IoT devices must and will be more intelligent in the future. However, due to the lack of benchmarks representative to the diverse IoT applications, there are limited architecture performance studies on IoT devices. This paper presents IoTBench, the first benchmark suite targeting at IoT edge-device applications. This suite includes seven representative programs from three major IoT categories. We investigate the computational demand and execution efficiency of these benchmarks running on a popular IoT device platform. We also analyze and characterize the energy consumption of IoTBench using analytic approaches. Overall, IoTBench establishes a foundation for innovative architecture design of IoT edge devices.

To efficiently support continuous retrieval for continuous media, many approaches based on the striping strategy that is implemented on a multi-disk drive have been proposed. However, the striping strategy only considers the fixed bandwidth of an object. Moreover, the aggregate bandwidth of the multi-disk drive is determined when the object is striped. Furthermore, for some applications, such as self-learning systems, users may quickly browse some familiar objects while they will slowly scan some unfamiliar objects. Therefore, we propose an efficient approach to supporting continuous retrieval with dynamic required bandwidths without reorganizing the whole striped object. From the performance study, we observe that the number of subobjects that must be moved is proportional to the initial required bandwidth and the current required bandwidth.

The goal of dynamic hashing is to design a function and a file structure that allow the address space allocated to the file to be increased and reduced without reorganizing the whole file. We propose a new scheme for dynamic hashing in which the growth of a file occurs at a rate of n+k/n per full expansion, where n is the number of pages of the file and k is a given integer constant which is smaller than n, as compared to a rate of two in linear hashing. Like linear hashing, the proposed scheme (called linear spiral hashing) requires no index; however, the proposed scheme may or may not add one more physical page, instead of always adding one more page in linear hashing, when a split occurs. Therefore, linear spiral hashing can maintain a more stable performance through the file expansions and have much better storage utilization than linear hashing. From our performance analysis, linear spiral hashing can achieve nearly 97 percent storage utilization as compared to 78 percent storage utilization by using linear hashing, which is also verified by a simulation study.

One of the most important challenges in a video-on-demand system is to support interactive browsing functions such as fast forward and fast backward. There are several possible approaches to implementing these functions which require additional resources. Although the segment-sampling strategy proposed by Chen et al. (1995) can support browsing at any desired speed without any additional resource, this strategy only considers the display of a single continuous object at a time. In this paper, we propose a new strategy, called the multi-sampling strategy, which generalizes the approach of the segment-sampling strategy so that it can support continuous display of multiple objects at different display speed rates, simultaneously, without any additional resource.

In this paper, we propose an efficient conflict-resolution approach based on the multi-disk architecture for the insertion/deletion operations on continuous media that are split up into blocks and placed in various locations on the disk, without reorganizing the whole data. When a new subobject is inserted after subobject i, it will be assigned with an identification number (i + 1) and be inserted into a disk in which the retrieval of the new subobject does not conflict with the retrieval of any other subobject, where a conflict means a pair of two consecutive subobjects that are stored in the same disk have to be retrieved simultaneously. However, a new conflict on the same disk may occur since all the identification numbers of subobjects after subobject i are increased by one. Only when such a new conflict occurs, one movement operation is required, so does the case of a deletion operation. Moreover, to reduce those additional movement cost, a deferring approach is proposed at the cost of an additional buffer. In this approach, n data insertions are deferred and stored in a buffer. Then, the system starts to insert those data after an optimal insertion sequence is determined. Based on this approach, two strategies are proposed: the conflict-resolved-first-deferring strategy (the CRFD strategy) and the conflict-resolved-last-deferring strategy (the CRLD strategy). From our performance analysis, we will prove that the CRLD strategy with an insertion sequence according to the ascending order of identification numbers is an optimal strategy based on the proposed deferring approach.