Moore’s Law has ruled the development of computers for over 50 years (actually i

Question

Moore’s Law has ruled the development of computers for over 50 years (actually it has been officially tracked starting 1958).

The trend is evident in High Performance Computing (HPC) field as shown in this diagram on the right from Top500 website.

Notice that in the period depicted in the diagram from 1994 to 2016, there had been dramatic changes in system architecture of super computers. Hardware evolved from vector computers to Massively Parallel Computing to Clusters and now accelerators. In the mean time, parallel computing tools shifted from PVM to MPI and OpenMP. Popular interconnections changed a few times between propriety connectors and fast switches. This diagram is a vivid testimony that Moore’s Law is followed through a combined efforts of many things together   

Top500 website contains a wealth of statistics on these fast computer systems. Browse the website then answer Questions 1-3 using your knowledge in operating systems (Question 4 requires further reading of extra materials).

3) In HPC, researchers chase after extreme speed which takes a well balanced system comprised of solutions of software, hardware and networking to achieve. Conduct individual research and demonstrate your understanding of trade-offs and comprises that must be made in building a balanced system: Assuming that you are asked to provide computational support to a Bioinformatics research group on DNA sequencing. Explain how are you going to set up the computers and network environment for this mission. How would you adjust the plan if you are asked to support a high frequency trading financial firm in New York City with the only goal being to make more profit?

Performance Development 100 TFlaps 10 Tflops IT Flops 100 GFlops 1 Flops 100 MFlops 1996 1998 2000 2002 2004 2006 2010 2012 2014 2016

Explanation / Answer

supercomputer operating systems have undergone major transformations, based on the changes in supercomputer architecture. While early operating systems were custom tailored to each supercomputer to gain speed, the trend has been to move away from in-house operating systems to the adaptation of generic software such as Linux

Since modern massively parallel supercomputers typically separate computations from other services by using multiple types of nodes, they usually run different operating systems on different nodes, e.g. using a small and efficient lightweight kernel such as CNK or CNL on compute nodes, but a larger system such as a Linux-derivative on server and I/O nodes.

While in a traditional multi-user computer system job scheduling is, in effect, a tasking problem for processing and peripheral resources, in a massively parallel system, the job management system needs to manage the allocation of both computational and communication resources, as well as gracefully deal with inevitable hardware failures when tens of thousands of processors are present.[76]

Although most modern supercomputers use the Linux operating system, each manufacturer has its own specific Linux-derivative, and no industry standard exists, partly due to the fact that the differences in hardware architectures require changes to optimize the operating system to each hardware design.

2. Describe improvements that should be made over the current popular OS (your answer to Question 1) in regard to processes and memory management to make it more suitable for HPC over next five years.

Supercomputers are the fastest and most powerful computers available, and at present the term refers to machines with hundreds of thousands of processors. They are the super-stars of the high–performance class of computers.We define high–performance computers as machines with a good balance among the following major elements:

• Multistaged (pipelined) functional units.

• Multiple central processing units (CPUs) (parallel machines).

• Multiple cores.

Reduced instruction set computer (RISC) architecture (also called superscalar) is a design philos-ophy for CPUs developed for high-performance computers and now used broadly. It increases the arithmetic speed of the CPU by decreasing the number of instructions the CPU must follow.

In order to achive exascale bye the early 2020s, a transition to a faster and more general memory interface will be necessary to access bot close and distant memory devices using a unified protocol. This memory protocol will have a notion of near va. far, larse vs. small, and fast ve. slot. It loads or stores domain that can be used to map addresses at large scale, without necessarily trying to maintain coherence using a hardware directory protocol.

While co-packaged memeory will provice the critical growth in bandwidth, greater capacity will have to be supplied through other means. The good news is that several technologies are lining up as psersistent replacement ot DIMMS, with greater capacity, comparble latency, and bandwith. Hewlett Packard Enterprise has been hard at work on memristor for several years now, and the fruits of those efforts are nearly ready to go to market. With such a device, we can now envision a world where the memory becomes the center of the system architecture or, in other words, memory-centric computing. Once attached to a serial memory controller, the persistent memory now lives on the fabric and can be accessed from anywhere. Such memory can then be co-located closer to the processors in a more distributed way, or aggregated in larger pools allowing for the re-unification of storage tiers such as the burst buffer and the parallel file system. The Machine is our strategic vehicle to advance memory-centric computing and other enabling technologies, which aims to integrate standard cores, application-specific cores, memory, management, and fabric all in a single package

4. Remote Direct Memory Access (RDMA) enables efficient memory access from one computer to another. Read about it using Google Scholar, ACM Digital Library or any professional literature tools. Write a 150-200 words (references extra) review on RDMA’s applications and potentials in HPC. Find relevant information in cutting edge research and cite them (minimum 5) properly in IEEE/CS format.

RDMA supports zero-copy networking by enabling the network adapter to transfer data directly to or from application memory, eliminating the need to copy data between application memory and the data buffers in the operating system. Such transfers require no work to be done by CPUs, caches, or context switches, and transfers continue in parallel with other system operations. When an application performs an RDMA Read or Write request, the application data is delivered directly to the network, reducing latency and enabling fast message transfer.

However, this strategy presents several problems related to the fact that the target node is not notified of the completion of the request (single-sided communications).

Much like other high-performance computing (HPC) interconnects, as of 2013 RDMA has achieved limited acceptance due to the need to install a different networking infrastructure. However, new standards such as iWARP enable Ethernet RDMA implementation at the physical layer using TCP/IP as the transport, combining the performance and latency advantages of RDMA with a low-cost, standards-based solution. The RDMA Consortium and the DAT Collaborative have played key roles in the development of RDMA protocols and APIs for consideration by standards groups such as the Internet Engineering Task Force and the Interconnect Software Consortium.

Hardware vendors have started working on higher-capacity RDMA-based network adapters, with rates of 40 Gbit/s reported.Software vendors, such as Red Hat and Oracle Corporation, support these APIs in their latest products,[citation needed] and as of 2013 engineers have started developing network adapters that implement RDMA over Ethernet. Both Red Hat Enterprise Linux and Red Hat Enterprise MRG have support for RDMA. Microsoft supports RDMA in Windows Server 2012 via SMB Direct.

Common RDMA implementations include the Virtual Interface Architecture, RDMA over Converged Ethernet (RoCE), InfiniBand, and iWARP.

Related Questions

Navigate

• Browse (All)
• Browse M
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.