Rather‐high‐frequency sound scattering by swimbladdered fish
Summary (3 min read)
Introduction
- The authors goal is to design broadcast schedules that minimize the waiting time, i.e., the amount of time the client needs to wait in order to obtain the most recent update.
- Fig. 1 depicts an example of a data broadcast system the authors consider.
- Moreover, the idle time can vary for different clients.
- This correspondence focuses on the design of broadcast schedules that minimize the amount of time the client needs to wait (after it is ready) to obtain the desired update.
A. Universal Schedules
- The authors begin by observing that deterministic schedules do not behave well in the presence of an adversarial client.
- Indeed, as shown in Fig. 3(c), such a client misses the beginning of the current update and needs to wait an entire period to receive the next update, resulting in a long waiting time.
- Their goal is to design a random schedule that minimizes the worst case expected waiting time of a client, where the expectation is taken over the probability distribution governing the schedule.
- A natural approach to limit the power of the adversarial client is to bound its adaptivity to the schedule.
- Universal schedules make no assumption on the distribution of clients’ idle times.
C. Our Results
- The authors present (for the restricted adversarial clients discussed above) a universal broadcast schedule that guarantees a worst case expected waiting time of 1= p 2 ' 0:7 time units, regardless of the clients’ access patterns.
- The authors work mainly addresses adversarial clients with limited adaptively to the schedule that can be quantified by one time unit.
- One of the important characteristics of the schedule is the number of updates it sends over a period of time.
- The authors show that, under certain restrictions on the server, this is the best possible schedule.
D. Organization
- The remainder of this correspondence is organized as follows.
- In Section III, the authors focus on clients with unit adaptivity and prove their main results.
- In Section IV, the authors briefly discuss schedules for highly adaptive clients.
- Finally, conclusions are presented in Section V.
A. Random Schedules
- As mentioned in the Introduction, the authors are interested in designing universal schedules for delivering a series of data updates from a single information source over a broadcast channel.
- The length of the interval is chosen without loss of generality, as their techniques (with an appropriate scaling) can be applied for time intervals of an arbitrary length.
- In a random schedule, the interleaving times are random variables.
- The authors define the client’s waiting time as the length of the time interval between t and the beginning of the transmission of the next packet.
B. Adversarial Clients
- The authors goal is to design schedules that perform well for any client, regardless of the viewed history of the schedule.
- This allows to simplify the analysis, without any loss of generality.
- The authors say that an adversarial client is !-adaptive if its actions at time t depend only on the history V 2 V(S; t !) of the schedule S at time t !.
- Note also that the expression sup0 t<!EWT (S; t) bounds the maximum waiting time for requests placed at times t < !, when the client does not have any knowledge of the schedule’s history.
- For a real random variable X , the authors denote by FX(t) = Pr[X < t] its cumulative distribution function, by X = 1 0 (1 FX(x))dx its expectation, and by fX its probability density function (when exists).
C. Transmission Rate
- The transmission rate of a schedule S = fX1;X2; . . .g is defined to be the expected fraction of the time the channel is in use.
- Definition 4 (Transmission Rate r): Let S = fX1;X2; . . .g be a random schedule and let Rt be the expected number of packets sent in S up to time t.
- The transmission rate of S is defined to be r = lim t!1.
III. UNIVERSAL SCHEDULES FOR ! = 1
- The authors study the design of optimal universal schedules for adversarial clients with a degree of adaptivity of one time unit (i.e., ! = 1).
- The authors schedules are defined by a single random variable X .
- That is, all interleaving timesXi in their schedules are independent and have the same distribution as X .
A. Optimal Schedules
- Each random schedule in the family the authors present is associated with a parameter , which is equal to the expected value ofX , i.e., = E[X].
- 2) If event A happens, the client needs to wait X1j A t time units until the transmission of the first packet begins.
C. Optimal Schedules for Large Rates
- This corresponds to the situation where excessive slackness results in a high waiting time.
- Second, notice that the tradeoff curve the authors present has a knee phenomenon.
- That is, increasing the rate beyond 2 1+ p 2 ' 0:82 has little effect on the worst case ex- pected waiting time until r reached a value of approximately 0:95.
IV. HIGHLY ADAPTIVE ADVERSARIAL CLIENTS
- The authors discuss schedules for adversarial clients with very small degree of adaptivity !.
- For this extreme case the authors give a tight analysis of W (S; !).
- Let n be a sufficiently large integer, also known as Theorem 6.
- For a client’s request at time t 1 n , Lemma 1 implies that the expected waiting time is bounded by 1 1 4n (for n large enough).
V. CONCLUSION
- The authors defined the notion of universal schedules that guarantee low waiting time for any client, regardless of its access pattern.
- The authors studied the performance characteristics and the design of universal broadcast schedules, focusing on adversarial clients whose adaptivity is bounded by one time unit.
- Moreover, the authors have shown that this is the best possible schedule.
- For larger values of r the authors have presented a tight analysis of the tradeoff between the transmission rate and the minimum worst case expected waiting time.
- For smaller values of !, these expressions do not differ significantly from those appearing in this work and an analysis of similar nature may be performed.
Did you find this useful? Give us your feedback
Citations
419 citations
384 citations
Cites methods from "Rather‐high‐frequency sound scatter..."
...The average target strength can be calculated according to the swim bladder morphometry of the species and the tilt angle distribution observed (Foote, 1985)....
[...]
257 citations
238 citations
Cites background or methods from "Rather‐high‐frequency sound scatter..."
...Because the presence, structure, and orientation of a swimbladder is species dependent (Jones and Marshall, 1953; Whitehead and Blaxter, 1964; Alexander, 1970), using geometric shapes to model ®sh backscatter inadequately represents asymmetrical swimbladders (Foote, 1985)....
[...]
...Recent models replicate anatomical detail of the swimbladder and body (Foote, 1985; Foote and Traynor, 1988) or generalize animal morphology using combinations of regular shapes such as gas-®lled (Do and Surti, 1990; Clay, 1992) and ¯uid-®lled cylinders (Clay, 1991)....
[...]
160 citations
Cites background from "Rather‐high‐frequency sound scatter..."
...There also exist theoretical models for calculating the backscattered cross section of fish as functions of swimbladder presence, fish size, orientation and frequency (Foote, 1985; Furusawa, 1988)....
[...]
...The average target strength can be calculated according to the swimbladder morphometry of the species and the tilt angle distribution observed (Foote, 1985)....
[...]
References
109 citations
92 citations
48 citations
27 citations
23 citations