A Fault Tolerant Mechanism Of Functional Language Based On NVRAM

High-performance computers are widely used in various fields of national security and economic. For example, they are widely used in nuclear weapons research, weather forecasting, oil exploration, bio-pharmaceuticals, animation, etc., which greatly promoted the technological level and economic development. Now with the development of high performance computer systems, expanding the scale of high performance computing, high-performance computing challenges facing reliability issues, the E-Class average time to failure of the computer system in the future is only about 30 to 60 minutes, so it's necessary to support more efficient and fault-tolerant technology.Existing fault-tolerant technology as checkpoint technology,the most widely used one. With the increasing size of the system and the system MTBF is increased, the traditional checkpoint technology spending will be close to the average E-class computer system failures time, which makes high-performance computer system unable to provide effective services.The new non- volatile RAM( NVRAM) with features of high-speed access offers the development of fault-tolerant technology new opportunities. NVRAM can replace DRAM as the main memory of a computing system, allowing the application to run in non-volatile memory space, therefore no need to back up data to disk, reducing the cost and difficulty of fault-tolerant technology. N VRAM can also become a storage instead of a disk or storage intermediate layer, the characteristics of the byte-random access and high-speed read can greatly improve the efficiency of fault-tolerant technology. Changes in the architecture will promote the development of a new type of fault-tolerant technology. NVRAM's emergence provide the fault-tolerant technology to a new direction to develope of new fault-tolerant technology. NVRAM is already been used on the new fault-tolerant technology, but due to changes in the underlying hardware features, existing operating systems does not adapt to the characteristics of the NVRAM, unmatching the changes in the new architecture. Therefore to achieve the fault-tolerant technology requires a simple and efficient language model supports from user- level applications to simplify the complexity of the application layer and the physical layer.In summary, the existing high-performance computer systems face the challenge of reliability, the traditional fault-tolerant technology has been unable to meet the needs o f the rapid development of high-performance computing. In support of the rise of new storage components and applications NVRAM, the fault-tolerant technology is expected to achieve a breakthrough, reducing overhead and improving the efficiency of fault-tolerant technology. While the existing NVRAM-based fault-tolerant techniques require extensive modification to the system, a language model is needed to simplify the system to support the simple and efficient fault-tolerant systems.This study is focused on how to achieve simple and efficient fault-tolerant systems by language model supporting. This paper studies the language model which supports fault tolerance, implement the system to support fault-tolerant operation. The main work is in three aspects as follows:1. The fault-tolerant language modelSince the conventional fault-tolerant technology based NVRAM is lack of upper support from the language model, its implementation is more complicated and very difficult to apply. Traditional command-style language model has uncertainty due to its side effects when there is multiple execution, and can not meet the needs of fault tolerance. In this paper, a language model for fault tolerance is designed based on an analysis of the existing language. Combined with the actual needs of fault tolerance, to identify and adjust the features that not support fault tolerance in semantics. Depending on the language model we design a fault-tolerant system with a simple and efficient fault-tolerant language, and analysis the functional languages which is close to fault-tolerant language model and have the advantages of fault-tolerance. In order to design a fault-tolerant system based on functional language.2. Design and implementation of fault-tolerant data management mechanismLanguage model directly reflected in the data structure in the memory and the memory allocation and management of data. What's more, data management has direct impact and effectiveness on fault-tolerant manners. This paper studies and analyzes the data management mecha nisms of functional languages. Through the analysis of its impact on fault-tolerant technology, adjusted to the requirements of fault-tolerant models, to designe a data management mechanisms to to support fault tolerance while achieving specific memory data structured management and data consistency maintenance.3. Design and implementation of fault-tolerant runtime systemLanguage model can not run without the support of the underlying runtime system. To support fault-tolerant language model, we study and analyze the runtime system of functional languages, analyzing of the details of the program execution in-depth, and combined with demand of fault-tolerance to design and adjusted to achieve the support of the upper fault-tolerant language model. Implement a runtime system with fault-tolerant functional language support to achieve fault-tolerant result that can dynamically save state incase power down and resume execution after recovery.