We introduce a framework and optimal ‘fresh’ caching for a content distribution network (CDN) comprising a front-end local cache and a back-end database. The data content is dynamically updated at a back-end database and end-users are interested in the most-recent version of that content. We formulate the average cost minimization problem that captures the system’s cost due to the service of aging content as well as the regular cache update cost. We consider the cost minimization problem from two individual perspectives based on the available information to either side of the CDN: the back-end database perspective and the front-end local cache perspective. For the back-end database, the instantaneous version of content is observable but the exact demand is not. Caching decisions made by the back-end database are termed ‘push-based caching.’ For the front-end local cache, the age of content version in the cache is not observable, yet the instantaneous demand is. Caching decisions made by the front-end local cache are termed ‘pull-based caching.’ Our investigations reveal which type of information, updates, or demand dynamic, is of higher value towards achieving the minimum cost based on other network parameters including content popularity, update rate, and demand intensity.

In this work, we consider the problem of ‘fresh’ caching at distributed (front-end) local caches of content that is subject to ‘dynamic’ updates at the (back-end) database. We first provide new models and analyses of the average operational cost of a network of distributed edge-caches that utilizes wireless multicast to refresh aging content. We attack the problems of what to cache in each edge-cache and how to split the incoming demand amongst them (also called “load-splitting” in the rest of the paper) in order to minimize the operational cost. While the general form of the problem comes with an NP-hard Knapsack structure, we were able to completely solve the problem by judiciously choosing the number of edge-caches to be deployed over the network This reduces the complex problem to a solvable special case. Interestingly, our findings reveal that the optimal caching policy necessitates unequal load-splitting over the edge-caches even when all conditions are symmetric. Moreover, we find that edge-caches with higher load will generally cache fewer but relatively more popular content. We further investigate the tradeoffs between cost reduction and cache savings when employing equal and optimal load-splitting solutions for demand with Zipf($z$) popularity distribution. Our analysis reveals that equal load-splitting to edge-caches achieves close-to-optimal for less predictable demand ($z< 2$) while also saving in the cache size. On the other hand, for more predictable demand ($z>2$), optimal load-splitting results in substantial cost gains while decreasing the cache occupancy.

Due to the advances in machine learning techniques, recommender systems nowadays are capable of learning and influencing the users’ decisions. Hence, recommendations became an important facility to reduce the cost (or increase the profit) of the operators of the demand networks. In this paper we formulate and study the problem of dynamically optimizing the demand shaping, through content recommendation, and proactive caching. The formulated problem suffers from the curse of dimensionality, so we devise an approximate algorithm optimizing only over a short look-ahead window. The approximate problem is not convex, as such we utilize non-convex optimization techniques to tackle the problem. To verify the efficiency of our proposed solution, we establish a lower bound on the minimum achievable cost and contrast it with our solution.

Existing content caching mechanisms are predominantly geared towards easy-access to content that is static once created. However, numerous applications, such as news and dynamic sources with time-varying states, generate ‘dynamic’ content where new updates replace previous versions. This motivates us in this work to study the freshness-driven caching algorithm for dynamic content, which accounts for the changing nature of data content. In particular, we provide new models and analyses of the average operational cost both for the single and distributed edge caching scenarios. In both scenarios, we characterize the performance of the optimal solution and develop algorithms to select the content and the update rate that the user(s) must employ to have low-cost access to fresh content. Moreover, our work reveals new and easy-to-calculate key metrics for quantifying the caching value of dynamic content in terms of their refresh rates, popularity, number of users in the distribute edge caching group, and the fetching and update costs associated with the optimal decisions. We compare the proposed freshness-driven caching strategies with benchmark caching strategies like cache the most popular content. Results demonstrate that freshness-driven caching strategies considerably enhance the utilization of the edge caches with possibly orders-of-magnitude cost reduction. Furthermore, our investigations reveal that the distributed edge caching scenario, benefiting from the multicasting property of wireless service to update the cached content, can be cost-effective compared to the single edge caching, as the number of edge caches increases.

