THE INFORMATION SOURCE FOR THE DATA CENTER INDUSTRY

Data Center Knowledge Guide to

Meet Apollo
Reinventing HPC and the Supercomputer

Bill Kleyman,
Cloud, Virtualization and
Data Center Architect
November, 2014

Brought to you by
 Meet Apollo Reinventing HPC and the Supercomputer

Summary CONTENTS
Administrators, engineers and executives are now
Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
tasked with solving some of the worlds most com-
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
plex challenges. This could revolve around advanced
computations for science, business, education, phar- Section 1 Understanding New
High-Performance Demands. . . . . . . . . . . . . 2
maceuticals and beyond. Heres the challenge many
New types of workloads and applications . . . . . . 2
data centers are reaching peak levels of resource con- New business and user demands. . . . . . . . . . . . . . . 2
sumption; and theres more work to be done. So how Scalability, management and growth. . . . . . . . . . 3
are engineers and scientists supposed to continue
Section 2 Overcoming HPC Challenges
working around such high-demand applications? and Creating Innovation . . . . . . . . . . . . . . . . . 4
How can they continue to create ground-breaking Resource utilization and scale. . . . . . . . . . . . . . . . . 4
research while still utilizing optimized infrastructure? Agility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
How can a platform scale to the new needs and Density and consolidation . . . . . . . . . . . . . . . . . . . . . 4
Money (and budgets) . . . . . . . . . . . . . . . . . . . . . . . . . 4
demands of these types of users and applications.
Creating Innovations A Liquid
This is where HP Apollo Systems help reinvent the Cooling Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
modern data center and accelerate your business.
Section 3 Fueling your
Next-Generation Compute Model with
Introduction Intel-powered Apollo . . . . . . . . . . . . . . . . . . . . 6
Innovation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
New applications are being deployed which require
Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
more performanceand that means more compute
Future innovation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
resources as well. These applications and data sets are
Benefits of cloud computing. . . . . . . . . . . . . . . . . . 7
critical to helping an organization run and stay ahead
of the competition. Most of allthese applications are Simplifying access to HPC. . . . . . . . . . . . . . . . . . . . . 7
gathering data, quantifying informationand pro-  ynamic content delivery and
D
graphics acceleration. . . . . . . . . . . . . . . . . . . . . . . . . 7
ducing critical results. One of the biggest challenges
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
surrounding the modern, high-intensity, applications
revolves around resource consumption and economies Final Thoughts. . . . . . . . . . . . . . . . . . . . . . . . . . . 8
of scale. Remember, were not just talking about server
platforms here. Its critical to understand where power,
cooling, and resource utilization all come into play. This is
why modern organizations requiring high levels of pro-
cessing power must look at new hardware systems capa-
ble of providing more powerwhile using less space
and being more economical. Its time to look at power-
fulmodularsolutions which can break all the norms
around high performance computing and data gather-
ing. This whitepaper explores HP Apollo Systems, where
it can be deployed and how it directly impacts mod-
ernhigh-performanceapplication requirements.

