Figure 10 shows the reachability after adding random wait mechanism to MD and flooding. The figure clearly shows that MD and flooding are improved significantly in all static, mobile, and on-o_ cases. In the static case, MD and flooding are better than the previous result (Figure 8) and the reachabilities are close to 100%. In the mobile case, MD and flooding maintain the same level of reachability. In the on-off scenario, however, advances on MD and flooding can be clearly identified. These results show that random wait mechanism is effective in reducing collisions and thus improve reachability.
Figure 11(a) shows cumulative distribution of delay for static environment. Contrary to the previous result, MD performs slightly worse than OPP and TPP here. This is because the random wait mechanism incurs a longer data delivery latency. It does not have the long tail as in TPP though. As expected, the curve of flooding shifts to the right as what happened to MD. Figure 11(b) shows also the trend that the curves of MD and flooding shift to the right. But MD performs as well as OPP. TPP is the worst in latency. Figure 11(c) shows similar results to Figure 11(a) except the overall
amount of data received in MD and flooding simulations is higher.
These results show a trade-off between reachability and latency. The cost of a higher reachability is often the longer latency. One may trade-off a little reachability for much shorter latency. Whether to implement random wait depends on the
requirements from the applications. If reachability is more important, turning on
random wait will be a good option. If the latency is more critical, it is more appropriate to turn off the random wait mechanism.
Figure 10. Reachability for static, mobile, and on-off case after adding random wait to MD and Flooding
Figure 11. Latency.
4.5. CONCLUSION
Inspired by the physics of magnets and nails, we propose magnetic diffusion, a simple and yet efficient dissemination mechanism for mission-critical data. MD is able to identify all the shortest paths available from the data source to sink. By transmitting data over the multiple shortest paths, MD performs well in the timeliness and
reliability of data delivery while the overhead and energy consumption of MD is kept low. These properties are confirmed by the simulations. Therefore, we conclude that MD is particularly suitable for sensor network applications in healthcare and
workplace safety. For these applications, the timeliness and reliability of data, and the energy efficiency of the system are all required properties.
五、分析與討論之二 — Fast Forwarding: Suffix Trees for Fast Sensor Data
5.1. INTRODUCTION
On the address-centric Internet, communication nodes are numerically addressed, for example, 140.112.42.220. A source node sends data by the destination node’s address.
Forwarding, also known as the routing table lookup problem, involves how, given the destination address of an incoming data packet, each intermediate router locates a matching entry in the routing table. From the matching entry, the router identifies the network interfaces (or ports) towards the next hops that the data packet should be forwarded further. Efficient algorithms and data structures such as [50] are proposed to speed up the number matching, which lead to the design of very high-speed IP switches today. Motivated to achieve high-speed forwarding for sensor networks, we seek efficient algorithms and data structures to speed up string matching for
data-centric sensor networks.
In data-centric wireless sensor networks [51], nodes are no longer addressed. Data do not carry the destination address. Instead, each sink node sends an explicit interest through the network to draw in a particular type of data. The intermediate nodes in turn disseminate the data based on the data content rather than the destination node’s address. This data-centric style of communication is particularly promising for that it alleviates the effort of node addressing and address reconfiguration in large-scale mobile sensor networks.
Figure. 12. An example of interest matching.
Forwarding in data-centric sensor networks is particularly challenging. It involves matching of the data content, i.e., string-based attributes and values, instead of numeric addresses. This content-based forwarding problem is well studied in the domain of publish-subscribe systems. It is estimated in [52] that the time it takes to
match an incoming data to a particular interest ranges from 10s to 100s milliseconds.
This processing delay is several orders of magnitudes higher than the propagation and transmission delay.
Consider the MicaZ [53] and TinyOS [54] sensor network development platform. The packet size limit is 36 bytes. The wireless radio transmits at 100s kbps. The
transmission delay is thus at the scale of 1s milliseconds. Assume 10 meter radio range and 2x108m/s propagation speed The propagation delay can be found at the scale of 0.01s microseconds. The processing delay is evidently the bottleneck of the per hop forwarding delay. The interest lookup delay is contributed by 2 levels of matching - interest and predicate matching. Each interest may consist of multiple predicates. At the higher level, the system needs to identify, among various interests, a particular interest that matches the incoming data. At the lower level, the system verifies whether a predicate in an interest matches the incoming data. Illustrated in Figure 12 are 3 example interests composed by a number of predicates. Interest 1 is looking for anything related to the nslab group. Interest 2 looks for data about a faculty member whose name is polly, and interest 3 looks for data about all 2nd year and above master students. The incoming data in Figure 12 matches both interests 1 and 2.
Much of the recent work [55][56][57][58] focus on the strategies of structuring interests or content types to enable fast interest matching. Their objective is to reduce the number of predicate matching required. Our work complements these earlier studies in that we focus on improving the efficiency of individual predicate matching.
Predicate matching involves attribute matching and value matching. Attribute matching is essentially an exact string matching problem. One common practice to speed up attribute matching is to fix the bit position of all possible attributes in the data packet as well as the routing table. This method, although simplifies the attribute matching process, will be memory and bandwidth consuming when the number of different data types is high. When there are different sensors to be added to the network, the system will not be easily extendable without changing the packet format and interest table data structure. Value matching is also a string matching problem, when the data type is string. Depending on the operator of the predicate, value matching may require exact or sub-string matching. In essence, the efficiency of predicate is determined by the efficiency of the string matching algorithm used.
For efficient string matching, prior work [7] suggests the use of ternary search tree (TST). It is a string matching algorithm with O(|P|+log(N)) time complexity and O(|S|)
space complexity. P denotes the input string, typically an attribute or value string in the incoming data. S denotes the training word set which concatenates all the strings appear in the entire interest table. N denotes the total number of strings in the training set. To speed up the string matching process further, we propose and evaluate the use of suffix tree (ST), a linear time string matching algorithm that can be easily extended to perform efficient prefix, suffix and substring matching. ST has an O(|P|) time and O(|S|) space complexity [59][60].
Although the large-scale performance of TST and ST’s memory requirement is the same, we find, in real implementations, the amount of memory required by ST is significantly higher than that of TST. This is a serious problem for sensor nodes in which the memory space is very limited. To tackle the problem, we further propose a scheme to optimize the memory consumption for ST.
To observe how the algorithms will perform in practice, we implement TST and ST, as well as a simple hash-based method for comparison. The experimental results show that the computation time is reduced by 29% to confirm a match and 48% to confirm a non-match at best. The memory usage of ST is improved by 24% with the memory optimization scheme.
Our contribution is three-fold: (1) we identify ST for scalable sensor data forwarding, (2) we propose a novel optimization for ST to reduce the space requirement, and (3) we implement four string matching algorithms and evaluate how well they will perform in practice. The remainder of this thesis is organized as follows. The related work is presented next in Section 5.2. We describe then in Section 5.3 the string matching algorithms. Next in Section 5.4, we provide an analytical comparison of the algorithms. In Sections 5.5, 5.6, and 5.7 we detail the experimental setup, results, and our findings on how the algorithms perform in practice.