hp apollo slidecast

© Copyright 2014 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.

HP Apollo SystemsJohn Gromala, Senior Director, Hyperscale, Product Management, HP

Servers

© Copyright 2014 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.2

HP Confidential until June 9, 2014

Solving global problems requires greater…

• Geophysical Sciences

• Energy Research & Production

• Meteorological Sciences

• Government

• Academia

• Research & Development

• Life Sciences

• Pharmaceutical

• Entertainment

• Media Production

• Visualization & Rendering

• Computer-Aided Engineering

• Electronic Design Automation

AccessibilityPerformance Efficiency

• Financial Services

http://intranet.hp.com/enterpriseservices/ww/site/industries/manufacturing.html

http://intranet.hp.com/enterpriseservices/ww/site/industries/hls.html

http://intranet.hp.com/enterpriseservices/ww/site/industries/me.html

http://intranet.hp.com/enterpriseservices/ww/site/industries/fsi.html

http://intranet.hp.com/enterpriseservices/ww/site/industries/ps.html



Reinventing HPC to accelerate the world of tomorrow• HP’s advanced HPC compute solutions to

solve the world’s most difficult problems

– Rapidly ramp performance for accelerated results

– Maximize rack-scale density and energy efficiency

– Unleash HPC with an infrastructure that is affordable, less complex, and easy to manage

– Flexibility to precisely meet workload needs

• Expanding access to HPC resources via HPC Innovation Hubs, and HPC as a Service

Introducing: The new Apollo family



10/100/1000Base-T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53

Speed: Green=1000Mbps, Yellow=10/100Mbps Duplex: Green=Full Duplex, Yellow=Half Duplex

54

51 5249 50Green=10Gbps, Yellow=1GbpsSFP+

Unit

SYS

ModeGreen = SimplexYellow = Duplex

HP A5800 Series Switch JG225A

10/100/1000Base-T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53


54


Unit

SYS

ModeGre en = SimplexYellow = Duplex


10/100/1000Base-T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53


54


Unit

SYS



10/100/1000Base-T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53


54


Unit

SYS



42

41

40

39

38

37

36

35

34

33

32

31

30

29

28

27

26

25

24

23

22

21

20

19

18

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

42

41

40

39

38

37

36

35

34

33

32

31

30

29

28

27

26

25

24

23

22

21

20

19

18

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

47

46

45

44

43

47

46

45

44

43

The best performance for your budget

Leading performance per $ per watt• up to 4x more performance/$/watt• up to 60% less floor space

Rack scale efficiency• 160 x 1P servers per rack with

10 hot-pluggable dual-server trays per 5U chassis

• Maximize rack-level energy efficiency

Tailor to the workload for lower TCO• Mix compute, accelerator, storage

and networking to fit workload needs

Introducing the HP Apollo 6000 System

Electronic Design Automation• Computer chips• Cellular phones• Pacemakers• Controls for automobiles and satellites • Servers, routers and switches

Monte Carlo Simulations• Investment analysis• Financial derivatives• Physical sciences• Engineering• Computational biology• Computer graphics• Gaming



HP Apollo 6000 System Overview

First available tray• ProLiant XL220a dual-server tray

• Front serviceable

• Rear cabled solution

• Max power of ~169W per tray

Rack scale• 160 nodes per 48U rack

• 5U chassis (1.0m deep rack)

• 20 nodes per enclosure

• Front service, rear cabled

10/100/1000Base-T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

5 3


54

51 5249 50

Green=10Gbps, Yellow=1GbpsSFP+

Unit

S Y S

M o d eGreen = SimplexYellow = Duplex

H P A5 8 0 0 S e r i es

S w i tc h J G2 2 5A

10/100/1000Base-T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53


54

51 524 9 50


Unit

S Y S


H P A5 8 00 S e r i es


10/100/1000Base-T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

5 3


54

51 5249 50


Unit

S Y S


H P A5 8 0 0 S e r i es


42

41

40

39

38

37

36

35

34

33

32

31

30

29

28

27

26

25

24

23

22

21

20

19

18

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

42

41

40

39

38

37

36

35

34

33

32

31

30

29

28

27

26

25

24

23

22

21

20

19

18

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

High performance computing

CPU

DIM

Ms

2x 1P

CPU

DIM

Ms

Shared power & cooling• Efficient pooled power shelf supports up to 6 chassis

• N, N+1, 2N redundancy configs

• 12 volts DC output with max power of 15.9kW

• Advanced Power Manager

Rack-level shared infrastructure for efficiency and flexibility

• Highest frequency per core

• Intel E3-12xx v3 Haswell

• CPU core generation ahead

• Single-threaded applications

• Max turbo frequency of 4GHz

• Low latency: No 2P cache coherency



Next generation density and efficiency at scale

Dell M620 2 1/2 racks

20% more performance*60% less rack space 46% less power**

$3M

HP Apollo 6000 System with ProLiant XL220a Servers, 1 rack

TCO savings10/100 /1000Base -T

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53

Speed: Green=1000Mbps, Ye llow =10/100Mbps Dup lex: Green=Fu ll Dup lex , Ye llow =Hal f Dup lex

54

51 5249 50Green=10Gbps, Ye llow =1GbpsSFP+

Unit

SYS

ModeG re e n = S im p le x

Y e llo w = Du p lex


10 /100 /1000Base -T33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53

Speed: Green=1000Mbps, Ye llow =10 /100Mbps Dup lex: Green =Fu ll Dup lex, Ye llow =Hal f Dup lex

54

51 5249 50Green =10Gbps, Ye llow =1GbpsSFP+

Unit

SYS

ModeG re e n = S im p lexY e llo w = Du p lex

HP A 5800 Series Switch JG225A

10/100 /1000Base -T

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53


54


Unit

SYS

ModeG re e n = S im p le xY e llo w = Du p lex


10/100 /1000Base -T

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4817 18 19 20 21 22 23 24 25 26 27 28 29 30 31 321 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