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 1

 Meet Apollo Reinventing HPC and the Supercomputer

Section 1 Understanding New High-Performance Demands

Its a different kind of computing world out there. In February 2012, the European Commission
The demand for more compute performance for announced that it has adopted a plan to
applications used by engineering, risk modeling, double spending on HPC to 1.2 billion, with
or life sciences is relentless. If you work with work- much of that money aimed at the installa-
loads like these, your success depends on optimiz- tion of additional petascale supercomputer
ing performance with maximum efficiency and systems.
cost-effectiveness along with easy management for In January 2013, the EU launched two
large-scale deployments. 1+ Billion, 10-year research investments
But here is the real challenge (in brain research and graphene).
New business and user demands.
High performance computing is being restricted by
the very thing that enables it: technology Science, academia, and even Bitcoin now
use high performance computing technolo-
Computers are the engines that drive the discov- gies to process massive amounts of informa-
eries that truly impact all of our lives. The mas- tion. Consider this exampleThe National
sive compute power they provide allows leading Renewable Energy Laboratory (NREL) focuses
research institutions in government, academia, on creative answers to todays energy chal-
and industry to run the simulations and analytics lenges. From fundamental science and energy
that are behind the most amazing advances in sci- analysis to validating new products for the
ence and technology. So what are some of the new commercial market, NREL researchers are dedi-
demands out there? What are some different appli- cated to transforming the way the world uses
cations behind supercomputer platforms? Lets energy.
look at a few:
New types of workloads and applications.
Super computers are fundamentally changing
the way we gather, quantify and process infor-
mation. Also, consider thisthe amount of
data were working with has increased dramati-
cally. The sheer number of data sets currently
traversing the modern data center is enough
to make anyones head spin. Governments and
massive organizations are investing billions
of dollars into high performance computing.
Why? The speed of technological development
now allows us to conduct research faster than
ever beforeand make life-saving decisions.
Predicting earthquakeseven if just a few
seconds beforehandcan be absolutely criti-
cal. In fact, a 2013 IDC report on Big Data and
HPC Trends indicates the following:

Source: HP Case Study National Renewable Energy Lab

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 2

 Meet Apollo Reinventing HPC and the Supercomputer

HPC technology has been used to speed break-

Because liquid cooling is 1,000X more efficient than throughs in science and engineering by enabling
air cooling, platforms like the HP Apollo 8000 System organizations to transition their research from the
physical laboratory to the digital world of simula-
allows you to take advantage of higher performance tions and computer analysis. Recent innovations in
components. That allows extremely dense configura- HPC technology are making these programs and
tions that deliver hundreds of teraflops of compute applications accessible to enterprise customers,
who are using them to enhance research, product
power in a very compact space while cooling up to
design, and development efforts. Ultimately, this
80 kW of power and supporting up to 144 servers in allows enterprises to gain a distinct competitive
a single rack. edge.

Its important to get a few things out of the

Thats the case with the organizations new way before we continue:
high-performance computing (HPC) system.
ff The amount of data traversing an HPC
While providing an astounding amount of
environment will only continue to grow.
compute power to drive renewable energy
research, the NREL Peregrine System is break- ff There will be an even greater need to run
ing new ground in energy-efficient comput- more parallel workloads to process even
ing with an innovative liquid cooling system more information.
developed by HP. ff As cloud, IT consumerization, and big data
The need for easy scalability, management, continue to evolvenew requirements will
and growth. be placed around the modern data center
From resource to applicationyour entire and the HPC workloads being hosted.
compute platform must be able to scale ff Your information and the speed at which you
quickly. However, if more performance requires can process it is absolutely critical to your
too much space and consumes too much business. And, in many cases, people directly
powerit cant meet the demands of the rely on this critical information to do very
market and the users. New technologies are specialized tasks.
now being developed to deliver unparalleled
performance without sacrificing efficiency.
To meet the dynamic demands of the market NowThere are new platforms which break
and the userHPC and supercomputer plat- through the barriers of traditional computing to
forms now utilize management and efficiency enable a high-density, energy-efficient, HPC solu-
technologies to create unparalleled levels of tion that uses a groundbreaking warm-water liq-
manageability. Consider thisBecause liquid uid cooling system to deliver faster, more energy-
cooling is 1,000X more efficient than air cool- efficient, and more sustainable infrastructure for
ing, platforms like the HP Apollo 8000 System research computing workloads than ever before.
allows you to take advantage of higher per- Lets examine how this new type of environment
formance components. That allows extremely is created.
dense configurations that deliver hundreds of
teraflops of compute power in a very compact
space while cooling up to 80 kW of power and
supporting up to 144 servers in a single rack.

