,
several
small
application
and,
according
to
the
business
responsibility
boundary.
This
is
what
I
call
this
easy
understanding,
service
and
Engineers
can
improve
system,
positive
story
and
engine.
The
engineers
can
choose
the
best
technology
for
each
subsides,
but
sometimes
engines
may
have
first
difficulty
in
architecture,
microservices,
airport
security
and
debt
management.
A
Engineers
Engineers
would
like
to
implement
Microsoft
in
their
enterprise
system
because
of
the
needs
of
the
system,
majority
and
business.
That
agility
may
have
been
necessary
in
some
cases
to
increase
overall
system
majorities,
so
split
business
application
and
that
in
the
database
according
to
the
responsibility
boundary
and
make
it
the
service
the
dead
base.
First
Hobbies
button
is
a
famous
way
to
split
the
dead
in
Microsoft
seats.
As
a
result,
database
began
to
be
distributed
and
updating
their
turn
into
this.
Video
DB
increase
the
risk
of
the
inconsistencies
and
what
is
the
inconsistencies?
A
I
describe
the
examples
of
the
inconsistencies
when
the
transfer
service
transfers
the
money
from
the
bank
here
to
bank
B
and
Bank
B,
which
were
the
money
by
the
bank
account
and
Bank
B
deposit.
The
same
amount
of
the
bank
B
account.
In
this
scenario.
A
bank
ba
cannot
deposit
the
money
for
some
reason,
like
a
failure
and
inconsistency
happened
between
the
Banquets
amount
and
the
bank
pays
amount.
A
A
A
What
is
a
complex
distributed
transaction,
the
Microsoft's
textbook
Microsoft
Spartan
with
example,
name
jabber
include
the
benefit
of
the
microservice
and
the
challenges
of
them.
Applying
the
global
transaction
management
to
the
microservices
and
mentions
JB
must
be
X
compliant
and
cannot
show
the
modern
DB
technology
in
other
nosql
and
second
technology
stacker
derivative
to
the
famous
framework
that
include
the
transaction
monitor
such
as
the
Java,
ee
or
spring,
and
third,
consistency
must
be
given
out
to
some
extent
in
the
distributed
system
from
cap.
A
Saga
design
protein
is
a
most
famous
approach
in
the
microservice
architectures
Saga
Link's
business
process
together
ninja
work
workflow
as
the
process
fails,
a
compensation
transaction
is
called
to
cancel
the
update
and
ensures
atomicity,
but
please
look
at
the
call
sample
of
the
business
application,
which
is
implementation
of
the
transaction
management
and
routing
the
business
world.
As
a
result,
the
business
application
becomes
complex.
A
Why
is
it
complex?
Why
is
it
complicated
dented?
Because
three
interesting
cells
inconsistencies
fractal
and
distributed
transactions?
They're
very
complex
event?
There
are
three
main
codets.
First
business
error.
Second
system
failure
last
right
right,
conflict,
the
first
step
of
the
complication.
Here's
the
update
is
prevented
by
the
business
partition
in
the
DB.
It
will
prevent
this
update,
for
example,
with
the
row
more
than
the
Maximus
balance
or
out
of
balance.
It
will
be
necessary
to
count
the
service
that
have
already
been
updated
to
achieve
this
opinion.
A
The
second
step
of
the
complication,
for
example.
When
process
goes
down,
you
are
in
the
transaction
execution,
the
transaction
state
will
be
lost
and
the
inconsistency
will
happen
to
prevent
the
inconsistency
by
System.
Failure
like
this,
a
constant
mechanism
is
needed,
but
failure
handling
is
also
challenge
where
it
can
knocker
at
the
various
times
and
how
to
set
the
transaction
depends
on
the
case
of
the
error.
A
Concurrency
control
is
needed,
the
lock
is
the
type
of
the
concurrency
control
and
the
lock
information
itself
becomes
a
state.
The
log
information
must
be
the
persistent
after
that,
the
call
to
deal
with
the
concurrency
control
will
be
exported
to
the
application
logic
play
it.
Look
at
the
picture
below
what
we
originally
wanted
to
do
was
simply
to
have
the
two
service
update
when
distributed
transactions
are
implemented.
A
There's
three
factors
complicate
the
application
and
Industrial
structures.