We introduce a framework and provably-efficient schemes for ‘fresh’ caching at the (front-end) local cache of content that is subject to ‘dynamic’ updates at the (back-end) database. We start by formulating the hard-cache-constrained problem for this setting, which quickly becomes intractable due to the limited cache. To bypass this challenge, we first propose a flexible time-based-eviction model to derive the average system cost function that measures the system’s cost due to the service of aging content in addition to the regular cache miss cost. Next, we solve the cache-unconstrained case, which reveals how the refresh dynamics and popularity of content affect optimal caching. Then, we extend our approach to a soft-cache-constrained version, where we can guarantee that the cache use is limited with arbitrarily high probability. The corresponding solution reveals the interesting insight that ‘whether to cache an item or not in the local cache?’ depends primarily on its popularity level and channel reliability, whereas ‘how long the cached item should be held in the cache before eviction?’ depends primarily on its refresh rate. Moreover, we investigate the cost-cache saving trade-offs and prove that substantial cache gains can be obtained while also asymptotically achieving the minimum cost as the database size grows.

We introduce a framework and provably-efficient schemes for `fresh' caching at the (front-end) local cache of content that is subject to `dynamic' updates at the (back-end) database. We start by formulating the hard-cache-constrained problem for this setting, which quickly becomes intractable due to the limited cache. To bypass this challenge, we first propose a flexible time-based-eviction model to derive the average system cost function that measures the system's cost due to the service of aging content in addition to the regular cache miss cost. Next, we solve the cache-unconstrained case, which reveals how the refresh dynamics and popularity of content affect the optimal caching. Then, we extend our approach to a soft-cache-constrained version, where we can guarantee that the cache use is limited with arbitrarily high probability. The corresponding solution reveals the interesting insight that `whether to cache an item or not in the local cache?' depends primarily on its popularity level, whereas `how long the cached item should be held in the cache before eviction?' depends primarily on its refresh rate. Moreover, we investigate the cost-cache saving tradeoffs and prove that substantial cache gains can be obtained while also asymptotically achieving the minimum cost as the database size grows.

In this work, we consider the problem of ‘fresh’ caching at distributed (front-end) local caches of content that is subject to ‘dynamic’ updates at the (back-end) database. We first provide new models and analyses of the average operational cost of a network of distributed edge-caches that utilizes wireless multicast to refresh aging content. We attack the problems of what to cache in each edge-cache and how to split the incoming demand amongst them (also called "loadsplitting" in the rest of the paper) in order to minimize the operational cost. While the general form of the problem comes with an NP-hard Knapsack structure, we were able to completely solve the problem by judiciously choosing the number of edge-caches to be deployed over the network. Interestingly, our findings reveal that the optimal caching policy necessitates unequal load-splitting over the edge-caches even when all conditions are symmetric. Moreover, we find that edge- caches with higher load will generally cache fewer but relatively more popular content. We further investigate the tradeoffs between cost reduction and cache savings when employing equal and optimal load-splitting solutions for demand with Zipf(z) popularity distribution. Our analysis reveals that equal load-splitting to edge-caches achieves close-to-optimal for less predictable demand (z2), optimal load-splitting results in substantial cost gains while decreasing the cache occupancy.

In this work, we provide a comprehensive analysis of stability properties and delay gains that wireless multicasting capabilities, as opposed to more traditional unicast transmissions, can provide for content distribution in mobile networks. In particular, we propose a model and characterize the average queue-length (and hence average delay) performance of unicasting and various multicasting strategies for serving a dynamic user population at the wireless edge. First, we show that optimized static randomized multicasting (we call it `blind multicasting') leads to stable-everywhere operation irrespective of the network loading factor (given by the ratio of the demand rate to the service rate) and the content popularity distribution. In contrast, traditional unicasting suffers from unstable operation when the loading factor approaches one, although it outperforms blind multicasting at small loading factor levels. This motivates us to study `work-conserving multicast' policies next that always outperform unicasting while still offering stable-everywhere operation. Then, in the worst-case of uniformly-distributed content popularity, we explicitly characterize the scaling of the average queue-length (and hence delay) under a first-come-first-serve multicast strategy as a function of the database size and the loading factor. Consequently, this work provides the fundamental limits, as well as the guidelines, for the design and performance analysis of efficient multicasting strategies for wireless content distribution.

