Topic: SPARC Decentralised SuperComputer (Read 175 times)

In our era of exponential data growth, industry and academia are realizing that they must achieve a fundamental transformation in how operations are run and innovation is achieved. This transformation will require a step change in thinking and an openness that may not have been present before. The key lies in cloud services adoption. Due to a convergence of multiple technologies, including ubiquitous wireless communication, real-time analytics, machine learning, and embedded systems, the commoditization of compute power resources is inevitable. Already, cloud computing has become a highly demanded service or utility due to the advantages of high computing power, cheap cost of services, high performance, scalability, accessibility as well as availability.
As such, enterprises are gradually divesting from their data centers and moving applications workloads to the public cloud. According to the CSA survey report, in 2016, 60.9% of applications workloads were still in enterprise data centers [1]. By the end of 2017, however, fewer than half (46.2%) will remain there. This is, in part, due to new applications primarily being deployed in the cloud, and because enterprises plan to migrate 20.7% of their existing applications to the public cloud [1].

Cloud computing, which includes Software-as-a-Service (SaaS), Infrastructure-as-a-Service (IaaS), and Platform-as-a-Service (PaaS), is projected to increase from $67B in 2015 to $162B in 2020 attaining a Compounded Annual Growth Rate CAGR of 19% as shown in Figure 1 [2][3]. Notably, the International Data Corporation (IDC) forecasts that the High-Performance Computing (HPC) sector will experience a CAGR of 8 percent bringing total market to USD $31.4B by 2019 [4].

Examples of computation-intensive sectors:
• Big Data and Business Analytics (BDA)
Forecasts worldwide revenues for BDA will reach $150.8 billion in 2017, an increase of 12.4% over 2016. Commercial purchases of BDA-related hardware, software, and services are expected to maintain a CAGR of 11.9% through 2020 when revenues will be more than $210 billion.
Science, Technology, Engineering and Mathematics (STEM) researchers and industries
For example, the market for neural network software is projected to grow from USD $7.17B in 2016 to $22.55B by 2021 [5]. Driven by uncategorized and newly digitized data, neural networks are computation-intensive sorting machines.
• Video and Image rendering
Rendering and visualization software is expected to grow at a compounded annual growth rate (CAGR) of 30.03% from 2016 to 2020 [6]. 3D enabled display devices, virtual and augmented reality, and high-end video games are driving this growth.
• Advanced Risk Analytics
Risk analytics has a predicted CAGR of 15.3% bringing the market from USD $17.60B in 2017 to $35.92B by 2022 [7]. Emerging technologies such as artificial intelligence enable institutions to improve underwriting decisions and increase revenues while reducing risk costs.
In summary, emerging technologies and big data analytics are fueling a growing demand for high- performance computing.

SPARC has identified the following opportunities and challenges in the existing market:
1) High growth-rate of HPC requirements. A large market is available for competitive cloud-based computing services.
2) A vast supply of computing resources exists in the form of underutilized personal and commercial computer assets. These resources can be networked to form a distributed IaaS platform providing high-performance computing power to science and industry. This architecture incurs limited additional infrastructure, operation and maintenance costs.
3) Existing centralized IaaS providers such as Google, Amazon and Microsoft incur higher costs to acquire, operate, maintain, update and expand costly infrastructure and dedicated hardware. Higher costs ultimately result in higher pricing; their customers pay for it.
SPARC’s value proposition
SPARC is the first North American globally-distributed IaaS providing consumers with the most cost- effective, flexible, and easily accessible cloud-based high-performance computing. To achieve this, we are
1) Initially leveraging existing demand and infrastructure. The Berkeley Open Infrastructure for Network Computing (BOINC) is a computing platform comprised of 1,128,897 compute nodes. At 17.1 petaFLOPS, it is the third most powerful supercomputer on Earth and supports computing projects in astrophysics, mathematics, and biology, for example. The alpha version of our network connects to the BOINC network and rewards participant nodes with Science Power and Research Coins (SPARC) for computational work performed. Researchers and developers requiring computing power purchase SPARC tokens from an exchange and attach them to their projects. These tokens are distributed to the participant nodes in proportion to work performed. SPARC tokens can then be exchanged directly for computing power from the network or traded for conventional currency on an exchange.
2) Recapturing wasted computing power. We have identified inefficiencies in existing blockchain currencies. For example, in the competition-based process of hashing a digital transaction on the Bitcoin network, approximately 5,000 peta hashes per second are discarded. This amount of wasted compute power is equivalent to approximately 64,000x the combined power of the Top 500 Supercomputers. Instead of hashing to determine proof-of-work, SPARC is compensating participant nodes for actual work performed. We are effectively recapturing wasted computing resources and making them available to science and industry.
3) Varying hardware allocation instead of varying pricing. Whereas Amazon EC2, Google Cloud Compute and other services vary the price as a function of system loads, our network varies the hardware dedicated to your service via an auction process. We allow you to set your price point and let the network compete for available jobs. This ensures 100% network usage, and potentially high rates of power at low cost without added setup for your project.
4) Expanding infrastructure by connecting new compute assets to our global cloud. Our platform is designed to operate across all intelligent devices; smart phones, tablets, laptops and computers connected to the SPARC network generate revenue for their owners by performing computations for user projects. Computer power consumers can also participate as suppliers earning SPARC tokens that can be used to fund their own usage, or be traded for conventional currency on an exchange. This is an example of demand creating supply.
5) SPARC foresees that imminently arriving self-driving automated vehicles, each containing Massively Parallel Processing (MPP) units, when not in operation, will network together forming one of the most powerful distributed supercomputers on Earth. The Internet of Things allows for any device to participate in our distributed computing network, earning SPARC for device owners and providing compute power for users.
6) Due to its globally-distributed architecture, SPARC incurs minimal costs relating to infrastructure, operations, and maintenance. This enables us to provide the most cost-effective and the most powerful computing services in the world.

Conclusion:
SPARC offers a truly disruptive solution to growing global demand for cloud-based, high- performance computing. Compared to centralized IaaS platforms, our network delivers cost-effective, flexible, and powerful computing power. The SPARC network can endlessly expand its infrastructure by connecting underutilized computer assets to the network, and recaptures wasted computing power from proof-of-hash blockchains. Our distributed architecture is sustainable and eco-friendly, promoting full usage of existing hardware instead of commissioning and operating new infrastructure. SPARC envisions a world in which anyone with a laptop and an internet connection will have the power of a supercomputer at their fingertips.