Cloud Storage and Intel Power Management

agreeablesocietyΤεχνίτη Νοημοσύνη και Ρομποτική

29 Οκτ 2013 (πριν από 3 χρόνια και 11 μήνες)

92 εμφανίσεις

1

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Cloud Storage and

Intel Power Management

Usage Oriented Reference Architecture

Enrique Castro
-
Leon, Ph.D.

End User Platform Integration

October 2009

Rev 1.0

2

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Summary


The Cloud Computing Reference Architecture Model


Historical context for Cloud Storage 2.0


Cloud storage yesterday, today and tomorrow


The growth of the digital universe and power considerations


Use cases for storage services producers and consumers


Usage Oriented Reference Architecture for Cloud Storage


CPU, servers, storage devices, rack, row, whole data center


How it works within and across architectural building blocks


Data center system
-
level use cases: platooning

3

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Cloud Applications Reference Architecture Model

Usage:
Discovery,
exploration phase, market
segment analysis

Audience:
C level
executives

Content:


High Level
conceptual
a
rch

Example:

Cloud
infrastructure RA

Usage:
Solution architecture discussion

Audience:
CTO, IT architects, IT Managers

Content
:
Usage
specific/environment
g
eneric

Example:
Cloud Storage Power Management

Usage:
App development/prototype discussions

Audience:
IT architects &
engineers

Content:

Usage
specific/environment
s
pecific

Example:


Power management for cloud application

Usage:
Strategic planning and product
dev

Audience:
Platform architects and engineers

Content:

Solution
concept proposals

Example:


Next
gen. power mgt
sol’n

proposal

Usage:

Defining
product
r
oadmap

Audience:


Prod. Dev. engineers

Content:
Project dev. documentation

Example:
Prod.
Reqs
. &
impl
. plan

High Level

30000’

General

Purpose

Reference

Architecture

(GPRA)

Technical

1000’

Usage Oriented

Reference Architecture

(UORA)

Deep technical

Special Purpose Reference
Architecture (SPRA)

Concept &

Design

Product Roadmap

Development &

Implementation

Product Hardware, Software
Architecture

End
Users & SIs

ODMs & OEMs

4

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Cloud Storage 2.0

Yesterday


New name for an old idea


IBM*, Sun* early pioneers in
service
-
based storage back to
early 90s and before


“The Network is the Computer”


Economies of scale through
provider specialization


Roadblocks to scaling


Networking costs, standardization
and security


Customer refusal to share physical
infrastructure kills economies of
scale

Today


Vision is good, but primary driver is
now economic


DC disk storage expensive


Storage to grow 300% by 2010**


Progress in standardization


Virtualization


Makes logical isolation of tenants
practical


Power consumption


Spoiler for cloud storage providers


Tomorrow


Size of digital universe


2.25 x 10
21

bits in 2007 or 281
billion gigabytes**


Expected to grow tenfold by
2011**


Emerging applications and usages


Not about corporate storage
anymore


Bring consumers and emerging
markets into the cloud


New data types


Transition from analog to digital
TV and digital photography


Surveillance data


Sensor based networks


RFID

**Source
: John F.
Gantz

et al., “The Diverse and Exploding Digital Universe”,
IDC
March 2008

Cloud Storage Use Cases

Cloud
Storage

Traditional

dedicated

corporate

storage

Storage Virtual
Infrastructure

Services

Digital media

creation &

distribution

Rich

Internet

Applications

Data

Presence

Service

Application &

Data Storage
Services

Online

Storage
Providers

Corporate

storage

services

Business Use

Cases


In
-
sourced
vs. Out
-
sourced

Backup

Personal

store
-
n
-
sync

Family

Movies
-
2
-
go

Social

Computing

Other
corp

cloud
-
based

apps

Storage
-
based

applications

Streaming

Storage providers

Consumer use cases

(individual & corporate users

Storage

consumers

In the Office

6

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Challenges & Opportunities (consumers)


Challenges


Risk of data loss


Solve backup problem


Proliferation of storage media


Unwieldy data management


Data volume increasing


Digital media: audio, video, pictures, data


Media obsolescence


Data retrieval, migration


Opportunities


A la carte storage services


Data presence (Pi Corporation*), backup (
Mozy
*)


Data stewardship


DR, offsite storage

7

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Challenges & Opportunities

(corporate & service providers)