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 3

 Meet Apollo Reinventing HPC and the Supercomputer

Section 2 Overcoming HPC Challenges and Creating Innovation

It all starts with what we knowand that is while gear as they can and then rely on cheap labor
the human imagination is limitless, the massive to handle the operational complexities. But is
space and energy requirements of traditional com- this the best approach? Could putting in place
puters are becoming limiting factors on the growth a better high performance computing infra-
of compute powerand slowing the pace of structure be a more optimal use of funds?
innovation. The clock is always ticking to find the Expanding the thinking around HPC and spe-
answer, find the cure, predict the next earthquake, cifically how the HP Apollo platform operates
and create the next new innovation. Thats really can not only help save dollars around infra-
why high-performance computing (HPC) is always structureit can greatly improve the elasticity
striving to find the answers faster to engineering, of your platform.
scientific, and data analysis problems at scale.
With all of this in mindThe only way to get past
But there are still some challenges to overcome. these challenges and to new kinds of solutions is
Resource utilization and scale. through a system capable of delivering perfor-
Islands of compute can become a real prob- mance as well as efficiency. Andthe best way to
lem for HPC environments. Organizations get to that point is through unparalleled innova-
have many islands of compute due to either tion. Days of traditional rack and data center man-
departmental boundaries (commercial) or agement are over. New concepts around power,
the mechanics of grant funding which gives cooling and density are defining how HPC plat-
researchers money to support hardware for forms scale and operate.
their research.
Agility. Creating Innovations A Liquid
The critical nature of data and the workloads Cooling Example
running these data sets must be agile and Regulating temperatures in the modern data cen-
capable of scale. Quickly flexing the amount of ter can be a challenging task. In this data center
resource available to individuals or teams can cooling survey, Data Center Knowledge set out to
be a real challenge. find out how those responsible for managing data
Density and consolidation. center environments coped with cooling them. It
Consolidation is very common in enterprise IT quizzed people with managerial roles in modern
environments because those applications in data centers to determine their current practices
that type of infrastructure are generally not all and attitudes toward data center cooling.
that resource intensive. However, HPC adminis-
trators outside of enterprise IT and in the world
Here are some of their findings:
of science and/or academia generally never ff There are significant benefits to using it, espe-
want to reduce the amount of hardware their cially newer technologies such as immersion.
workloads are utilizing. In fact, they are always ff Significant numbers of people are interested
looking for more hardware to allow them to in it, but many still think that it involves a
run bigger problems or to generate better major upheaval.
answers for existing challenges. ff Many people dont know what would prompt
Money (and budgets). them to consider it, but they are interested in
Organizations are still concerned about how it, suggesting a large and untapped market
much money is being spent on physical for a technology with lots of potential.
systems and how best to spend the money
to maximize value to end-users. Often HPC
sitesespecially academicbuy as much

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 4

 Meet Apollo Reinventing HPC and the Supercomputer

Its no wonder that cooling is so important to the Apollo Systems are designed to make it practical
people responsible for keeping the nations data to recycle the warm water used to cool the system
centers running. 85% of respondents answering as a heat source for their facilities. The U.S. National
the question agreed or strongly agreed that reduc- Renewable Energy Labs (NREL) did just that when
ing cooling costs is one of their highest priorities, they leveraged the supercomputers unique design
and the same proportion are always on the lookout to create one of the greenest data centers in the
for new opportunities to reduce cooling overhead. worldusing the waste heat from their supercom-
puter to heat adjacent office and lab space.
This is where some of the latest technologies and
innovations around cooling can really help out.
Liquid cooling not only enables higher-perfor- Leveraging the efficiency of HP Apollo
mance components, but also allows you to use 8000, we expect to save $800,000 in
the heat transferred to the water for facilities
operating expenses per year. Because we
heatreducing costs, and your carbon footprint.
When the National Renewable Energy Lab (NREL) are capturing and using waste heat, we
deployed an Intel-powered HP Apollo 8000 System, estimate we will save another $200,000
they slashed their costs by $1 million USD a year, that would otherwise be used to heat the
while putting their money where their values are. building. We are saving $1 million per year
Apollos new dry-disconnect server technology
keeps components cool and dry.
in operations costs for a data center that
cost less to build than a typical data center.
Systems like the HP Apollo 8000 System offer the
Steve Hammond,
worlds first 100-percent liquid-cooled supercom-
Director, Computational Science Center, NREL
puter with built-in technology that protects the
hardware. Rememberone of the most important
ingredients behind modern HPC requirements is Organizations which process massive amounts of
scalability. Apollo Systems work with a rack design dataand have to do it quicklysimply cannot
supporting up to 144 servers per rack. This trans- compromise with the infrastructure they deploy.
lates to about four times the teraflops per rack Whether youre in science, academia, government,
compared to air-cooled designs, and the energy- or even or compute-intensive enterpriseyour
efficient design helps organizations eliminate up to HPC environment is absolutely critical to you.
3,800 tons of carbon dioxide waste from their data Even more so is the need for this platform to run
center per year. optimally, with high levels of resiliency, and even
greater levels of efficiency. So, what
makes up this kind of system? What
should data center managers and
organizations look for when their
need revolves around powerful HPC
and super computing platforms?
Most of allWhat are the next-
generation innovation ingredients
behind high performance computing?
Lets take a look.