In this work, we provide a comprehensive analysis of stability properties and delay gains that wireless multicasting capabilities, as opposed to more traditional unicast transmissions, can provide for content distribution in mobile networks. In particular, we propose a model and characterize the average queue-length (and hence average delay) performance of unicasting and various multicasting strategies for serving a dynamic user population at the wireless edge. First, we show that optimized static randomized multicasting (we call it `blind multicasting') leads to stable-everywhere operation irrespective of the network loading factor (given by the ratio of the demand rate to the service rate) and the content popularity distribution. In contrast, traditional unicasting suffers from unstable operation when the loading factor approaches one, although it outperforms blind multicasting at small loading factor levels. This motivates us to study `work-conserving multicast' policies next that always outperform unicasting while still offering stable-everywhere operation. Then, in the worst-case of uniformly-distributed content popularity, we explicitly characterize the scaling of the average queue-length (and hence delay) under a first-come-first-serve multicast strategy as a function of the database size and the loading factor. Consequently, this work provides the fundamental limits, as well as the guidelines, for the design and performance analysis of efficient multicasting strategies for wireless content distribution.

The dominant portion of smartphone traffic is generated by apps that involve human interactivity. Particularly, when human users receive information from a server, they spend a few seconds of information processing before taking an action. The user processing time creates an idle communication period during the app session. Moreover, the generation of the future traffic depends on the service of the current query-response pair. In this paper, we aim at leveraging the properties of such interactions to reap quality-of-experience gains. Existing schedulers, both in practice and theory, are not designed in view of the aforementioned traffic characteristics. Theoretical works predominantly focus on scheduling of traffic that is either generated independently or directly controlled, but not governed by the specific dynamics caused by human interactions. Schedulers in practice, on the other hand, employ round-robin and processor-sharing methods to serve multiple ongoing sessions. We show that neither of these approaches is effective for serving apps that involve human interactivity. Instead, we show that optimal scheduling for interactive traffic is non-randomized over packets, which we call action-based, as it avoids breaking ongoing service of actions in order to align human response times with the service of other actions. Since the design of optimal action-based policy is computationally prohibitive, we develop low-complexity suboptimal action-based policies that are optimal for two ongoing sessions. Our numerical studies based on a real-data trace reveal that our proposed action-based policies can reduce total delay by 22% with respect to packet-based equal processor sharing.

We investigate dynamic content caching in relay networks where an intermediate relay station (RS) can adaptively cache data content based on their varying popularity. With the objective of minimizing the time average cost of content delivery, we formulate and study the problem of optimal RS cache allocation when the popularities of data content are unknown apriori to the network. While optimal dynamic cache control suffers the curse of dimensionality, we develop a fundamental lower bound on the achievable cost by any caching policy. Inspired by the structure of such lower bound, we develop a reduced-complexity policy that is shown numerically to perform close to the lower bound.

We investigate the performance of dynamic proactive caching in relay networks where an intermediate relay station caches content for potential future use by end users. A central base station proactively controls the cache allocation such that cached content remains fresh for consumption for a limited number of time slots called proactive service window. With uncertain user demand over multiple data items and dynamically changing wireless links, we consider the optimal allocation of relay stations cache to minimize the time average expected service cost. We characterize a fundamental lower bound on the cost achieved by any proactive caching policy. Then we develop an asymptotically optimal caching policy that attains the lower bound as the proactive caching window size grows. Our analytical findings are supported with numerical simulations to demonstrate the efficiency of the proposed relay-caching.

In this paper we characterize the proactive diversity gain of a cognitive network with predictable primary and secondary requests. Network performance is analyzed under two proposed proactive service policies that preserve higher priority for the primary user. The first policy preserves the primary diversity bound as if there is no secondary user in the network, whereas the second policy boosts the secondary diversity with guaranteed higher primary diversity. For each policy, we derive diversity gain bounds for primary and secondary users. We show that the predictability of secondary requests can remarkably boost quality of service (QoS) of the secondary user compared to the previous literature when secondary requests are nonpredictable. We provide numerical simulations to validate our analytical findings and demonstrate performance merits.

