Papers by Berksan Serbetci
The work establishes the exact performance limits of stochastic coded caching when users share a ... more The work establishes the exact performance limits of stochastic coded caching when users share a bounded number of cache states, and when the association between users and caches, is random. Under the premise that more balanced user-to-cache associations perform better than unbalanced ones, our work provides a statistical analysis of the average performance of such networks, identifying in closed form, the exact optimal average delivery time. To insightfully capture this delay, we derive easy to compute closed-form analytical bounds that prove tight in the limit of a large number Λ of cache states. In the scenario where delivery involves K users, we conclude that the multiplicative performance deterioration due to randomness – as compared to the well-known deterministic uniform case – can be unbounded and can scale as Θ( logΛ/loglogΛ) at K=Θ(Λ), and that this scaling vanishes when K=Ω(ΛlogΛ). To alleviate this adverse effect of cache-load imbalance, we consider various load balancin...
ArXiv, 2021
This work presents a new way of exploiting nonuniform file popularity in coded caching networks. ... more This work presents a new way of exploiting nonuniform file popularity in coded caching networks. Focusing on a fully-connected fully-interfering wireless setting with multiple cache-enabled transmitters and receivers, we show how nonuniform file popularity can be used very efficiently to accelerate the impact of transmitter-side data redundancy on receiverside coded caching. This approach is motivated by the recent discovery that, under any realistic file-size constraint, having content appear in multiple transmitters can in fact dramatically boost the speed-up factor attributed to coded caching. We formulate an optimization problem that exploits file popularity to optimize the placement of files at the transmitters. We then provide a proof that reduces significantly the variable search space, and propose a new search algorithm that solves the problem at hand. We also prove an analytical performance upper bound, which is in fact met by our algorithm in the regime of many receivers. ...
We investigate the decoding delay performance of a communication network in which a single source... more We investigate the decoding delay performance of a communication network in which a single source is transmitting data packets to a single receiver via multiple routers. Network coding is applied to all data packets at the source at each transmission opportunity. Receiver receives network coded packets from routers and decodes them. We define the delay as the time between arrival of a data packet at the source and decoding of all the packets served in the busy period of the source queue starting from the arrival of that data packet. We show that the delay can be expressed in closed-form.
2020 18th International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOPT), 2020
The work presents a new way of exploiting non-uniform file popularity in caching networks. Focusi... more The work presents a new way of exploiting non-uniform file popularity in caching networks. Focusing on the interference channel with cache-enabled transmitters and receivers, we show how non-uniform file popularity can be used to accelerate the impact of transmitter-side data redundancy in coded caching. This approach is motivated by the recent discovery that under realistic file-size constraints, having content appear in multiple transmitters can boost multiplicatively the speed-up factor attributed to coded caching.We formulate the problem through an optimization algorithm, which seeks to optimize the number of transmitters each file is cached at, as a function of that file’s popularity. Part of the optimization effort involves a biconvex problem; such problems are traditionally solved by heuristic Alternate Convex Search methods that generally do not guarantee the global optimum. To avoid this, we follow a more involved path which includes the design of a new search algorithm tha...
The work establishes the exact fundamental limits of stochastic coded caching when users share a ... more The work establishes the exact fundamental limits of stochastic coded caching when users share a bounded number of cache states, and when the association between users and caches, is random. This association can greatly affect performance, which improves when the association is more balanced across the caches, and which deteriorates when this association becomes less uniform. Our work provides a statistical analysis of the average performance of such networks, quantifying the effect of randomness by identifying in closed-form, the exact optimal average delivery time. To insightfully capture this delay, we derive the exact scaling laws of the optimal average delivery time. In the scenario where delivery involves $K$ users, we conclude that the multiplicative performance deterioration due to randomness - as compared to the well-known deterministic uniform case - can be unbounded and can scale as $\Theta\left(\frac{\log\Lambda}{\log\log\Lambda}\right)$ at $K=\Theta(\Lambda)$, and that ...
The work establishes the exact performance limits of stochastic coded caching when users share a ... more The work establishes the exact performance limits of stochastic coded caching when users share a bounded number of cache states, and when the association between users and caches, is random. Under the premise that more balanced user-to-cache associations perform better than unbalanced ones, our work provides a statistical analysis of the average performance of such networks, identifying in closed form, the exact optimal average delivery time. To insightfully capture this delay, we derive easy to compute closed-form analytical bounds that prove tight in the limit of a large number $\Lambda$ of cache states. In the scenario where delivery involves $K$ users, we conclude that the multiplicative performance deterioration due to randomness -- as compared to the well-known deterministic uniform case -- can be unbounded and can scale as $\Theta\left( \frac{\log \Lambda}{\log \log \Lambda} \right)$ at $K=\Theta\left(\Lambda\right)$, and that this scaling vanishes when $K=\Omega\left(\Lambda...