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 5

 Meet Apollo Reinventing HPC and the Supercomputer

Section 3 Fueling your Next-Generation Compute Model

with Intel-powered Apollo
New workload and data demands forced engineers Ingredients of HP Apollo: innovation
at HP to get creativereally creative. New innova- Multiple HP innovations make the Intel-powered
tions around liquid cooling and even intelligent HP Apollo 8000 System stand out from the compe-
Cooling Distribution Unit (iCDU) Racks has allowed tition. It leads the way in teraflops per rack for accel-
Apollo to break new ground in the HPC supercom- erated results. Its efficient liquid cooling is delivered
puting world. without the risk, while paving the way to data cen-
ter energy recycling and a more sustainable future.

HP Apollo was built from the ground up to transform 1. D

 ry-disconnect server trays contain sealed
heat pipes that cool server components
the data center by delivering improvements in total while keeping them dry
cost of ownership, performance and density at rack 2. T
 he HVAC power distribution system
scalebringing the potential power of HPC to eliminates most conversion steps to exceed
organizations of all sizes. Energy Star Platinum certification levels
3. T
 he HP Apollo 8000 iCDU Rack maintains
sub-atmospheric pressure for operational
resiliency and serviceability
NREL challenged HP and Apollo to meet
these demands: 4. Q
 uick-connect, modular plumbing kit can
be operational quickly as opposed to multiple
;; Enabling your environment to deliver
days or weeks
up to four times the teraflops per rack
than the old data center.
;; Operate a next-gen environment while
still lowering operational costs.
;; Reducing energy consumption with
dynamically intelligent systems.
;; Maximizing space efficiency and density
of the entire data center.
;; Supporting tens of thousands of users
running hundreds or even thousands
of parallel workloads.
The industry answer to all of these challenges now
revolves around the new HP Apollo HPC platform.
HP Apollo was built from the ground up to trans-
form the data center by delivering improvements in
total cost of ownership, performance and density at
rack scalebringing the potential power of HPC to
organizations of all sizes.
Engineers developed new server and resource
delivery methodologies to architect the HP Apollo
System.

