Vice President and General Manager of High Performance Computing, Intel
According to our analyst friends across the board, Intel is on a hot track. The company is in the process of actively transforming the company and go-to-market plan in a direction that takes it from traditional HPC and towards a new technical computing track. One of the key players behind this strategy is Intel’s Raj Hazra, who appears to be Intel’s technical computing rising star. Already he’s made waves in 2014 with the hiring of former Penguin Computing CEO, Charles Wuischpard as the Vice President of the Data Center Group and General Manager of Workstations and Technical Computing. It will be interesting to see what else Hazra has up his sleeve for 2014. We caught up with him to ask about his thoughts on the year and the direction of supercomputing.
HPCwire: Raj, you have your fingerprints on so much going on at Intel that is relevant to the HPC industry at large. What are some of the most important initiatives coming from Intel that you expect to have an impact on the HPC space in 2014 and beyond?
Raj Hazra: Let me actually start by highlighting some of the key challenges that we, in the HPC industry, are facing today and then tie those into the actions we are taking to address them.
First, while Moore’s law has enabled us to increase the number of transistors we can fit on a given die, thereby increasing the number of cores in our processors, most applications haven’t evolved at the same pace to take advantage of this exponential growth in computing power. To address this challenge, we will – in collaboration with the community – visibly expand our investments through additional Intel Parallel Computing Centers (IPCC) focused on “modernizing” applications. We’ve received strong feedback from end users that the IPCC program helps fill a unique role in preparing for the manycore machines of the future – by augmenting their capabilities with technical assistance, access to hardware, and resources to close other critical resource gaps. IPCCs are the tip of a very large iceberg. We’re also working with open standards bodies on programming models, with OEMs on 3-way collaborations with their key customers, with ecosystem partners for tools, compilers, and middleware that improve developer efficiency, and many other activities.
Secondly, it has been well documented that only a fraction of the SMB’s worldwide – who would benefit from HPC – actually have access to HPC technology. To address this challenge, Intel will collaborate with partners to train and mentor companies – like SMB manufacturers – to help bring the social and economic benefits of HPC to these communities. The Ann Arbor, Michigan-based UberCloud Experiment, awarded by HPCwire at SC 13, serves as a good example of these efforts and early results of our broader initiative can be seen online through this link: https://vimeo.com/69181845, pw: #intel
HPCwire: Intel knows better than anyone that Big Data is transforming the landscape in some pretty remarkable ways. Can you discuss what these changes will mean for the high performance computing landscape now and into the future?
Raj Hazra: The system building blocks for Big Data solutions are all amongst those used in long-term proven solutions found in our diverse HPC market segment. Most notably are those that provide scalable I/O at all levels, computation at various density and scale to match the needs, and reliability in both capability and capacity class systems. In other words, we see Big Data as a new domain of technical computing that will greatly leverage the work of the existing domains, yielding rapidly developed and deployed solutions.
We primarily see this as a further maturing of the HPC landscape, where additional value will be placed in non-traditional spaces, especially those that aren’t addressed by single-facet, CPU-centric, and other non-holistic benchmarks such has HP-LINPACK, SpecFPRate, etc. Given the myriad big data solutions stacks, developing a meaningful solutions metric to help guide systems decisions is an immediate and pertinent task for this emerging HPC community.
The challenge will remain that one solution does not fit all. So, Intel’s strategy will continue to be that of providing the comprehensive requisite building blocks and total solutions portfolios to address the uniqueness of any given challenge.
HPCwire: As you look down the road at 2014 and beyond, what rends do you see developing that are worth taking note of?
Raj Hazra: There are a few worth consideration:
- Heterogeneous computing becomes the next step, not a detour – not just in high-end, but mainstream HPC. The combination of the Xeon processor, Xeon Phi co-processors, and the common programming model supported by them will expand its impact across of all HPC by delivering the desired performance/W requirements, while, at the same time, enabling developers to maintain programming flexibility, performance and efficiency.
- Big Data transforms HPC; First real business value propositions from use of analytics beyond niche using HPC. HPC will expand beyond the existing, well-established frontiers to tackle new challenges (workloads) such as big data and enable the transformation of data into high-value knowledge.
- Domain optimized HPC Clouds – the promise of HPC with the value of the cloud model. The powerful economic forces of the cloud model along with its lack of ability to address traditional HPC workload requirements will accelerate the development of domain optimized HPC cloud solutions, thereby transforming industry models for capital planning, software development & delivery, security, scalability, and redundancy.
- Integration – driving fundamental changes in system architecture: Integration of the right features into silicon will reduce power, reduce costs by eliminating chips, increase performance by having components closer together, increase scalability by being able to abstract multiple integrated features under one set of libraries, and enable unprecedented level of innovation by providing very dense computing solutions. Intel, with its leadership in process technology, is in a unique position to drive this change.
- Everyone wants their own – renewed interest across the geos in not just investing in “having” the best of class, but also in “making” that best of class: While each of the geos have their own motivations, the common goal across them is to not just field world class systems, but innovate and develop their own homegrown technologies (CPU, fabric, software) to power these world class systems – with HPC being the targeted beachhead segment. Interesting times ahead.
HPCwire: On a personal note, can you talk about your personal life? Your family, background, any hobbies?
Raj Hazra: I have a Spouse and 2 girls (22 and 14). I love the northwest – mountains, oceans and outdoor activities. For hobbies I enjoy cooking and doodling with music of all kinds.
HPCwire: One last question – is there anything about yourself that you can share that you think your colleagues would be surprised to learn?
Raj Hazra: I play in a local band – as a pianist and a singer.