This
is
the
root
cause
of
the
complication
of
the
transaction
management.
The
Saga
itself
has
only
business
error
capability
and
cannot
deal
with
the
failure
and
oh
right
right.
Conflict
in
particular,
The
Saga
is
serious
reflowed
when
it
comes
to
the
right
red
conflict.
Therefore,
no
matter
how
hard
we
try,
we
cannot
prevent
the
inconsistencies.
A
There
are
already
asterism
mechanisms
in
the
Academia
to
prevent
the
right
right
conflict
in
the
distributed
transaction.
That
is
a
principle
of
commitment
ordering
paper
from
1992..
This
paper
say:
that's.
The
only
mechan
is
to
prevent
the
right
right
conflict
and
the
distributed
transactions
or
two
phase,
locking
or
avoiding
cascading
abort
and
Server
doesn't
have
this
technology
today.
I
don't
mention
the
detail,
but-
and
this
is
this
academic
opening.
A
However
many
microservices
Engineers,
don't
know
the
fact
why
we
think
that
academic
affiliation
is
fragmented,
microservices
boundaries
to
like
the
quest
Richardson
prefer
to
cite
a
semantic
acid
and
the
cycle
paper,
but
felon
and
right
right
conference
solutions
are
inadequate
in
these
papers.
On
the
other
hand,
there
is
a
faction
that
concerning
how
to
prevent
pharaoh
and
right
right
conflict
automatically
in
the
fraction.
There
are
a
commitment
or
running
paper
and
distributed
consensus
for
the
failure
paper
for
now
on
where
X
Prime.
What
happened
when
we
implement
this
theory
about
this
function?
A
One
of
the
technology
function
is
the
two
phrase
comment
which
was
called
the
end
pattern
of
the
Microsoft
seats.
Two
friends
commit
provide
concurrency
control
r
surface
committee
acquires
a
lock
during
the
DB
update
and
the
Lux
release
after
the
agreeing
atmosphere.
This
is
so-called
two-phase,
locking
two
friends
commit
aggregate
as
a
DB
update,
possibility
like
a
prepare,
okay
and
finalize
to
commit
when
all
other
prepare
okay,
they
say
the
so-called
Atomic
commitment.
A
The
two
phrase
comet
has
two
merits
for
the
engineers.
The
first
is
to
prevent
in
constant
inconsistency
by
the
right
right,
conflict
and
second,
the
exposure
of
the
transaction
controller
function
to
the
application
can
be
very
small.
At
least
one
line
like
a
JTA
Atma
transaction.
This
reduced
business
lodge
coupling.
A
We
have
tried
to
implement
a
separate
transaction
management
architecture
shown
here.
I
apply
my
prototype
automatic
transaction
annotation
as
a
library
to
the
transaction
application
on
caucus,
a
prototype
that
ddvc
driver
is
applied
to
the
bank
application
and
caucus
from
there.
The
resource
manager
processes
launched
as
assigned
Comfort
bank
kpb
and
transaction
managers
started
as
a
separated
port.
A
The
opportunity
API
after
the
API
call
the
transaction
added
first
in
January
to
the
jdbc
drive
and
in
the
bank
services
driver
internally
passed
the
transaction
ID
and
update
SQL
to
the
resource
manager.
The
resource
manager
process
updated
SQL
to
the
DB.
In
addition,
we've
already
update
the
SQL
is
a
lock
by
detect
for
an
update
foreign.
A
After
that
admin
commitments
to
start
the
DB
return
update
of
their
video
like
a
preview,
okay
when
usually
to
the
transaction
manager.
If
I
return
the
previous
okay,
that
transaction
manager
sends
a
commit
to
the
resource
manager.
If
everyone
prepare
and
this
font,
the
transaction
manager
sends
the
route
back
to
the
resource
manager.
At
the
same
time,
the
transaction
manager
passes
the
result
of
the
atomic
commitment
in
the
transaction
log.
A
A
That
is
a
single
point
of
Valor
problem
of
the
transaction
manager
in
particular
that
transaction
manager
process
is
down
in
the
atomic
commitment
phrase,
and
the
transaction
state
is
lost.
Dblock
is
not
released
and
spontaneous
rollback,
brexit
atomistic.
