Hewlett Packard Enterprise Performance Optimized Datacenters bring HPC faster to end users

HPC datacenters today are like sports cars. They’re tuned for peak performance and optimized for high speeds. But many datacenters today require costly and time-consuming upgrades if not entire design and construction to cope with increasing power density featured by HPC clusters.

Like owners of a newly-delivered sports car, no HPC customer wants to contend with “some assembly required.” They just want to turn the key and start driving. Fortunately, HPE has since 2008 been proposing factory built modular datacenters that are fully equipped with IT and sent to the customer as fully operating datacenters.

HPE’s containerized solution is called the Performance Optimized Datacenter, or POD. POD with its simplicity and rapid deployment times count for some of the leanest and most attractive containerized solutions in the HPC marketplace today.

Why might you be interested in HPE PODs?

• Lower construction & acquisition costs. PODs can cost up to 45% less than their equivalent bricks and mortar solutions.

• Reduced construction & deployment time. PODs are modular and simple to configure, with only 15 part numbers that can be used to design a whole datacenter. And however they’re ready to be used in customers’ premises, PODs are typically built and shipped to the customer in just a matter of weeks.

• Reduced operating expenses. PODs practically define HPC energy efficiency, with PUE ratios as low as 1.25. (PUE stands for Power Usage Effectiveness, a ratio of a datacenter’s total energy use over its IT equipment’s energy use. The ideal PUE score is a number just above 1).

• Higher equipment density. POD solutions go from 370 U of rack space upwards — to 2200 and beyond. PODs can also fit up to an average 27kW/rack, with a maximum of 45 kW/rack.

• Scalability through repeatable modular design. POD customers like the solution’s ability to scale with their business, to “pay as you grow.”

In July 2014, for instance, Airbus deployed HPE PODs in both Toulouse, France and Hamburg, Germany, boosting the aerospace company’s HPC performance to 1,200 teraflops.

• Visit the HPE POD dedicated webpage

• Contact your  HPE POD experts

Partners or customers interested in learning more about integrating PODs into their IT infrastructure should contact HPE POD experts at POD-emea@hpe.com. 

Upcoming Events:

• For its 8th edition, the famous OSL HPC & Technical Excellence Symposium (TES) will take place at HPC Competency Center in Grenoble, France from March 14th to March 18th 2016. This is a great opportunity for HPE and its partners to be trained on the latest HPC and Big Data technologies and on the most recent Open Source developments dedicated to these activities: a real added value for your customers.

Registration will open mid - January, for any questions related to this major partner event, do not hesitate to contact Marine Terrier (marine.terrier@hpc-pulse.com).

• HPE Technology & Solutions Summit (TSS) – Cannes, February 1-5 2016. More information are available on the event webpage: http://www.hpetssevents.com/
• And many other partnerships in progress.

The massively parallel HP Moonshot ( http://hp.com/moonshot ) solutions now include three new types of cartridges. These cartridges offer a return on investment unmatched in terms of performance , cost , density and power consumption.

The M300 cartridge single-server based on Intel ® Atom CPU 2.4 GHz C2750 / 8 cœurs/64 bits / 20 W with 32 GB of RAM.
- Hearts CPU up to 360 frame is 3240 42U Rack with Hearts.
- This cartridge perfectly meets the needs of calculation integer .

APU M700 Quad cartridge - based server (CPU + GPU). Each server has an AMD Opteron ™ X2150 / 15 W : 4 Cores CPU 64 Ghz bits/1.5 generation Jaguar + 128 cores GPU AMD Radeon HD 8000.
- Up to 720 CPU and Hearts Hearts 23040 GPU chassis or 6480 207 360 + GPU Hearts Rack 42U.
- This cartridge perfmet provide service VDI guy on dedicated servers (HDI : Host Desktop Service) instead of virtual servers.
- It also opens new horizons for computing on GPU.

M800 Quad server cartridge -based DSP . Each server has a Texas Instrument Keystone II 66AK2H : 4 Hearts ARM Cortex -A15 8 C66x + DSP Hearts .
- Up to 720 CPU and 1440 Hearts Hearts DSP chassis or 6480 + 12960 ŒURS DSP 42U rack .
- Ideal for application of signal processing.

Next Moonshot cartridges will be  available with Leap4, first half of 2014, stay tuned ...

Servers for standard calculations compute nodes are not been forgotten : The new SL Series 2500 now allows a high density incorporating the latest Intel Xeon Ivy bridge.. 2U chassis with up to 8 sockets and offering a flexible CPU / GPU on 2U formats. The cost structure is optimized by sharing power supplies and cooling system by air within the chassis . 

Big Data ( http://hp.com/bigdata) now represents a major concern in the reflections of our customers. Again , HP innovates by proposing a solution where data are getting closer compute . The SL 4540 . 3 models are proposed : 1 servers 60 discs, 30 discs 2 servers , 3 servers 15 discs. All in a standard 5U chassis . This is the perfect solution that fit the storage fitting applications like Mongo DB, Vertica but also massively parallel distributed processes as Hadoop , Cloudera and MapR data. These solutions, erasure coding features enables a very competitive cost in giga really optimized for resilience.

The software with the latest version of Cluster Management utilities (CMU) Version 7.2 :

Improved usability of the product, ease of deployment . CMU now deploys a 'smart' integrated GUI leveraging meta- cluster information terminal.

Supports HP Moonshot m100 cartridges, m300 and m700 . Many improvements (speed, relevance, meta- information) in his analysis cmudiff difference engine ( coupled to the opensource pdsh )

Integrating technology diskless " oneSIS " which will probably be adopted in future versions of the product.

• The next meeting of the Club HPC spring ring its 10th Edition. Do not hesitate to contact emmeline.anterrieu @hp.com for more information on this event.

• As every year, HP will be present at Forum TERATEC beginning July 2014

• HP today announced the implementation of a Centre of Excellence around the GPU with NVIDIA (hp.com/go/nvidia-computing-center) as well as a technology center around the latest Intel Xeon technology , Xeon phi and interconnect TrueScale . ( HP.com/go/centerofexcellence)

• HP will be at the end of January sales ANSYS Conference

• End of March , the great gathering of HP technology partners Technical Excellence Symposium (TES) 2014 HPC Competence Center EMEA Grenoble . This is an opportunity for HP and its partners converge on value-added solutions benefiting from the latest technologies. A real added value for our customers. xavier.delore @ hp.com is your contact for this event essential partner .

• And many other partnerships in progress.

Super Computing 2013 was an opportunity for HP to announce the partnership between the National Research Laboratory of Energy of the United States (NREL ) and HP http://www.nrel.gov/ through our Apollo solution rack. Apollo Rack is a solution of HPC high energy performance cooled by warm water . For NREL , the water recovered at the exit of the system ( heated by the heat of the components ) can heat the campus buildings . Consequently, less than 1 PUE is shown . Note that this solution is installed in Denver, where the harsh winters require resilience of seamless service . Beyond PUE performance , criteria flop performance and densities were preferred .

The name of this machine system:  HP Peregrine !

The first phase of 1.2 Petaflop based on Intel Xeon and Xeon E5 technology Phi has passed acceptance in November 2013 and is now in production phase.