2017 IEEE Wireless Communications and Networking Conference (WCNC), 2017
In this work we investigate optimal geographical caching in heterogeneous cellular networks where... more In this work we investigate optimal geographical caching in heterogeneous cellular networks where different types of base stations (BSs) have different cache capacities. Users request files from a content library according to a known probability distribution. The performance metric is the total hit probability, which is the probability that a user at an arbitrary location in the plane will find the content that it requires in one of the BSs that it is covered by. We consider the problem of optimally placing content in all BSs jointly. As this problem is not convex, we provide a heuristic scheme by finding the optimal placement policy for one type of base station conditioned on the placement in all other types. We demonstrate that these individual optimization problems are convex and we provide an analytical solution. As an illustration, we find the optimal placement policy of the small base stations (SBSs) depending on the placement policy of the macro base stations (MBSs). We show ...
We investigate optimal geographical caching in heterogeneous cellular networks, where different t... more We investigate optimal geographical caching in heterogeneous cellular networks, where different types of base stations (BSs) have different cache capacities. The content library contains files with different popularities. The performance metric is the total hit probability. The problem of optimally placing content in all BSs jointly is not convex in general. However, we show that when BSs are deployed according to homogeneous Poisson point processes (PPP), independently for each type, we can formulate the problem as a convex problem. We give the optimal solution to the joint problem for the PPP deployment. For the general case, we provide a distributed local optimization algorithm (LOA) that finds the optimal placement policies for different types of BSs. We find the optimal placement policy of the small BSs (SBSs) depending on the placement policy of the macro BSs (MBSs). We show that storing the most popular content in the MBSs is almost optimal if the SBSs are using optimal place...
2019 International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOPT)
We consider caching of video streams in a cellular network in which each base station is equipped... more We consider caching of video streams in a cellular network in which each base station is equipped with a cache. Video streams are partitioned into multiple substreams and the goal is to place substreams in caches such that the residual backhaul load is minimized. We consider two coding mechanisms for the substreams: Layered coding (LC) mechanism and multiple description coding (MDC). We develop a distributed asynchronous algorithm for deciding which files to store in which cache to minimize the residual bandwidth, i.e., the cost for downloading the missing substreams of the user's requested video with a certain video quality from the gateway (i.e., the main server). We show that our algorithm converges rapidly. Finally, we show that MDC partitioning is better than the LC mechanism when the most popular content is stored in caches; however, our algorithm enables to use the LC mechanism as well without any performance loss.
2019 IEEE Information Theory Workshop (ITW)
The work considers the K-user cache-aided sharedlink broadcast channel where each user has access... more The work considers the K-user cache-aided sharedlink broadcast channel where each user has access to exactly z caches of normalized size γ, and where each cache assists exactly z users. For this setting, for two opposing memory regimes, we propose novel caching and coded delivery schemes which maximize the local caching gain, and achieve a coding gain larger than 1+Kγ (users served at a time) despite the fact that the total cache redundancy remains Kγ irrespective of z. Interestingly, when z = K−1 Kγ , the derived optimal coding gain is Kγz + 1, matching the performance of a hypothetical scenario where each user has its own dedicated cache of size zγ.
IEEE Transactions on Communications
The work establishes the exact performance limits of stochastic coded caching when users share a ... more The work establishes the exact performance limits of stochastic coded caching when users share a bounded number of cache states, and when the association between users and caches, is random. Under the premise that more balanced userto-cache associations perform better than unbalanced ones, our work provides a statistical analysis of the average performance of such networks, identifying in closed form, the exact optimal average delivery time. To insightfully capture this delay, we derive easy-to-compute closed-form analytical bounds that prove tight in the limit of a large number Λ of cache states. In the scenario where delivery involves K users, we conclude that the multiplicative performance deterioration due to randomnessas compared to the well-known deterministic uniform case-can be unbounded and can scale as Θ log Λ log log Λ at K = Θ (Λ), and that this scaling vanishes when K = Ω (Λ log Λ). To alleviate this adverse effect of cache-load imbalance, we consider various loadbalancing methods, and show that employing proximity-bounded load balancing with an ability to choose from h neighboring caches, the aforementioned scaling reduces to Θ log(Λ/h) log log(Λ/h) , while when the proximity constraint is removed, the scaling is of a much slower order Θ (log log Λ). The above analysis is extensively validated numerically.