This
is
a
common
noun
of
theoristic
Hazard.
A
To
prevent
the
futuristic
Hazard,
it
is
necessary
to
deal
with
the
system
switch
operation
of
the
persistent
straight
called
primary
backup
high.
Above
it,
fragmentation
can
easily
occur
in
the
crowd
crowd
environment
if
fragmentation
makes
it
difficult
to
prevent
a
heuristic
hazard.
There
are
two
reasons
why
the
cloud
environment
and
primary
backups
are
incompatible.
A
First,
primary
backups
uses
of
fencing
requirement
equipment
to
switch
systems,
but
the
promise
for
the
primary
Buckle
is
a
Fencing.
Equipment
equipment
must
be
able
to
receive
the
failure
native
notification
immediately
in
an
environment
such
as
a
cloud
sometimes
of
our
notifications,
delay
and
a
system
does
not
switch
over.
A
Second,
the
single
coordinator
itself
becomes
a
single
point
of
failure,
like
a
primary
back
on
the
two
threads
coming
to
or
saga.
Therefore,
the
mechanism
to
overcome
the
single
coordinate,
the
single
point
of
failure
will
end
up
having
a
single
point
of
error
and
by
taking
a
counter
measure
against
it,
the
system
will
gradually
become
the
dependent
on
the
hardware.
The
crowd
vary
is
terrible,
structured
Hardware
from
the
software
stack,
but
if
you
increase
the
dependence
on
the
hardware,
there
is
no
point
in
the
building.
A
The
system
in
the
crowd
our
discussed
from
the
theoretical
aspect
of
computer
science.
Again
there
is
a
so-called
failure
model
in
Academia.
The
model
describes
the
cellularity
of
the
deal
with
the
I.T
system.
Failure,
Sega
and
two-phase
comic
can
only
operate
in
the
favor
of
free
environment.
Primary
backup
is
a
product
that
works
properly
in
synchronous
file.
A
Stop
that
is
an
environment
where
there's
no
communication,
delay
or
break
up
the
crowd
is
an
environment,
whereas
the
infrastructure
is
hidden
and
distributed
and
happened,
communication
device
sometimes
so
it
is
very
difficult
to
apply
second
to
phase
content,
primary
backup,
In
The
Crowd
by
action,
Theory
and
two
for
its
committed.
This
Theory
before
the
crowd
is
popular.
So
new
theory
is
required
to
construct
a
distributed
transaction
in
the
cloud
native
systems,
that
is
to
incorporate
the
consensus
in
the
failure
environment
Theory,
the
Dr,
Lampard
and
Dr
Lynch
have
worked
on.
A
Let
me
explain
what
paxus
is
taxes
says.
The
consensus
mechanism
is
using
a
Google
spanner
for
the
multiple
database
replication
and
blockchain
Ledger
replication.
What
is
the
consensus?
Consensus
mean
that
the
group
of
the
multiple
participant
obtaining
a
single
result-
this
is
a
bit
confusing,
so
I
will
use
an
example
of
the
parliament
system
of
the
Greek
as
a
Greek
island
of
Texas
in
the
parliament
democracy.
A
A
A
The
majority
rule
so-called
Quorum
mechanism
works
here.
Even
if
one
of
the
five,
let's
read
the
leaves
the
previous
World
result
in
the
later
boat
Riders
will
be
same.
In
other
words,
if
one
process
is
down,
but
the
majority
of
the
processes
are
still
alive,
consistency
can
be
assured.
This
is
nominal.
The
pigeon
Hall
principle
and
the
mathematical
Theory
it
is
very
difficult
to
explain.
Sorry
if
you
are
interested
in
this
Theory,
please
look
it
up.
A
So
where
should
this
practice
mechanism
be
mounted?
A
simple
solution
is
a
replica
simple
solution
is
to
replace
the
primary
backup
storage
area
with
no
SQL
implemented
and
graph
to
a
paxus
or
something,
for
example,
destroying
the
transaction
log
in
the
radius.
But
this
approach
is
not
efficient
in
the
term.
There's
a
number
of
communication
for
consistency
and
transaction
termination.
A
The
problem
with
the
heuristic
is
Aggregates
the
update
ability
and
the
resource
participating
the
distribute
transactions.