Challenges


Exponential growth of digital footprint


New media, Web 2.0 driving storage requirements


Cost of in
-
house data storage


Energy usage can’t be ignored


Opportunities


Virtualization driving storage usage/delivery


Service oriented IT with SLA
-
driven products


Retention, privacy, access, management, security


Service decoupled from underlying technology


Cloud based storage


Integrated power management for storage devices


Organization
-
wide, uniform policies


Advanced info capture, search, discovery, classification


8

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Power Management in Cloud Storage

Efficiency on the rise,
but


Number of devices
increasing faster than
efficiency gains


Storage devices


Devices driving demand for
cloud storage


Net power consumption on
the rise from move to
denser form factors


4U
-
> 2U
-
> 1U
-
> Blades

Goals for Cloud Power
Management in
hosting side


Attain “impedance match”


Server power draw & DC
infrastructure


Reduce stranded power
and power over
-
provisioning


Increase operational
flexibility and ability to
manage power related
events


Appliance paradigm

Racking

density

Nameplate rack

p
ower draw

range

(
KW)

4U

1



4

2U

2


8

1 U

3


16

Blade

4


25+

9

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Power Management Data Center Use Cases

Active power capping to reduce stranded power


Maximize servers in rack under fixed rack power


Maximize utilization of branch circuits

Power & thermal monitoring and event management


Monitor, subscribe/unsubscribe to power & thermal events


Trim power or turn off equipment as response to emergencies

Allocate power to workloads to fulfill SLA


Prioritize power allocation to nodes in a group


Hierarchical power allocation

Power forecasting in data center planning


Use historical records for power dispatching


Use forecasts to negotiate favorable electricity rates

Power metering and chargeback services


Historical power data enables fine grained metering & chargeback


9

10

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Usage
-
Driven Power Mgt Architecture

Main architectural pattern


Nested feedback loops

Power management conceptual levels


Row and higher levels including whole data center


Appliance or rack


Server group


Single
-
node or server + storage subsystem


CPU/Memory

11

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Intel Power Management Technologies
Portfolio for Hyper
-
scale Environments

Power control through nested control loops

Data Center

CPU P / T state control

ACPI, low
power SSDs

Intel® Intelligent Power

Node Manager

(single node)

Intel® Data Center Manager

(10
-
10,000+ nodes)

OEM, ISV, SI solutions

10 W

100 W

1 KW

10 KW

Row

Rack

100 KW

Server

Chipset

CPU

Power Control Range

Hardware entity

Implementation technology

IPMI, Data
Center

Management Interface

Mem

Solid State Drives


Legend:


3
rd

Party Mgmt SW


Intel


Server OEM


11

12

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Cloud Storage Power Management Usage Framework

12

Intel Confidential

ACPI P
-
state control

Power control

algorithms

Recursive composition

Node level

power management

Data Center Management Systems

Power feedback loops & state estimation physical models

Management S/W
(multi
-
level)

Other P/T sensors,
controls & algorithms

Rack
Optimization

Maximize server
rack capacity in
power constrained
racks

Branch Circuit

Optimization

Optimize DC
pwr

infrastruct

Reduce stranded power

Platooning

&

workload
rebalancing

Optimize workload ramp
-
up,

energy usage in virtualized
environments

Chargeback

Fine grained
reporting, billing,
auditing


P/T Monitoring

Control power
consumption against
forecast targets

13

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

S

PSMI

S

PMBus

S

PMBus

CPU

NM

PMBus

Group
Policies

DCM

Di sk Array

Encl osure

Intelligent

PDU

Row Power
Manager

Building
Management
System

Intel DCM technology

cloud plug
-
in

Node Manager
performs vol tage &
frequency scal ing to
keep server power
wi thi n assi gned node
power target

Power to

storage

subsystem

DCM assigns power targets i n

i ndi vi dual nodes dynami cally

subject to compute
subsystem power constrai nts

Per node actual

Power

consumed

Actual

Power

consumed

Cloud storage
application

Rack
Policy

Engine

Storage

control

Node

control

Gl obal pol i cy:
Appl i ance power not
to exceed 5 KW

Di sk Array

Encl osure

Di sk Array

Encl osure

1

Reduce basel i ne

power i n storage
subsystem
through

use of SSDs

2

Cl oud storage power

pol i cy engi ne budgets storage
and node power to meet 5KW
objecti ve


3

Cl oud storage pol i cy
engi ne manages storage

power through

sel ecti ve spi n
-
downs

4

5

Cl oud storage pol i cy
engi ne sets power target

for
compute & storage

subsystems

6

7

Storage
subsystem set
power

Compute
subsystem set
power

Actual

Power

consumed

Actual

Power

consumed

Cloud Storage Systems Integration Framework

14

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Cloud Storage Power Management
Reference Architecture

