Challenges and Opportunities in VLSI for Systems Dependability

Abstract

This chapter describes the scope, activities, and results of a research program entitled, “Fundamental Technologies for Dependable VLSI Systems (DVLSI for short henceforth)” which began in 2007 and ended in 2015. The program, funded by JST (Japan Science and Technology Agency ) under the CREST (Core Research of Evolutional Science and Technology) initiative, consisted of 11 projects and addressed problems in dependability of electronic systems from various different angles. VLSI is a complex system in its own right and involves a number of potential hazards that arise internally from aging in elements or those that can be caused by external disturbances such as ionizing radiations. Coping with these phenomena has always been a challenge in semiconductor engineering and this program as well. Fabrics (physical structures) robust against threats, bit-error correction methods, and logic-level redundancies have been extensively studied. To go further, challenges of 3-D integration, chip-area (on-chip and across-chip) network, and wireless packaging have been taken on. Exploiting the potential of VLSI in solving problems in systems that call for hard real-time response and/or synchronicity as in robotics and wireless telecommunications has been addressed as new great opportunities for VLSIs. Advanced ways of verification and test for VLSIs have also been dealt with. We will begin this chapter by going over the background of VLSIs for electronic systems and reviewing the necessity of dependability. We will then describe how this multi-project program of CREST DVLSI was formed and conducted. The university-industry collaboration in goal-oriented management efforts is highlighted as essential. A summary of results obtained follows.

1 VLSI in Electronic Systems and Their Dependability

1.1 Pervasiveness of VLSI

The VLSI (Very Large Scale Integration of semiconductor circuits) and software (computer program) are two great enablers of electronic systems, a synonym to modern-day convenience. Personal computers and cell phones, almost indispensable personal items these days, are good examples. Figure 1.1 shows a simplified block diagram of a personal computer. It is seen that VLSI chips such as a microprocessor [1,2,3], and semiconductor memories [4], e.g., RAM (Random Access Memory) and NVM (Nonvolatile Memory) , are the most important parts among others. Important peripheral devices such as HDD (Hard Disk Drive), communications control, and monitoring display have built-in processors as well. The PC (Personal Computer) is a typical general-purpose computer where users run various different application programs. High-performance (Super-) computers are at the highest end of general-purpose computers.

Fig. 1.1

A simplified block diagram of a PC (Personal Computer) to illustrate the use of VLSIs as key components

Figure 1.2 depicts the power train (power generation and transmission) in a hybrid electric-gasoline-engine vehicle which uses a number of ECUs (electronic control units) . Each ECU has at least one microprocessor “embedded” and is thus an electronic system in its own right. The automobile these days is a typical embodiment of embedded computing [5]. A high-end car these days uses as many as 80 microprocessors for various subsystem and module-level control [6]. Actually, the VLSI has provided the biggest momentum to improve the quality and reduce the cost of products or services of electronic systems. This is true with most of complex systems products, which may be mechanical (stationary or mobile), aerodynamic, electrical, electromechanical, electromagnetic, optical, electro-optical, or chemical. Because these systems generally need control for precision and throughput, which is hard to achieve were it not for the VLSI and program control. Automobiles, aircrafts, rockets, robots, chemical plants, utilities, medical devices, ATMs (Automatic Teller Machines), data storages, and agricultural plants of today are good examples of computer-embedded systems. They would not have existed without the VLSI as their key components for smart control. It is almost funny that we are accustomed to call these computer-embedded electronic systems “dedicated systems.” Although the purpose of the system is certainly “dedicated”, for example, to automotive control, computers (microprocessors) have actually found far more general and voluminous applications in embedded control than in “general-purpose” computing by PCs and HPCs (High-Performance Computers) .

Fig. 1.2

Courtesy, Toyota Motor Corporation

Electronic control units in the power train of a hybrid electric and gasoline-engine vehicle to illustrate use of VLSI-powered ECUs (Electronic Control Units).