If
this
is
a
primary
backup,
it
will
be
delayed
or
not
switching
the
crowd.
That's
a
price
and
permanent
system
to
the
atomic
commitment
results.
Prepare
of
datability
makes
the
input
for
the
factors
protocol
the
Richer
process
to
the
prepare,
okin
and
voltage
token
After
exiting
the
practice
protocol
finalizer
coming
to
a
log
back
sensualizer.
They
have
a
failure,
tour
and
standard
availability.
A
A
Let's
look
at
the
examples
of
this
using
the
kubernetes,
the
countered
are
most
the
same
as
the
two
friends
coming
but
and
the
lower
left
corner
plays
the
place,
the
Redundant
coordinator
in
the
port,
where
prepare
it
as
a
port
for
the
banquet
and
about
for
a
pot
for
the
bank
B
here
we
will
first
update
the
bid.
This
update
process
is
not
really
different
from
the
two
Frets
commit
and
next
sender
prepare
from
the
transfer
service
to
the
site
called
or
the
bank
service.
A
At
that
time,
the
site
can
create
a
box
to
collect
the
contents
of
sporting
result
and
the
transaction
state
of
the
bank
subsides
next.
The
day
B
returns
a
prepare.
Okay.
At
this
point,
the
science
curve,
the
transaction
State
as
a
preview.
Okay,
after
that,
the
side
call
of
the
bank
broadcast
this
prepare
or
get
to
the
Redundant
coordinators.
A
One
of
the
Redundant
coordinates
that
go
down
if
the
majority
of
the
coordinators
are
still
alive.
The
parameters
can
commit
immediately
and
if
redundant
coordinators
act
more,
the
coordinate
has
a
variety
is
crossed
to
the
100
percent
and
the
probability
of
the
theorist
adults
can
be
as
close
to
their
percent
as
possible
and
the
current
implementation,
the
Redundant
coordinate,
is
pursed
into
the
single
pod,
but
you
have
to
coordinated
the
geographically
distributed
or
rearranged.
It
will
be
possible
to
provide
like
a
days
after
recovery
and
so
on.
Now.
A
A
The
second
edition
is
that's
separated
of
business
application
and
transaction
control
will
allow
us
to
solve
the
problem.
White
Houston
died,
the
FBI
in
interface,
between
business,
application
and
transaction
control,
components
find
the
standard
we
can
create
a
library
in
the
language
such
as
a
go
or
JavaScript.
That
works
the
same
as
the
trend
at
map
transduction,
and
it
starts
from
the
cap
theorism.
A
A
The
cap
seller
is
a
Capcom
gesture
as
a
hypothesis
that
the
distributed
system
can
not
satisfy
consistency
and
availability
and
partition
tolerance
all
at
the
same
time
and
in
the
crowd
about
Liberty
and
the
partition
tolerance
importance.
So
consistency
in
the
ensure
about
eventual
consistency
and
contract.
We
have
been
flying
the
Google
or
Dr
Lynch's
academic
Pro
of
the
cap,
as
the
partition
does
not
occur.
Both
consistency
and
viability
can
be
achieved
and
purchasing
really
occurs
in
the
real
system,
even
and
even
if
it
does
occur.
A
Fox
Texas
can
prevent
the
lack
of
consistency,
because
paxos
provides
a
redundancy
and
wait
for
the
comet
during
the
passion
no
current,
and
when
Consciousness
and
departure
lands
are
required.
One
hundred
percent
of
our
ability
cannot
be
achieved.
However,
redundancy
can
bring
your
ballability
at
close
to
100
as
possible,
in
other
words
using
distribute
consensus
as
a
localism,
such
as
the
foxes
who
arrived
or
something
enough,
correct
consistency
and
availability
and
partition.
Tolerance
can
all
be
satisfied
and
simultaneously.
A
The
forest
and
the
final
point:
we
are
still
considered
the
issue
that
the
DB
needs
to
be
the
except
for
compliant
and
that
more
than
DB
is
actually
no
SQL
can
be
accepted.
However,
this
is
can
be
solved
by
the
important
transaction
wrap
on
no
SQL
like
the
scroll,
DB
or
cherry
gaussian
paper
approach.