Page 56 - SAMENA Trends - May-June 2023
P. 56

REGIONAL & MEMBERS UPDATES  SAMENA TRENDS

        Huawei and IEEE Kazakhstan Subsection Jointly Release the HPC Lossless

        Ethernet and AI Fabric Network Technical White Paper


        Huawei   and   the   IEEE   Kazakhstan
        Subsection  have jointly released a  white
        paper that sheds light on lossless Ethernet,
        a  network  solution that  promotes high-
        performance   computing   (HPC)   and
        AI  scenarios  thanks  to  its  unrivaled
        advantages in network performance
        compatibility,  cost-effectiveness  and
        flexibility. The report, titled HPC Lossless
        Ethernet and AI Fabric Network Technical
        White Paper, was launched at the Huawei
        Network Summit, which took place on the
        first  day  of  the  annual  flagship  Huawei
        Middle East & Central Asia (ME&CA) Tech
        Carnival,  which is  taking  place in  Almaty,
        Kazakhstan for the first time, from June 5
        - 7. Under the 'Leading Digital Infrastructure
        for New Value Together' theme, Huawei's
        ME&CA  Tech  Carnival  brings  together   educational institutions and 80 colleges and   application  system.  The  objective  of
        industry opinion leaders  to explore the   universities and support research in fields   network optimization  is to maximize the
        infinite  possibilities  of  digitalization.  In   such as  meteorological prediction and   throughput and minimize the latency of the
        attendance at the launch ceremony of the   modeling, life sciences, fluid mechanics and   entire network, which involves the following
        white paper were  Arthur Wang,  President   oil exploration. Dr. Fahem Al Nuaimi, CEO   technologies:
        of the Data  Center  Network  Domain,   of Ankabut, said, "Ankabut wants  to play   Flow  control:  Identifies  cyclic  buffer
        Huawei  Data  Communication  Product   its role in promoting cutting-edge research   dependencies and eliminates the necessary
        Line,  Dr.  Carlo  Molardi,  IEEE  Kazakhstan   and development  and  collaboration  for   conditions  for generating  them,  resolving
        Subsection,  Liu  Gui,  Vice  President  of   member  institutions  and  the  country.  An   the PFC deadlock problem and enhancing
        Enterprise Business Group, Huawei Middle   advanced network is crucial for this role.   network  reliability.    Congestion  control:
        East  &  Central Asia  Region;  and Fahem   Working  with  partners  like  Huawei,  we   Draws on the AI algorithm to dynamically
        Al Nuaimi, CEO, Ankabut, the United Arab   can create a  future-proof network  that   adjust the explicit congestion notification
        Emirates. Around the world, many countries   helps advance our mission of making UAE   (ECN)  thresholds  for maximized network
        and regions are proactively formulating   universities leaders in research." According   bandwidth  and  minimized  network
        policies conducive to developing HPC and   to the white paper, lossless Ethernet   latency.  Traffic  scheduling:  Adopts  NSLB
        AI  to  promote  digital  transformation  and   technology adopts intelligent remote direct   technology to load balance network-wide
        lossless Ethernet is and will keep playing a   memory  access (RDMA) and  network   traffic,  achieving  90%  ultra-high  network
        key role in these efforts. For example, Saudi   scale  load  balancing  (NSLB),  which  help   throughput  and  improving  AI  training
        Arabia has formulated Saudi Vision 2030,   to achieve zero packet loss and 90% ultra-  efficiency by 20%. To converge and optimize
        aiming to develop HPC and AI technologies   high network throughput. The report begins   the  network and application  system, in-
        to  improve  global  competitiveness.  It   with the HPC network architectures: Clos,   network  computing  (INC)  technology
        also  established the King Abdulaziz  City   Multi-Rail, and directly connected topology   is  used  on  HPC  networks.  Designed
        for  Science  and  Technology  (KACST)  to   (DCT).  Clos  is  a  multistage  architecture,   for  Message  Passing  Interface  (MPI)
        support the R&D of HPC and AI projects.   with each switching unit connected to all   communication,  this  technology  enables
        Qatar also released National Vision 2030 to   switching units at its lower stage. It is strict-  network devices to participate in computing,
        propel HPC and AI technologies. The Qatar   sense  non-blocking,  re-arrangeable,  and   minimizing the job completion time.  HPC
        Computing  Research  Institute  is  Qatar's   scalable. Multi-Rail adopts cell switching   and AI are becoming converged and ready
        foremost research institute focusing  on   of modular devices to implement absolute   for everything,  and Lossless Ethernet  is
        HPC  and  AI  with  the  mission  to  promote   load  balancing.  DCT  features  ultra-  ideal for laying a solid foundation for HPC.
        technological  innovation.  The  UAE  also   large networking, low costs,  and a  small   This  will  help  to  fully  unleash  computing
        established the National AI Strategy 2031   number  of end-to-end  communication   power  across industries  to promote  the
        and appointed the world's first Minister for   hops.  The  paper  then  details  how  the   prosperity of the digital industry and fuel
        AI. Ankabut, UAE's advanced research and   software architecture improves HPC and AI   global digital transformation, underpinning
        education  network (NREN),  has  adopted   application performance from two aspects:   the construction  of a  fully-connected,
        lossless Ethernet  in its  supercomputing   network optimization and convergence   intelligent world.
        center to provide computing services for 35   and optimization of the network and

                                                                                                    56   MAY-JUNE 2023
   51   52   53   54   55   56   57   58   59   60   61