The more the benefits are drawn out of these systems and the more extensive their uses become over the population, the more heavily the human life depends on them. It is necessary therefore to see to it that these systems are available whenever they are needed. Because the VLSI is at the core of these systems as the workhorse, it is necessary to understand what the VLSI does in electronic systems, what would happen if it fails to function as expected, what could be done to prevent serious failures from happening, and what we can innovate further in realizing more dependable systems technologies. Actually, these are the subjects discussed in this book. (Let us call the systems that use VLSIs as key components “electronic systems” hereafter. The term VLSI system s may be used interchangeably.)

1.2 Necessity of Dependability

Dependability is never a single quality merit of a system. Central to the merit is rather the “performance” or “performance/cost,” in other words, “better fulfillment of the primary purpose” it is intended for. Table 1.1 shows the factors that would affect the decision a user would make in the procurement of a product or service offered in the marketplace. During early stages of market introduction, cost and or performance may be the most influential factors, but as a product category and its market mature, increased attention is paid to dependability for increased social and economic implications, and this is true now with all kinds of electronic systems. These days, dependability of an electronic system is an interest shared among all those concerned: producers, users, and service providers alike.

Table 1.1 Factors affecting the decision-making for procurement of a product or service

The requirements for dependability have been discussed in and among various government regulatory agencies, global/regional/national standards bodies, mission-oriented agencies, industrial associations, and academic societies. Figure 1.3 shows such organizations along with the documents they have published. It will be relevant to refer in particular to IEC 60300 [7] for dependability management, IEC 61508 [8] for functional safety in industrial process measurement, control and automation, and ISO 26262 [9] for the functional safety for road vehicles, since these will be frequently cited throughout this book.

Fig. 1.3

Organizations engaged in regulations, standards, and guidelines for dependability as part of product quality

2 Background and Motivation for the Program

2.1 What VLSI Has Brought About—A Historical Perspective

The VLSI has contributed to the progress in electronic systems in so many ways, which may be summarized as follows.

#1 Great number of devices integrated on a chip

As first observed by Gordon Moore and later named as Moore’s Law that has held up until very recently, the number of transistors integrated on a chip of VLSI silicon has doubled every 18 months [10]. It is interesting to review the progress that the VLSI made following what Gordon Moore predicted [11]. I will not go into that here, however, since there already are abundant references available for this history [12]. It is worthwhile to note here, however, that there is a very solid theoretical background to the scaling down the sizes (other physical parameters and operating voltages as well) of the transistor, the most basic element of VLSI that has underlain its progress [13]. The number of transistors in a microprocessor has actually increased from the mere 2300 of Intel 4004 in 1971 to the billions today [14]. The same is true with memory chips. In no other technologies has it ever been possible to integrate uniformly performing, reliable components the way VLSI has enabled, which has provided the most powerful driving force for the complex electronic systems [15].

#2 Variety of circuit functions realized on silicon

The VLSI rapidly evolved from the early days of chips with a few logic gates into a variety of circuit functions covering arithmetic, logic, memory, analog, and more. Memories include SRAM (Static Random Access Memory) , DRAM (Dynamic Random Access Memory) , ROM (Read-Only Memory), EPROM (Electrically Programmable ROM), EEPROM (Electrically Erasable and Programmable ROM), and Flash Memory [4]. The analog and analog–digital tier of the silicon circuitry is capable of small-signal and high-power amplification, and analog-to-digital and digital-to-analog conversion [16]. A very important type of products of VLSI called FPGA (Field Programmable Gate Array) emerged during the course of the development [17, 18]. Image sensors with billions of pixels have been used in cameras [19]. Micro-Electro-Mechanical (MEMS) is another direction the VLSI has taken to develop [20].

#3 Single-chip implementation of multiple circuit functions

Almost all the circuit functions described in #2 have actually been integrated in chips by now in the form of microprocessors used for personal computers, mobile communication devices, and computer-embedded electric, electronic, and software-controlled systems. The CMOS (Complementary Metal-Oxide Semiconductor), which emerged originally as low-power but low-speed integrated circuit technology, has since been exploited fully to realize all of the logic, memory, and coupled analog–digital functions, taking over the roles played by ECL, TTL and NMOS, and Bi-CMOS (hybrid bipolar and CMOS), realizing the highest density of integration by virtue of low power (virtually no power consumption when idle) inherent in that technology. This history is very well captured in Table 1.2 compiled by Makimoto et al. [21].