Chapter 3. A Low-Complexity Approach to Distributed Cooperative Caching with Geographic Constrain... more Chapter 3. A Low-Complexity Approach to Distributed Cooperative Caching with Geographic Constraints Chapter 3. A Low-Complexity Approach to Distributed Cooperative Caching with Geographic Constraints Note that the selection ofb (m) =b (m) in (3.25) uniformly at random requires the computation of J K. We use Fisher-Yates shuffle [33] to obtain the random permutation over J files and store the first K many files after shuffling to produce randomb (m) vectors. With Durstenfeld's extended Fisher-Yates shuffle algorithm [30], computational complexity is O(J).
2019 IEEE 20th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Jul 1, 2019
We consider caching of VoD contents in a cellular network in which each base station is equipped ... more We consider caching of VoD contents in a cellular network in which each base station is equipped with a cache. Videos are partitioned into chunks according to a layered coding mechanism and the goal is to place chunks in caches such that the expected utility is maximized. The utility depends on the quality at which a user is requesting a file and the chunks that are available. We impose alpha-fairness across files and qualities. We develop a distributed asynchronous algorithm for deciding which chunks to store in which cache.
ACM SIGMETRICS Performance Evaluation Review
Caching has received a lot of attention in the literature. Here we provide an overview of the wor... more Caching has received a lot of attention in the literature. Here we provide an overview of the work that is most closely related to the current paper. Namely, we survey the works about systems (networks) of caches. Building upon the approximation technique from Dan and Towsley [13], in [40] Rosensweig et al. proposed an approximation technique for a network of caches with general topology. Unfortunately, it is not easy to provide the performance guarantees of that approximation. Using the characteristic time approximation [16] (see also [21, 22]), Che et al. [12] provide a very accurate approximation for the cache networks with tree hierarchy. The characteristic time approximation is intimately related to the TTL-cache systems [18-20]. Then, in [23] Garetto et al. have shown how one can extend and refine the characteristic time approximation technique to describe quite general networks of caches. The recent upsurge in interest in the analysis of cache networks is motivated by two important application domains: Content Delivery Networks (CDN) [6, 9, 30] and Information Centric Networks (ICN) [11, 41, 48].
Finite-length polarization measures for polar coded systems
2012 IEEE 13th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), 2012
ABSTRACT In this paper, polarization performance of polar code generator matrices is analyzed. We... more ABSTRACT In this paper, polarization performance of polar code generator matrices is analyzed. We obtain the Bhattacharyya parameters of polarized channels generated by all possible lower-triangular 3 × 3 generator matrices and evaluate the corresponding polarization behavior of polar coded systems via both asymptotical polarization rate exponents and finite-length polarization distance graphs. A ranking of all lower-triangular 3 × 3 generator matrices is also provided.
Practical polar code construction using generalised generator matrices
IET Communications, 2014
ABSTRACT Polar coding is a recently proposed coding technique that can provably achieve the chann... more ABSTRACT Polar coding is a recently proposed coding technique that can provably achieve the channel capacity. The polar code structure, which is based on the original 2 × 2 generator matrix, polarises the channels, that is, a portion of the channel capacities approach 1, whereas the remaining channel capacities approach 0. Owing to the specific size of this original generator matrix, polar codes can only have code lengths equal to the powers of 2, resulting in inefficiency for codes of practical lengths. In this study, the performance of finite-length polar codes over the binary erasure channel is analysed. A normalised polarisation distance measure is defined and polar codes from different generator matrices showing different amount of polarisation are compared using this measure. Encoding structures for these generalised polar codes are proposed and polarisation performances in both asymptotical and finite-length cases are investigated for generator matrices of size 3 × 3 and 4 × 4. A generalised decoder is also proposed for this generator matrix and its erasure rate is compared with that of the original generator matrix. It is shown that polar codes that have performance similar to the original construction can be constructed and used for a variety of code lengths, not necessarily equal to powers of 2, using generalised generator matrices.
On the selection of generator matrices for polar coded communication systems
2012 20th Signal Processing and Communications Applications Conference (SIU), 2012
In this paper, the effects of the generator matrix selection on the polarization performance of p... more In this paper, the effects of the generator matrix selection on the polarization performance of polar coded communication systems is analyzed. We consider the problem of calculating the Bhattacharyya parameters for generator matrices of size larger than the standard size 2 × 2 and demonstrate that it is not an easy task to obtain the channel capacities in a recursive
Uploads
Papers by Berksan Serbetci