1.1 Preface and Motivation
Even to present days, the competitive strength of the electronic industry has been to concentrate upon reducing the cost and more addictive features to software and the hardware performance enhancement. We always try our best to research and develop newly inventing fields to assure we can develop more powerful project to avoid error within product content. In the past, computer technology usually were used to improve hardware performance and made easy-to-use software.
But, applying the great technology to reduce the cost is not a common strategy in business field.
More and more embedded products are equipped with computational power and a little free storage spaces. For such reason, they become more usable and flexible. When multifunction happens, it becomes unreliable in daily life. Since we already have computing power、free storage space device and it is essential for our daily life. We can make good use of these features to deal with some little issue which may consume excessive human resource but lower technique strength.
As a result, we can use distributed computing as the main concept to reduce our maintenance cost.
On the other hand, the embedded systems become increasingly unstable and complex. We must handle it with extreme care. In our experience, normally if the cable modem encounter malfunction in client side and the problem can not be simply solved by manual reset. What we can do is just replace it with a new cable modem, or recall it to factory side and try to find out the problem after return to the manufacture facility. In some cases, one of these cases is power cycle suddenly, the cable modem just lost it's calibration table and digital certificates, so that malfunction occurs. What we need is to re-calibrate it or find a method to preserve it's calibration tables. In this general practice, we can not avoid the increasing cost of maintenance. Because each tuner has their different features. Even if we keep the general calibration data, it can not totally meet the requirement of
１
every tuner. And it is almost impossible to keep each identical calibration data for every cable modem. The digital certificates is a kind of the company property, it's not allowed to store in the client side. Based on the budgeting control, the multiple system operator (MSO) and the factory would not willing to preserve this kind of unique data. They have to create a huge data base 、equip complex hardware and assign human resource to maintain. It means they have to give those extra costs to assure the integrity of the calibration data. The whole process is contrary with the economic benefits. In our point of view, the cable modems have sufficient computing power、storage spaces and the ample band width to maintain the calibration data for each other. Since it is almost essential in our daily life, and cable modems bare be power off. We can restored the calibration data without the presence of every cable modems that hold part of backup data. This concept is similar to a kind of application of Network Coding. The other feature of DOCSIS (Data Over Cable Service Interface Specification) 3.0 is high transmission speed, there is 4 channels bounding for upstream can provide about one hundred Mbps. It can be conducted under the network computing. The whole process of factory data recovery or data distributing can be quickly completed.
1.2 Research Objectives
As we mentioned in the previous section, we try to make all the process could be resolved inside the cable modems, and make sure the cable modem can self-handling each necessary processes without or with the minimum management from the MSO. It is called the CAROL (Cable modem Auto-Recovery Online) system. In this thesis, we focus on applying the concept of Organic Computing and make good use of the advantages of Network Coding. However, a system could be named organic if all of its components and subsystems are well coordinated in a purposeful manner.
Organic structures realize themselves as hierarchically nested processes, and structured to be able to meet upcoming challenges by goal-oriented reactions [1]. The Organic Computing is capable to achieve complex distributed systems with addictive abilities, such as configuration 、 self-organization 、 self-repair 、 self-optimization 、 self-protection [2] or adaptation. So that the
２
embedded systems become more life-like and can satisfy what we need. The embedded systems become free employees with no wadge. Therefore we believe the concept and principles of the Organic Computing are capable to satisfy our demands. Then we could propose a really implementable method which is based on a common system over DOCSIS network to make the cable modem auto recover their unique calibration tables, digital certifications and default settings.
Our goal is to make all of the process happen in client side, and simplified the multiple system operator's workload. So that there will be no necessity to conduct the routine maintenance in low-level system behavior.
1.3 Research Contribution
This thesis will discuss how to make a automatic recovery method in the cable modem. This is a proposed and completed method which includes the provision method for the MSO, mechanism that distributes the factory data, action flows which can retrieve the original data, and how to add customized log for tracing the life cycle of each cable modem. Additionally, we proposed equations as the tips for the developer who can customize the original method into an more efficient method based on their specific and unique demand. And we also provide the test results which were testified by some common tools. Such authorized tools are commonly used in industry. We can easily compare the traditional method and our system by these test reports. There are four merits of this thesis can be summarized as the following parts:
• The factory data is auto recovered and auto distributed
This is the most important part. We design some appropriate rules and patterns for internal behavior in the Hybrid Fiber-Cable network of the cable modems which should deliver some requested fairish functionality for protecting calibration data. To actually elevate the necessary degrees of freedom for adaptive behavior, we will definitely need the self-reconfigurable system for
３
this task. Choosing the recovery data distributing timing is the key of this system. In this thesis, we give a general introduction to this life-like method, and propose a generic calibration data exchanging architecture [3].
• Flexibility and multiple methods of recovery
According to the actual usage of the cable modems over the Hybrid Fiber-Cable network, we design a flexible method to fit any possible case. The life-like system must be prepare an environment for the self-organization between cable modems, whether the amounts of CM is large or small. The method must be designed as which could be easily resumed original status after being stretched or expanded.
As for the recovery plan, we should consider not only automatic but also manual operation.
The cable modems should be able to detect the changes in the surrounding environment, which in accordance with changes in the environment automatically set based on user's demand or behavior.
If the user, for example, the MSO wants to performed restoration manually, we should provide theses selective options for users.
• Less management
This system is a model of CM-CM interaction in which the calibration data processing has been thoroughly integrated into each objects and activities. In other words, this is an active protection system. Whether the activity is distribution or recovery, it should be achieved between the cable modems. This thesis provides a method to make all of the clients can cover each other without prejudice to common operations. After integration of all the cable modems into a group or many independent groups, this specific system becomes increasingly complex. This feature raises a new challenge for designing and implement .
４
• Traceable logs
There should be some logs for history inquiry. As we mentioned before, this system must be flexible. We have to trace the process flow and make some notes, then we can handle any case that we might image or not. For example, if increasing number of cable modems join this system, we can study the operations and then evolve the system by the logs. To achieve this goal, we should apply the second research objective and try to design the log generator. The principle is to make the log generator simple and effective, and integrated into each important point in this system. Thus, it is important that input should be limited and simplified as much as possible. The application generates logs automatically, and the log is used only for research and recording. This demonstrate the need for new methods to evolve and control the behavior of this flexible systems and to create trustworthy systems which adapting to our needs.
５