Table 1.2 Evolution of CMOS to encompass broader applications over time. CMOS has gradually outperformed other circuit technologies and enabled the integration of various different circuit functions on a single chip of VLSI [21]

#4 Application functions and accelerated processing

During the course of evolution in VLSI, what is now called the ASIC (Application-Specific Integrated Circuit) [22] has evolved. The ASIC contrasts to general-purpose integrated circuits such as standard memories and microprocessors. ASICs with specific system- or subsystem-level functions have often been developed in-house at a systems house, or at a semiconductor house to the order of a systems house, for signal processing in telecom, image-processing applications (routers and switches, data compression, data correction, display control), for example. Some of these application functions that were originally developed for ASICS such as efficient display control, encryption, and decryption for secure data transmission have been integrated in a general-purpose microprocessor. There are other types of VLSIs that evolved into high-performance, dedicated computation to complement microprocessors. In this category are DSP (Digital Signal Processor) [23] and GPU (Graphic Processing Unit ) [24].

#5 Abundance of on-chip resource

The availability of an abundance of circuit resource has been exploited to introduce fault tolerance to the VLSI. The use of redundant bits for error correction was first used in DRAMs and SRAMs, easily accommodating a few defective bits to the effect of salvaging partially defective chips and thus drastically lowering the average memory prices. The introduction of error correction dramatically improved the tolerance of semiconductor memories against radiation-induced soft errors. (Please refer to paragraphs below). The fault-tolerant technology is used in flash memories in a more sophisticated fashion to optimize the memory retention and write–erase endurance. Error-correcting codes and encoding techniques are used to avoid physical interference of charges in the neighboring cells [25, 26]. Recent multiple-processor chips as well as FPGAs are capable of performing redundant concurrent calculation and then having a vote for the correct result to be robust against faults in a part of the chip. Two of most advanced VLSI architectures are shown in Figs. 1.4 and 1.5 for illustrative purposes. Figure 1.4 shows a powerful integration of a multi-core processor and an FPGA which includes security features such as AES (Advanced Encryption Standard) , SHA (Secure Hash Algorithm), and RSA (Rivest–Shamir–Aldeman encryption) [27]. Figure 1.5 is a microprocessor for automotive applications. Security features to support ISO 26262 have been integrated [28].

Fig. 1.4

A functional block diagram of an integration of a multi-core processor and an FPGA. Courtesy, Xilinx Corporation

Fig. 1.5

A functional block diagram of a multiple-core microprocessor for automotive applications. Various safety and security features such as redundancy and access guard are integrated to support ISO 26262 for road vehicles. Courtesy, Renesas Electronics

#6 Stable manufacturing and sourcing

The remarkable progress in the precision manufacturing technology for semiconductors and its rapid proliferation amongst players throughout the world in a competing as well as collaborating business environment has brought about high quality and stability in the sourcing of the VLSI, contributing tremendously to the build, maintenance, and maintenance support of the electric and electronic systems in terms of cost and availability. This has allowed systems houses to use multiple sources to secure procurement of key components.

#7 Distribution of reusable IPs

It has been made possible by the development of commercial practice in the semiconductor industry to distribute the rights to use the whole or parts of the design of an existing VLSI. Commerce of rights to use a semiconductor design (IP, Intellectual Property as it is called) that has proven to work has enabled reuse and helped realize more complicated chips in shorter time and with less cost of development. The last two items (#6 and #7) are a socioeconomic rather than technical phenomenon, which is worth noticing here discussing the impact of VLSI. Figure 1.5 in which a microprocessor IP and an FPGA IP are integrated is a good example.

The progress in VLSI technologies described above has been the contributors to progress in electronic systems, providing ever higher performance at ever lower prices, as well as dependability in compact, integral packages.