Many popular smartphone apps involve human interaction through the entire session; e.g. apps for web browsing, making reservations and online gaming. In this work, we characterize bi-directional interactive app traffic in the timescale of seconds, that is shaped by the human interaction. We collect and analyze a dataset comprising 1500 interactive app sessions. The combined uplink and downlink traffic bursts are the outcome of user-server interactions, which we label as actions. Within each action, we discover high correlation between the number of uplink and downlink packets reaching 0.98. Our study reveals that the distribution of action duration and interarrival can be approximated with exponential or gamma distributions. The analysis provides insights on the temporal characteristics of bi-directional packet bursts associated with actions, during an app session. We show that action-based service at access points (APs), where actions constitute the service units rather than packets, can reduce service delay by 50%.

Mobile data users are known to possess predictable characteristics both in their interests and activity patterns. Yet, their service is predominantly performed, especially at the wireless edges, “reactively” at the time of request, typically when the network is under heavy traffic load. This strategy incurs excessive costs to the service providers to sustain on-time (or delay-intolerant) delivery of data content, while their resources are left underutilized during the light-loaded hours. This motivates us in this work to study the problem of optimal “proactive” caching whereby, future delay-intolerant data demands can be served within a given prediction window ahead of their actual time-of-arrival to minimize service costs. To that end, we first establish fundamental bounds on the minimum possible cost achievable by any proactive policy, as a function of the prediction uncertainties. These bounds provide interesting insights on the impact of uncertainty on the maximum achievable proactive gains. We then propose specific proactive caching strategies, both for uniform and fluctuating demand patterns, that are asymptotically-optimal in the limit as the prediction window size grows while the prediction uncertainties remain fixed. We further establish the exponential convergence rate characteristics of our proposed solutions to the optimal, revealing close-to-optimal performance characteristics of our designs even with small prediction windows. Also, proactive design is contrasted with its reactive and delay-tolerant counter-parts to obtain interesting results on the unavoidable costs of uncertainty and the potentially remarkable gains of proactive operation.

In this work, we formulate and study the profit maximization problem for a wireless service provider (SP) that encounters time-varying, yet partially predictable, demand characteristics. The disparate demand levels throughout the course of the day yield excessive service cost in the peak hour that substantially hurts the reaped profit. With the SP's ability to track and statistically predict future requests of its users, we propose to enable proactive caching of the peak hour demand ahead during off-peak times. Thus, network traffic will be smoothed out, while end-users' activity patterns are undisturbed. In addition, the SP is able to assign personalized pricing policies that strike the best balance between enhancing the certainty about the future demand for optimal proactive caching and maximizing the revenue collected from end-users. Comparing the proposed system's performance to the baseline scenario of the existing practice of no-proactive service, we show that the SP attains profit gain that grows with number of users, at least, as the first derivative of the cost function. Moreover, end-users that receive proactive caching services make strictly positive savings. Thus, we essentially demonstrate the win-win situation to be reaped through the exploitation of the consistent users' activity.

Multiple transmitting antennas can considerably increase the downlink spectral efficiency by beamforming to multiple users at the same time. However, multiuser beamforming requires channel state information (CSI) at the transmitter, which leads to training overhead and reduces overall achievable spectral efficiency. In this paper, we propose and analyze a sequential beamforming strategy that utilizes full-duplex base station to implement downlink data transmission concurrently with CSI acquisition via in-band closed or open loop training. Our results demonstrate that full-duplex capability can improve the spectral efficiency of uni-directional traffic, by leveraging it to reduce the control overhead of CSI estimation. In moderate SNR regimes, we analytically derive tight approximations for the optimal training duration and characterize the associated respective spectral efficiency. We further characterize the enhanced multiplexing gain performance in the high SNR regime. In both regimes, the performance of the proposed full-duplex strategy is compared with the half-duplex counterpart to quantify spectral efficiency improvement. With experimental data and 3-D channel model from 3GPP, in a 1.4 MHz 8 × 8 system LTE system with the block length of 500 symbols, the proposed strategy attains a spectral efficiency improvement of 130% and 8% with closed and open loop training, respectively.