53


54


Unit

SYS

ModeG re e n = S im p le xY e llo w = Du p lex


42

41

40

39

38

37

36

35

34

33

32

31

30

29

28

27

26

25

24

23

22

21

20

19

18

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

42

41

40

39

38

37

36

35

34

33

32

31

30

29

28

27

26

25

24

23

22

21

20

19

18

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

47

46

45

44

43

47

46

45

44

43

The best performance for your budget


“We are seeing up to 35% performance increase in our Electronic Design Automation application workloads. We have deployed more than 5,000 of these servers, achieving better rack density and power efficiency, while delivering higher application performance to Intel silicon design engineers.”Kim Stevenson, CIO, Intel



Advancing the science of supercomputingThe New HP Apollo 8000 System

Leading teraflops per rack for accelerated results• 4X teraflops/sq. ft. than air-cooled systems• > 250 teraflops/rack

Efficient liquid cooling without the risk• 40% more FLOPS/watt and 28% less

energy than air-cooled systems• Dry-disconnect servers, intelligent Cooling

Distribution Unit (iCDU) monitoring and isolation

Redefining data center energy recycling• Save up to 3,800 tons of CO2/year (790

cars)• Recycle water to heat facility

Scientific Computing• Research computing• Climate modeling• Protein analysis

Manufacturing• Product modeling • Simulations• Material analysis



Dry disconnect servers• 100% water cooled

components• Designed for serviceability

Apollo 8000 System Technologies

Open door view of 4 compute & redundant CDU racks

Management infrastructure• HP iLO4, IPMI 2.0 and DCMI

1.0• Rack-level Advanced Power

Manager

Power infrastructure• Up to 80kW per rack• Four 30A 3-phase 380-

480VAC

Intelligent Cooling Distribution Unit• 320 KW power capacity• Integrated controls with active-active

failover

Warm water• Closed secondary loop in

CDU• Isolated and open facility

loop

Advancing the science of supercomputing



Differentiated: Dry disconnect servers

• Enables maintenance of servers without breaking a water connection• Inside the server tray, heat is transferred from components via vapor in sealed heat

pipes• Thermal bus bars on the side of the compute tray transfer heat to the water wall in the

rack• Water flows through thermal bus bar in the rack from supply-and-return pipes• Fluid fully contained under vacuum

New patented technology making a liquid-cooled system as easy to service as air-cooled

Thermal bus bars

Sealed heat pipes

Sealed heat pipes



Failure is not an option

• Dry disconnect servers: sealed heat pipes cool components

• Facility water isolated from IT loop• Takes ASHRAE spec water

• Secondary IT loop vacuum keeps water in place

• Intelligent Cooling Distribution Unit designed to minimize and isolate issues

• Comprehensive system insight and management built on Advanced Power Management and smart sensors

Efficient liquid cooling without the risk



Accelerating the science of supercomputing

IBM rack servers(Air cooled)

4X teraflop performance per sq. foot40% more FLOPs/watt28% less energy$1M savings

2.6X more energy for facilities infrastructure

up to 3800 tons more CO2/year

save over 1200 ft for a 1 petaflop system

annualized PUE 1.06 or better

HP Apollo 8000 system(Liquid cooled)



World’s largest supercomputer dedicated to advancing renewable energy research

• $1 million in annual energy savings and cost avoidance through efficiency improvements

• Petascale (one million billion calculations/ second)

• 6-fold increase in modeling and simulation capabilities

• Average PUE of 1.06 or better• Source of heat for ESIF’s 185,000 square

feet of office and lab spaces, as well as the walkways

• 1MW of data center power in under 1,000 sq. ft., very energy-dense configuration



Global impact: Reduce the carbon footprint• Top supercomputing performance has

grown 944X and the number of cores has exploded by 609X over 10 years*

• Power requirements have increased 5.5X, while power costs per kW have spiked

• The HP Apollo family is a new approach to HPC, with cost-effective and environmentally friendly HPC solutions

• Save while reducing carbon footprint

Save 3,800 tons of CO2 per year, ~amount of CO2 produced by 790 cars



HP Services for Apollo solutions

• Data center facilities• Workload migration• Big data, mobility,

virtualization design, planning and implementation

• Factory Express • Onsite installation • HP Education

• Datacenter Care supports your environment

• Proactive Care helps prevent problems

• Foundation Care helps solve problems faster

Support Consulting Services

Implementation

Data center planning and design services to achieve business outcomes

Ongoing support to help get connected and get back to business

Services to speed startup and build capabilities with new technology

Financing

• Available globally where HP Financial Services conducts business

• Technology refresh approach to allow for future scalability and upgrades

Flexible payment plan and terms to meet your needs



Extending HPC access to SMBs

HP – Intel HPC Innovation Hubs open up new opportunities to compete

• Partnering with universities, organizations and ISVs

• Access to HPC resources, modeling and simulation applications and expertise

• Result: Make vehicle components safer and more reliable

• Grow the industry, improve competitiveness in the global marketplace

Fueling innovation with HPC as a Service and HPC Innovation Hubs

HPC as a Service bridges the gap between public and private cloud• Access a common, open cloud platform

across standard IT cloud and HPC cloud initiatives

• Easy to use to increase competitive agility

• Open, efficient management of complex HPC resources

• Secure, scalable performance for your demanding workloads with leading cluster infrastructure and security integration

hp apollo slidecast

Technology

mbps duplex

simplex yellow

half duplex

hp confidential

hp servers

unit sys modegreen

u rack

tray rack scale