3 Threats and Opportunities for the VLSI Systems

Great many ingenuities and tremendous efforts in engineering and associated sciences have been put in to accomplish the colossal tower of VLSI technology as it stands, which has impacted electronic systems with so much socioeconomic momentum.

3.1 Threats Arising from Miniaturization

Suppose the precision printing and other manufacturing technologies continue to progress making the transistor and other device features even smaller, the VLSI engineering will be left with a pile of problems as follows to solve. Engineering has negotiated these problems of generic nature so far, but they will be much tougher to cope with in the future.

#1 Ionizing radiations and electromagnetic interference

There are the issues of various radiations in the environment that causes errors in the VLSI circuits. If a neutron from the outer space hits a VLSI chip, the electronic charges resulting from ionization in the semiconductor could cause errors in the VLSI circuits that could give rise to a system-level failure. This problem will be dealt with in Chap. 3 of this book. Electromagnetic interference is another persistent radiation issue. The voltage change induced by the alternating electromagnetic field generated off-chip (e.g., by an automotive engine igniter) or fed through the power line are a hazard that needs continued attention in the design of the VLSI. This problem will be handled in Chap. 4.

#2 Variations and degradation in device characteristics

The variation in sizes and other parameters of the transistor, which become more pronounced as it is scaled down, leads to variation in transistor characteristics, which in turn could cause deviation in delay times in the circuits. The latter could result in a system failure. This problem is addressed in Chap. 5. There are also multiple, persistent mechanisms that cause degradation in the characteristics of transistors and other components in VLSI over time and/or under the stress of operating voltage/current, temperature, etc. The time-dependent degradation mechanisms are the topic of Chap. 6.

3.2 Threats Arising from Scale and Complexity

Another aspect of problems in VLSI design for dependability is complexity-increasing scale and integration of different functions. A system consists of subsystems and modules with various different characteristics: processor, SRAM, flash memory, analog–digital components in hardware; and commands and sequences in software; some being offered as existing, already-proven parts, and some being newly developed and left to be proven. The complexity arises from the interactions of various objects such as these, consuming the time and human resource to make sure that they work in coordination in practical use cases.

#1 Connectivity

Interconnects and communications between subsystems are sources of system problems. Users of wireless telecommunications often experience loss of connection. Importance of securing minimal connectivity even under disaster conditions has been pointed out. It will be a challenge to mitigate or perhaps eliminate this problem in a wireless system with VLSIs with new functionalities. This is the topic of Chap. 7. Chapter 8 addresses connectivity in electronic systems and handles the challenges of wireless signal interconnects and wireless power supply for VLSI or system-level packaging.

#2 Responsiveness

A response within a certain specified length of time is often required in real-time systems. A soft real-time control is such that a late response is permissible to a certain extent as in the case of ATM as the user can wait for a second or two. A hard real-time control is such that this requirement is critical as in the case of robotics or automatic drive assistance. Meeting with the hard real-time response requirements in robotic applications and assuring synchronicity over the system-to-system handover in wireless applications are examples. This issue is dealt with in Chap. 9.

#3 Malicious attack s

Electronic systems are often the target of malicious attack of hackers who attempt to steal information, disrupt operation, etc., which poses a threat to systems security and reliability. Consideration for security and safety is adding more tasks for the VLSI systems design recently. This issue, which is becoming one of the greatest social concerns, is handled in Chap. 10.

#4 Design errors and test coverage

Complexity has to be dealt with in designing a VLSI system, but it tasks the process of verification, test , and validation of the systems as well. The mere number such as billions of transistors and ten million lines of source codes (operating systems alone) creates complexity, because experience tells that humans make an error in every 100 line of codes. Making certain that the design of an electronic system reflects the requirements specification has increasingly become a challenging task as complexity increases. Test coverage is therefore another important topic, which is undertaken in Chap. 11.

#5 Unknown threats and provisions

No design is perfect, particularly in light of changing threats, changing uses and changing use environments. Requirements specification, even though it will be prepared with utmost care may not be perfect. Unknown threats and provisions are discussed in Chap. 12.

