Data
Center
Evolu.

on
and
the
“Cloud”


Paul A. Strassmann
George Mason University
November 5, 2008, 7:20 to 10:00 PM

1

Hardware
Evolu.on


2

 Where
is
hardware
going?

  x86
con(nues
to
move
upstream

  Massive
compute
power
available


  Power
usage
grows
even
more
cri(cal

  Highly‐connected
compute,
networking,
and
storage

Emergence
of
high‐speed,
shared
interconnects

Shared
storage
becomes
less
expensive
and
more
pervasive

Networking
and
storage
riding
x86
improvement
curve




Virtualiza(on
transforms
powerful,
disparate
hardware

into
Virtual
Infrastructure


3

Non‐Virtualized
World


Exchange File/Print
Operating System Operating System
Operating System Operating System
Virtualization Virtualization

VPN CRM
Operating System Operating System
Operating System Operating System
Virtualization Virtualization

4

Enables
the
Virtual
Datacenter


Virtual Infrastructure

CPU
Pool

Memory
Pool

Storage
Pool

Interconnect
Pool

5

 Self‐Healing

Exchange CRM File/Print

APP APP APP APP APP APP

APP APP APP

Virtual Infrastructure

CPU
Pool

Memory
Pool

Storage
Pool

Interconnect
Pool

6

 Self‐Op.mizing,
Self‐Protec.ng

Datacenter


Moves
func.onality
tradi.onally
considered
“management”

into
the
infrastructure

Any
applica.on
gains
the
benefits

Provides
new
reasons
to
go
virtual


7

 Evolu.on
of
a
Virtual
Datacenter

IT Service IT Service Infrastructure Desktop
Delivery Continuity Optimization Management

Management
and

Automa(on


Virtual
Infrastructure

PlaDorm


Hypervisor

Hypervisor


8

The
Mobile
Workplace



9

Why
Use
Thin
Client?


10

Energy
Efficiency
–
Thin
Clients



11

So[ware
Evolu.on


12

 Where
is
so[ware
going?

  More
diversity
in
Opera(ng
Systems
and
Applica(ons

Simultaneous
use
of
Linux,
Solaris,
and
Windows
farms

Customized
open
source
stacks
and
diverse
versions/distribu.ons

  Services
becoming
disaggregated
and
distributed

SOA
leads
to
many
more
components
in
use

So[ware
as
a
Service
and
hosted
compu.ng
op.ons
growing

  Management
task
becomes
even
more
difficult

Need
a
way
to
normalize
management
approach

Can’t
s.fle
so[ware
progress
along
the
way


13

So[ware’s
Growing
Complexity

Process
 Process
 Process


App
Stack
A
 App
Stack
B
 App
Stack
C


DB2
9


Different
processes
must
be
executed
differently

depending
on
the
hardware,
Opera(ng
System,
or

applica(on
environment

14

 Legacy
Processes
Can
Be
Moved

Standard
SoOware
Containers
 Complex
applica(on

vService
A
 vService
B
 vService
C
 stacks
become

standardized

soOware
containers


App
Stack
A
 App
Stack
B
 App
Stack
C


With
common

processes
for

Virtual Infrastructure management

DB2
9


Physical
Hardware


15

Next
Phase:
Connec.ng
the
Clouds

Extends
Virtual
Infrastructure
beyond
single
datacenter


Uses
secondary
Data
Center
site
for
workload
overflow

Leverages
geographically
distributed
resources


Rents
resources
from
Service
providers
for
capacity
overflow


Maintains
IT
Service
Service
Level
Agreements


Virtual
Infrastructure

Virtual
Infrastructure


Secondary

 Primary

 Hos(ng



16

The
“Cloud”


17

 What
is
a
Cloud?

•  A
“cloud”
is
a
totally
reliable,
extensible,
and

manageable
so[ware
plaeorm
that
delivers
a
self‐
healing
and
self‐managing
datacenter.


•  It
aggregates
on‐premise
servers,
storage
and
network

into
“an
internal
cloud”
that
also
federates
with

“external
clouds”
of
compu.ng
capacity
that
frees
I.T.

from
the
constraints
of
hardware‐mapped
applica.ons.


•  The
“cloud”
guarantees
the
high
levels
of
availability,

low
latency,
security
and
scalability
for
all
applica.ons

independent
of
hardware
and
loca.on.



18

 How
the
“Cloud”
is
Different

•  Unlike
a
tradi.onal
Opera.ng
Systems,
which

are
op.mized
for
a
single
server,
the
virtual

datacenter
OS
–
the
“cloud”
‐
serves
as
the
OS

for
the
en.re
datacenter.


•  Datacenters
of
cloud
providers
and
So[ware

as
a
Service
companies
can
operate
the
virtual

data
center
using
cheap
commodity
hardware

because
the
compu.ng
is
redundant.



19

 A
Virtual
Data
Center
Delivers

•  A
set
of
infrastructure
services
to
seamlessly
aggregate

on‐premise
servers,
storage
and
network


•  A
set
of
cloud
services
to
federate
the
on‐premise

infrastructure
with
third
party
cloud
infrastructure


•  A
set
of
applica.on
services
to
guarantee
the
right

levels
of
availability,
security
and
scalability
to
all

applica.ons
independent
of
hardware
and
loca.on.


•  A
set
of
management
services
that
allow
to
proac.vely

manage
the
virtual
datacenter
and
the
applica.ons

running
on
it.