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 6

 Meet Apollo Reinventing HPC and the Supercomputer

Heres the other big ingredient: management workload. Furthermore, there is direct integra-
With HP Apollo 8000 System Manager, you can see tion with OpenStack platforms.
and manage shared infrastructure power, and envi- Simplifying access to HPC.
ronmental controls, from a single console. To address the growing demand for HPC, HP
;; Save by avoiding spend on serial also introduced HP Helion Self-Service HPC, a
concentrators, adaptors, cables, and switches. private cloud solution based on the HP Helion
OpenStack cloud platform that provides a self-
;; Flex to meet workload demands with
service portal containing high performance
dynamic power allocation and capping.
compute resources via a user-friendly applica-
tion interface. Customers have the option of
By using tools like HP Insight CMU as an efficient self-managing the solution, or selecting HP to
manage the system with a pay-for-use model.
and easy-to-use management utility for cluster
Dynamic content delivery and graphics
administration you create direct visibility into your acceleration.
HPC cluster. Thats rightgraphics acceleration and pro-
cessing is really impacting the HPC environ-
ment. The most exciting development here is
Furthermore, HP offers an efficient and robust util- that members of the HPC community continue
ity for the management of HPC and hyperscale to push each other to create new technologies
clusters. By using tools like HP Insight CMU as an to deal with the volume of data and the size
efficient and easy-to-use management utility for of problems theyre trying to solve. So, for the
cluster administration you create direct visibility HP Apollo 8000 System, Intel created the new
into your HPC cluster. Currently, HP Insight CMU Xeon Phi 7120d, and NVIDIA created the K40d.
is used at some of the largest HP cluster deploy- What does this mean? Engineers and research-
ments, supporting multiple Top 500 sites, where it ers can now see results faster than ever before?
is not only valued for its capability to install, modify,
and update images rapidly but also for its ability to
identify and isolate performance issues. Engineers and researchers can now see
results faster than ever before.
Platforms like Apollo are also helping
ff F or the HP Apollo 6000, look for the
innovate for the future.
ProLiant XL250a Gen9 Server which
Benefits of cloud computing.
has 1 x 2P server, along with up to two
The idea isnt to put absolutely everything in NVIDIA Tesla K40 or K80 GPUs for
the cloud. In fact, HPC systems require dedi- an additional ~14TFLOPs accelerated
cated hardware environments to process performance.
and compute at the level that they require.
ff F or the HP Apollo 8000 System, look for
However, cloud computing has introduced an
the Intel Xeon Phi 7120d in the ProLiant
element of scalability and extensibility into a
XL740f and the NVIDIA Tesla k40d for the
private/hybrid cloud platform. Imagine being
ProLiant XL750f.
able to add self-provisioning, which now
allows departments located at different sites
or researchers/engineers to provision/de-pro- Furthermore, HP is investing in new HPC Innovation
vision the resources they need for their proj- Hubs to ensure that HPC expertise and compute
ect. The amazing piece around cloud is that resources are available to enterprises through part-
automation in using technologies like VMware nerships with local universities, community organi-
vCloud Director allow you to dynamically zations, independent software vendors and chan-
allocate required resources for a given HPC nel partners.

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 7

 Meet Apollo Reinventing HPC and the Supercomputer

Final Thoughts
The next-generation of HPC will revolve around Data Center Knowledge (DCK) is a leading
the ability to process vast amounts of data. online source of daily news and analysis about
Furthermore, integrated capabilities with the cloud the data center industry. We cover the latest
will allow for an even greater amount of resource developments and trends driving the powerful
and supercomputing scalability. HP Apollo answers growth in demand for mission-critical facilities,
the very questions that revolve around some of the the challenges and opportunities presented by
most challenging computational tasks. high-density computing and its impact on power
and cooling, and the evolution of the industry
You now have the capability to fuel ground-break-
to include cloud computing and modular
ing research in science and engineering with the
latest technologywhile going greener than ever data centers.
before. For more information about
Data Center Knowledge, contact:
The powerful Apollo system can reach Edwin Rothrock at 513-322-5616
[email protected]
new heights of performance density with
more than 250 teraflops per rack with
ProLiant XL730f Gen9 Servers. That is up
to 4x the teraflops per sq. ft. and up to 40%
more FLOPs per watt than comparable
air-cooled servers. At the same time, it
helps reduce your carbon footprint, saving
up to 3,800 tons of CO2 per year.

So while you are changing the world with your

research results, you can reduce your impact on the
planet by reducing energy use. You can even take it
one step further and recycle the water to heat your
facilities, and even melt the ice on walkways.
This is the new type of HPC and supercomputing
architecture. One that is capable of true perfor-
mance and scale deliveryall while operating in
an ever-efficiency state. This model is no longer
limited by traditional concepts of technology.
Instead of being restricted by technologyplat-
forms like the Intel-powered HP Apollo infrastruc-
ture help enable new concepts and technologi-
cal innovations.

www.DataCenterKnowledge.com Contact: Edwin Rothrock 513-322-5616 [email protected] 8