3.3 Opportunities: Changing Markets and Increased Demands for Systems Dependability

Changes in the market environment that happened during the past 10 years are opening up new opportunities for VLSIs. First of all, certain types of electronic system products are receiving increasing requirements in privacy. Personal information stored in PCs, cell phones, or credit cards are prone to criminal plots and malicious attacks. Safety is an utmost requirement in robots in assistance of the handicapped or for hazardous mission in hostile environment. The same is true with automatic driving or drive assistance in road vehicles. Conformity to new safety standards such as described in Sect. 1.1.2 is now a must for the electronic systems design. These changes in markets and growing demands for safety and security pose great opportunities for VLSIs.

3.4 A Summary of Objectives

The threats and opportunities described in this section are mapped out in Fig. 1.6, which shows origins of threats to the dependability of electronic systems in terms of generation of faults and their escalation. Origins of faults are manifold. For example, noise charges generated in the semiconductor (bottom left) by a neutron of cosmic origin may lead to flipping in a logic or memory state, which may give rise to a failure of the system level, resulting in consequences with different levels of severity. Tampering of VLSI may also result in damages of varied severity. Bugs in circuit, logic, or program design could also cause failures to similar effects. Technological challenges therefore lie in the mitigation and containment of the threat by the design and test of VLSI. Opportunities for VLSI lie in realizing new functional features which could facilitate integration and enhance dependability of increasingly more complex systems.

Fig. 1.6

Propagation and containment of threats that could cause systems failure vertical positions of events or bugs are relative and arbitrary

4 The DVLSI Program

4.1 Vision, Scope, and Mission Statement

From what has been discussed in Sect. 1.3, we now arrive at a vision, scope , and mission statement for the DVLSI Program as follows [29]:

• To work on technologies that would help contain the threats against dependability within VLSI.

  New designs for dependability in physical, circuit, logic, and architectural aspects of VLSI will be explored. New methods of verification and test will be pursued as well to complement from a different angle. The VLSI, which has proven to work as most integral, most dependable parts of systems, needs further development to further improve dependability.

• To come up with ideas of new functionalities for VLSI which contribute to enhancing dependability at the system level.

  Systems in their most advanced form today as those used in electronic commerce, public telecommunications, management, robots, sensor networks, or so-called Internet of Things place challenges as described in Sect. 1.3.

• To provide a method for measuring the dependability of systems.

4.2 Program Start and Project Selection

The DVLSI program started with the appointment of the author to Research Supervisor in March 2007. In an arrangement customary to the CREST programs, we had the privilege of having distinguished advisors [30] from industry and academia shown in Table 1.3 join the Program Management to assist the Research Supervisor.

Table 1.3 DVLSI Program Advisors from industry and academia

The first RFP (Request For Proposals) was issued from JST in March 2007, the deadlines for submission set in May. The selection from the submitted proposals was conducted by the VLSI Program Management (Research Supervisor and Advisors), considering the relevance of the proposal from the following perspectives:

• If the proposal has captured essential problem(s) being experienced and/or overarching in practical VLSI design for dependability;

• What original and distinctively competitive ideas are presented to solve the problem(s) raised;

• If a target is set at a challengeable level and described as clearly and hopefully as quantitatively as possible with respect to the state of the art and on-going competing efforts throughout the world;

• What the likelihood of success in terms of PoC (Proof of Concept ) demonstration and expected successive industrial implementation is.

The selection process took a few months after the submission of proposals and was completed by August 2007. The same process of RFP, proposal submission, and project selection was repeated in 2008 and 2009 to finalize the selection. The eleven projects led by the Principal Investigators were awarded with the JST CREST funds over the 3 years between 2007 and 2009 as shown in Table 1.4 [31]. Table 1.5 is the list of Co-Investigators.