In this paper, we propose and study optimal proactive resource allocation and demand shaping for data networks. Motivated by the recent findings on the predictability of human behavior patterns in data networks, and the emergence of highly capable handheld devices, our design aims to smooth out the network traffic over time and minimize the data delivery costs. Our framework utilizes proactive data services as well as smart content recommendation schemes for shaping the demand. Proactive data services take place during the off-peak hours based on a statistical prediction of a demand profile for each user, whereas smart content recommendation assigns modified valuations to data items so as to render the users' demand less uncertain. Hence, our recommendation scheme aims to boost the performance of proactive services within the allowed flexibility of user requirements. We conduct theoretical performance analysis that quantifies the leveraged cost reduction through the proposed framework. We show that the cost reduction scales at the same rate as the cost function scales with the number of users. Furthermore, we prove that demand shaping through smart recommendation strictly reduces the incurred cost even below that of proactive downloads without recommendation.

The global adoption of smart phones has raised major concerns about a potential surge in the wireless traffic due to the excessive demand on multimedia services. This ever increasing demand is projected to cause significant congestions and degrade the quality of service for network users. In this paper, we develop a proactive caching framework that utilizes the predictability of the mobile user behavior to offload predictable traffic through the WiFi networks ahead of time. First, we formulate the proactive scheduling problem with the objective of maximizing the user-content hit ratio subject to constrains stemming from the user behavioral models. Second, we propose a quadratic-complexity (in the number of slots per day) greedy, yet, high performance heuristic algorithm that pinpoints the best download slot for each content item to attain maximal hit ratio. We confirm the merits of the proposed scheme based on the traces of a real dataset leveraging a large number of smart phone users who consistently utilized our framework for two months.

In this work, we investigate the profit maximization problem for wireless network carriers and payment minimization for end users. Motivated by our recent findings on proactive resource allocation, we focus on the scenario whereby end users harness predictable demand and WiFi connectivity in proactive data downloads, to minimize their expected payments. Carriers, on the other hand, utilize smart pricing schemes to differentiate between the off-peak and peak hour prices so as to reduce peak costs and maximize their profit.We formulate the tension between the carrier and end user as a two-player Stackelberg game in which the carrier assigns prices first, then the end user responds with optimized proactive downloads. We explore the equilibrium points under maximum and average price constraints, and study the impact of WiFi availability on the system's performance. In particular, we compare the new equilibria with the baseline scenario of flat pricing and no proactive downloads. Despite the potential uncertainty about future demand, and the freshness of proactively downloaded content, we characterize new equilibria points that yield win-win situation with respect to the baseline equilibrium.

We address the question of optimal proactive service and demand shaping for content distribution in data networks through smart pricing. We develop a proactive download scheme that utilizes the probabilistic predictability of the human demand by proactively serving potential users' future requests during the off-peak times. Thus, it smooths-out the network traffic and minimizes the time average cost of service. Moreover, we incorporate the varying economic responsiveness and demand flexibilities of users into our model to develop a demand shaping mechanism that further improves the gains of proactive downloads. To that end, we propose a model that captures the uncertainty about the users' demand as well as their responsiveness to the pricing employed by the service providers. We propose a joint proactive resource allocation and demand shaping scheme based on nonconvex optimization algorithms, and show that it always leads to strictly better performance over its proactive counterpart without demand shaping.