20

 Evolu.on
Towards
a
“Cloud”

•  Virtualiza.on
it
has
evolved
to
include
not
just
a

hypervisor
but
a
comprehensive
set
of
capabili.es
that

minimize
planned
and
unplanned
down.me,
ensure

applica.on
service
levels
and
automate
rou.ne

processes
in
the
datacenter.


•  Virtualiza.on
is
an
enabling
technology
that
frees
up

applica.ons
from
physical
servers
and
enables

hardware
independence.


•  Virtualiza.on
is
now
evolving
to
a
dynamic
datacenter

with
infrastructure
services
for
aggrega.ng
compute

capacity
and
for
automated
restart
of
redundant

datacenters
affected
by
so[ware
failure.



21

 Fault
Tolerance

•  Solu.ons
to
deliver
zero
down.me,
zero
data

loss
and
+Six
Sigma
availability.

•  Assure
execu.on
of
applica.ons
based
on

chosen
microprocessors.

•  Deliver
fault
tolerance
without
the
cost
and

complexity
of
hardware
duplica.on.


•  Quick,
simple
and
cost
effec.ve
backup
as
well

as
recovery
for
all
applica.ons.

 


22

 Security
&
Control

•  Offers
visibility
into
all
machine
resources
and

processes.

•  Monitors
and
controls
the
execu.on
of
all

applica.ons.

•  Stops
viruses,
rootkits
and
malware
before

they
can
infect
a
system.




23

 Scalability

•  Enables resource intensive applications to
run up to 200,000 I/Os per second.
•  Offers hot add capability for increasing
hardware resources.
•  Enables applications to scale without
disruption or planned downtime.
 


24

 Storage
Management

•  Dynamic storage allocation enables the
reduction of storage by >50%.
•  Uses memory only as required and tracks
actual usage for billing.
•  Reduces storage by sharing common user
data.
•  Links
all
storage
for
managing
capacity
and

improving
performance.

25

 Economics

•  “Cloud” data centers can be billed in actual
usage increments.
•  The infrastructure can obtain processing
power from “clouds” of external vendors.
•  Capacity can be distributed across several
data centers for handling of peak loads.

26

 “Cloud”
or
“U.lity”
Compu.ng


•  The
networked
infrastructure
provides

capacity
in
an
on‐demand
environment.


•  “Cloud”
compu.ng
offers
the
capacity
to
pay

only
for
actual
usage,
as
a
“u.lity”.

•  We
prefer
the
term
“u.lity”
compu.ng
to

describe
what
is
described
as
a
“cloud”.



27

 Megatrends

•  Progress
from
a
device
centric
world
to
a
person

centric
world.

•  Enables
connec.ng
from
anywhere,
by
any
means.

•  Offers
access
privileges
only
to
authorized
persons.

•  Allows
purchasing
of
computer
processing
power

independent
of
circuit
technology.

•  Makes
it
possible
to
associate
compu.ng
services

according
to
a
person’s
roles
or
loca.on.



28

 Direc.ons

•  Virtual
Infrastructure
aggregates
hardware
into

flexible
resource
pools
and
creates
a
truly
virtual

datacenter


•  Virtualiza.on
simplifies
so[ware
management,

and
extends
this
even
further
with
central

management
of
so[ware
configura.on

•  Virtual
mobility,
virtual
infrastructure,
enables

effec.ve
use
of
resources
that
are
local
or

remote,
owned
or
rented

29

Virtualiza.on
to
Mul.ple

Datacenters30



30

 Summary

•  The
current
prolifera.on
of
servers
has
been

costly,
unreliable
and
insecure.


•  It
has
decreased
the
quality
of
service.

•  Clustered
servers
cannot
offer
the
fault

tolerance,
scalability
and
the
favorable

economics
of
“Cloud”
compu.ng.

•  Migra.on
to
a
Service
Oriented
Architecture

must
include
the
“Cloud”
in
its
architecture.




31

Amazon
Elas.c
Compute
Cloud

(Amazon
EC2)


32

 What
is
a
Cloud
Service?


•  Amazon
Elas.c
Compute
Cloud
(Amazon
EC2)

is
a
web
service
that
provides
resizable

compute
capacity
in
the
“cloud”.
It
is
designed

to
make
web‐scale
compu.ng
easier
for

developers.


33

 Example
of
a
Service:
Elas.c
Block

Store
(EBS)

•  Provides
block
level
storage
volumes
for
use

with
EC2
applica.ons.


•  EBS
volumes
storage
that
persists

independently
from
the
life
of
an
applica.ons.


•  EBS
is
par.cularly
suited
for
applica.ons
that

require
a
database,
file
system,
or
access
to

raw
block
level
storage.


34

So[ware
Available
on
EC2


35

Usage
Pricing


36

Costs
Known
in
Advance


37

 Summary:
Key
Features
of
EC2

•  Allows
you
to
obtain
and
configure
capacity
on

demand.

•  Provides
complete
control
of
compu.ng
resources
and

lets
you
run
in
a
proven
compu.ng
environment.

•  Reduces
the
.me
required
to
obtain
and
boot
new

servers
in
minutes,
allowing
to
scale
capacity
as

compu.ng
requirements
change.

•  Changes
the
economics
of
compu.ng
by
paying
only

for
capacity
used.


•  Offers
tools
to
build
failure
resilient
applica.ons.


38