Table 1.4 Project subjects and PIs (Principal Investigators) in the DVLSI program

Table 1.5 Project teams consisting of the Principal Investigators and Co-Investigators

During the 3 years of selection process, it was fortunate to have the projects in the DVLSI Program cover key aspects of the problem of dependability rather comprehensively if not exhaustively. The projects address the aspects of functionality, design/verification tools, and test tools in most of the hierarchical layers from the physics, circuit to architecture, as shown in Fig. 1.7. The vertical axis of Fig. 1.7 is the systems hierarchy from the physical layer at the bottom to application at the top. On the horizontal axis are the segments of research products ranging from the design tools, test tools, and concepts in chips/circuits up to proposed solutions for dependable systems. Figure 1.8 is another roughly sketched project portfolio of the Program compiled from the project documents positioning the projects relative to the applications areas envisioned such as aerospace, plant control, transportation, automobiles, robots, information, telecommunications, medical, finance, and consumer appliances.

Fig. 1.7

Areas of technologies that the projects in the DVLSI program have covered in a plane defined by systems hierarchy on the vertical axis, from the physical layer to application, and segments of research products on the horizontal axis, from the design tools, test tools, through concepts in chips/circuits and up to proposed solutions for dependable systems. The names of the PIs heading up the projects are indicated in red

Fig. 1.8

Applications envisioned and approaches taken by the projects in the DVLSI program. The projects with their distinctive research focuses are positioned roughly relative to the broad spectrum of applications that range from aerospace, plant control/utilities/transportation, robot/automobile, information processing, wireless/telecom, finance/medical, to consumer electronics

4.3 Program Management

In view of the object of the CREST framework, in which technology innovations as a result of collaborative efforts within project teams are envisioned, and with ever-accelerating advancement in technology and realization in products taking place worldwide, the program management that consisted of the Research Supervisor and the Advisors adopted the following practice to help the projects effectively carry out the mission.

• Start out and keep interacting with industry to identify/refine the problems and objectives and have shared interest between the Program and industry if that has not been done enough (Actually this often was the case.),

• Come up with methods to solve the issues that compete favorably among similar efforts worldwide,

• Keep specifying and narrowing down possible applications or opportunities of PoC (Proof of Concept) demonstration,

• Keep interacting with industry to enable research results to get the concept proven and exited to the real world,

• Get the ideas patented and standardized.

The relationship between the Program and the outside world was envisioned as depicted in Fig. 1.9. It was always kept in mind to have a vertical (radial in Fig. 1.9), cross-layer interactions happening exchanging ideas and collaborating with each other. In the innermost core are the teams of Projects in the VLSI Program represented by the PIs (Principal Investigators). The layer surrounding the core is the semiconductor industry and EDA (Electronic Design Automation) industry. The semiconductor manufacturer layer is in turn enclosed in the systems industry layer, which is then to provide the products for the service provider industry (and mission-oriented government bodies) in the outer adjacent layer. The outermost space is the consumer or the general public.

Fig. 1.9

The DVLSI program and its intended cross-layer interactions with external partners

The DVLSI Program (center oval) had invited speakers, panelists, and commentators from the external organizations indicated in the outer shells attend the Program meetings to interact with the DVLSI Program. These organizations in effect formed special interest groups shown with elongated ovals in blue with the PIs indicated in red as the primary window of contacts. Some of these interactions have materialized into collaborative technology/product development and implementation. It was intended throughout the Program to have active interactions between the Program and the outside world first to obtain inputs in from, and then to promote exiting the research results back out into, the real world.

People outside the Program were invited from industries and mission-oriented government bodies such as JAXA (Japan Aerospace eXploration Agency ) to participate in discussions and collaborate with the teams throughout the term of the Program. It was intended that those invited form groups of special interest as depicted in long ovals as depicted in Fig. 1.9 with project teams of matching research topics.

5 A Summary of Results

5.1 What Has Been Accomplished

#1 Fundamental study of threats against VLSI dependability and means to mitigate them

There have been many important results obtained in the DVLSI Program out of the fundamental work of studying the nature of “threats” against the dependability of VLSI systems and means to mitigate/cope with them. Detailed account is given by Program researchers in the chapters and sections of Part II in this book, which is entitled, “The VLSI Issues in Systems Dependability.” Much of the fundamental, physical-/circuit-layer research work have been transferred to industry, or being engineered for products.