Set power (IPMI)

Power
change

Power metadata (IPMI)

BIOS ACPI

power & thermal


state control

i nterface

P
-
state


tabl e

Actual P
-
state

Operati ng system

or vi rtual ization

hypervi sor

Actual power

consumed

CPU power
control

NPTM policy

engine

Temp Mon

PSU Mon

Monitors

PMBus


P/S

PMBus

interface

Temp

sensor

NM firmware

Set P
-
state

Node power control

CPU power control

Power

CPU

Node Adapter Interface (pre
-
defined)

DC Control

CMS Core



Event Mgr

Policy
Data

Trending/
Query Data

NM

Node Adapter Servi ce (pl ug
-
i n framework)

DCM Network Communication Li b

Power Mgmt

Policy
Resolution

Aggregator

Resource
monitoring &
Mgmt


Discovery &
provisioning

DCMI

i PDU

i DRAC

Blade

DCM Node control

DCM
Console

Java in
-
process

3
rd
-
party

IB



Intel® Data Center Manager

Presentati on subsystem

(consol es, dashboards, GUIs)

Data Center Mgt

Building Mgt System

Operational Control

Pol i cies

Engi ne

Power

control

Info
col l ecti on

Pl anni ng &
Management

Moni tori ng &
Automati on

CMDB

Set power & power
metadata (REST)

Compute

Subsystem

set pwr &

pwr readout

(REST)

Apache

REST

SOAP

Native client

access API

Data

storage

Metadata

Servi ce

Metadata

storage

Storage

Application

Storage

service

Power

control

PDU

Power

readout

Cloud Storage Application

Power

Set power & power
metadata (REST)

Power

control

Appliance Group Manager

Other
envi ronmental
sensors (temp,
ai rflow, humidity)

KVM subsystem

15

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Cloud Storage Appliance
-
Level Replication

Rack level power and environmental sensor data (SOAP)

Branch circuit and per appliance power allocation

Application


power

policy (SOAP)

Row Power
Manager

The Row Power Manager can control multiple appliances

Cloud

storage

application

Rack Pol i cy

Engi ne

Storage

control

Node

control

Cloud

storage

application

Rack Pol i cy

Engi ne

Storage

control

Node

control

Cloud

storage

application

Rack Pol i cy

Engi ne

Storage

control

Node

control

Cloud

storage

application

Rack Pol i cy

Engi ne

Storage

control

Node

control

16

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Energy Management Use Cases

Power vs. Energy Management

t
1

t
2

t

Power

saved at

time t

Energy saved

between t1 and t2

is the area between

the two curves

17

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Integrated Power Management


Power Mgt mechanisms have operational restrictions


Node Manager: high workload


DBS: low workload


NHM power proportional computing: P
baseline

limited to about 50%


Pooled virtualized environment


Combine mechanisms


Extend operational envelope


Lower
Pbaseline


Example: Server platooning

18

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Advanced Use Case: Server Platooning

Active, high priority

workloads

Unconstrained power

Active workloads

Power capped

Idle

Asleep

(ACPI S3, S4 or S5)

100%

70%

50%

2%
-

25%

Takes 1 to 5 min

to bring node online

Power

consumption

Transitions for

increasing workload

Transitions for

decreasing workload

Platooning state

S

PSMI

S

PMBus

S

PMBus

CPU

NM

PMBus

S

PSMI

S

PMBus

S

PMBus

CPU

NM

PMBus

S

PSMI

S

PMBus

S

PMBus

CPU

NM

PMBus

S

PSMI

S

PMBus

S

PMBus

CPU

NM

PMBus

19

* Other names and brands may be claimed as the property of others.




Copyright © 2009, Intel Corporation.

Conclusions


Digital universe growing exponentially


Data spilling from the corporate data center into the cloud


Power usage growth slower but still unsustainable


Usage drives power management architecture


Paradigm of arbitrarily scalable systems


Composition & replication


Spans CPU to whole data center


Power control range from watts to hundreds of kilowatts


Power vs. energy management