We study the bounds and means of optimal caching in overlay Content Distribution Networks (CDN) that serve data with dynamic content to end-users who send random requests for the most up-to-date version of such content. Applications with such dynamic content are numerous, including daily news, weather conditions, stock market prices, social networking messages, etc. The service for such a dynamically changing content necessitates a fundamentally different approach than traditional pull-based (also called non-proactive) schemes. In particular, proactive caching is required to optimize the type and amount of content to be updated in the local servers of a CDN hence minimize the transmission and caching costs, subject to storage constraints. We study the metric of cost reduction achieved by proactive caching over non-proactive caching strategies. We introduce the notion of popularity to establish fundamental upper and lower bounds on cost reduction under different degrees of storage space constraints. We prove the lower bounds to achieve the optimal rate of increase achieved by the upper bounds as the database of items increases. In particular, for a general form of convex, superlinear and monotonically increasing cost functions, our results reveal that the optimal cost reduction scales as the cost function itself, or at least as its first derivative, depending on the number of popular data items, as well as the cache storage capacity.

This paper introduces the novel concept of proactive resource allocation for wireless networks, through which the predictability of user behavior is exploited to balance the wireless traffic over time, and significantly reduces the bandwidth required to achieve a given blocking/outage probability. We start with a simple model in which smart wireless devices are assumed to predict the arrival of new requests and submit them to the network$T$time slots in advance. Using tools from large deviation theory, we quantify the resulting prediction diversity gain to establish that the decay rate of the outage event probabilities increases with the prediction duration$T$. Remarkably, we also show that, in the cognitive networking scenario, the appropriate use of proactive resource allocation by primary users improves the diversity gain of the secondary network at no cost in the primary network diversity. We also shed light on multicasting with predictable demands and show that proactive multicast networks can achieve a significantly higher diversity gain that scales superlinearly with$T$. Finally, we conclude by a discussion of the new research questions posed under the umbrella of the proposed proactive wireless resource framework.

In a recent work, we have introduced the notion of proactive resource allocation in wireless networks whereby the predictability of user demands are leveraged to significantly enhance the spectral efficiency of the network in outage limited regimes. In this paper, we expand the horizon to the important scenario of multicast traffic. Our analysis reveals two additional types of gains that can be leveraged in this proactive multicast scenario. The first can be attributed to the basic nature of multicast traffic in which each request would represent a data source rather than a user, as it would in the unicast case. The second is the demand alignment phenomenon whereby the predictive network would wait to gather as much requests as possible and serve them altogether using the same resources. We analytically derive the impact of these advantages on the system diversity gain, which quantifies the exponential decay rate of the outage probability, and further illustrate the resulting gains via numerical results.

We investigate the problem of admission and power control considering a scenario where licensed, or primary, users and cognitive radios, or secondary users, are transmitting concurrently over the same band. The primary users share a common receiver and the interference on this receiver from secondary users should be strictly limited to a certain level. Each secondary link is assumed to have a minimum quality of service (QoS) requirement that should be satisfied together with the interference limit constraint, otherwise the secondary link is not admitted. Under those constraints, admission and power control for secondary users are investigated for two main optimization objectives. First, we maximize the number of admitted secondary links. Second, we maximize the sum throughput of the admitted secondary links. The first problem is NP-hard, hence we provide a distributed close-to-optimal solution based on local measurements at each user and a limited amount of signaling. For the second problem, which is non-convex, we propose a suboptimal algorithm based on sequential geometric programming. The proposed algorithms are compared with previously related work to demonstrate their relative efficiency in terms of outage probability, complexity and achievable throughput.

We investigate power allocation for users in a shared spectrum network. In such a network, the primary (licensed) users communicate under a minimum guaranteed quality of service (QoS) requirements, whereas the secondary users opportunistically access the primary band. Our objective is to find a power control scheme that determines the transmit power for both primary and secondary users so that the overall network throughput is maximized while maintaining the quality of service of the primary users greater than a specified minimum limit. In the assumed model, no interference cancellation is done at the receivers resulting in a non-convex optimization problem. It has been shown previously that binary power control almost always achieves the global optimum solution when no QoS constraints are imposed. This is not necessarily the case in our scenario, however. We introduce a distributed algorithm for "ternary" power allocation to be used when individual measurements are available at each node. We show via simulations the relative efficiency of the proposed algorithm compared to previously suggested ones. If a central controller exists with available information about the system parameters, we enhance the performance of the proposed algorithm through an iterative geometric programming (GP) algorithm and prove its convergence to a better solution than ternary power allocation.