It is due here to comment that Part II was contributed by many distinguished authors from outside the VLSI Program as well, who participated in the activities of the Program in the interactive way described in Fig. 1.9, and also kindly agreed to write succinct reviews for some of the chapters in PART II to identify the overarching issues and notable engineering efforts that had been made in the relevant area. Readers are referred to the papers in Part II for more elaborate account of the topics.

#2 Systems-/Solution-oriented results

The Program also brought forth several interesting innovative ideas for dependability at the systems and/or solutions layer. These are discussed in chapters of Part III in this book, which is entitled, “Design and Test of VLSI for Systems Dependability.” Many of them have been brought to the stage of demonstration in proof of concept (PoC) experiments, or preliminary implementation by the time of publication of this book. There are continued efforts being made on these proposals to have them implemented in practical systems. A survey conducted by the management of DVLSI Program on its closing in March 2015 said that about a dozen “exit” efforts were being undertaken between the DVLSI project teams and corporations exploiting the ideas and their demonstrations that had resulted from the Program research. It is hoped that we will see them materialized in tangible products and services in the not too distant future.

#3 Measurement of dependability

It was on the agenda for the DVLSI team since the beginning of the Program if it will ever be possible to establish quantitative metric(s) of dependability of a VLSI system. This subject was brought up to group discussions from time to time. However, we were not able to come up with a good result for quantitative metrics. Probably closest we have come to this topic is Fig. 1.10, which shows a Cartesian diagram. The horizontal axis is the robustness of the technologies built-in by design and represented by a product of technology “robustness factors” comprising variability resilience, soft-error resilience, noise immunity, aging resilience, timing/synchronicity robustness, and tamper resilience. The vertical axis shows the thoroughness of verification and test, and comprises of pre-silicon verification, post-silicon test, availability of field-test data, and MTTF information. By diagonally sectioning the Cartesian quadrant, it will be possible to categorize a design into a few different levels of dependability, which could be useful for auditing the design practice for dependability.

Fig. 1.10

Qualitative levels of VLSI systems measured from the robustness of design against threats and the thoroughness of verification and test

Same sort of idea may be used for assessing the dependability at the systems level. In fact, it is attempted in Chap. 2 of this book to describe risk analysis and engineering for dependability [6]. For systems, subsystems, or systems components that have been used for a considerable period of time well into their expected full lifetime with a good record of random failure/fault events archived, it would be possible to assess their dependability in terms of MTTF (Mean Time to Failure), MTTR (Mean Time To Repair), or FIT (Failure In Time), and use this knowledge to assess the dependability of the next generation of product.

Not only the above time measures, but other measurable dependability indexes such as rates of packet loss, bit errors, etc., at systems- or subsystems-level will be considered in the dependability. It is essential that archives of failure events and their analyses are built and made accessible for basic engineering researchers as those from the DVLSI Program. It is hoped that future project teams will be able to more effectively address the subject of dependability by having access to knowledge of actual failures and practice of dependability design in industry.

5.2 Outreach

Since the DVLSI program started in 2007, a project with objectives quite close to that of DVLSI started in Germany in 2012 [32] and then another in the United States [33]. DVLSI program extended invitation for scholars and engineers from outside Japan as well, including those who participated in the German and US programs to attend meetings of DVLSI, the 2012 JST International Symposium on Dependable VLSI Systems [34] and 2nd International Symposium in Dependable VLSI Systems [35], in particular. The DVLSI program had a number of other events of discussions to promote exchanges of ideas between the DVLSI researchers and people from industry and mission-oriented national organizations, e.g., JAXA, in Japan.

5.3 Conclusions

The ideas borne in the DVLSI program to mitigate threats and provide solutions to dependable systems presented in this book are abundant. They may still need more brush up and further engineering, but are believed to form part of foundation for dependable design and test of the VLSI and help improve the dependability of electronic systems of the future.