Hybrid transactional/analytical processing (HTAP) would overload database systems. To alleviate performance interference between transactions and analytics, recent research pursues the potential of in-storage processing (ISP) using commodity computational storage devices (CSDs). However, in-storage query processing faces technical challenges in HTAP environments. Continuously updated data versions pose two hurdles: (1) data items keep changing, and (2) finding visible data versions incurs excessive data access in CSDs. Such access patterns dominate the cost of query processing, which may hinder the active deployment of CSDs. This paper addresses the core issues by proposing an a nalyt i c offloa d e ngine (AIDE) that transforms engine-specific query execution logic into vendor-neutral computation through a canonical interface. At the core of AIDE are the canonical representation of vendor-specific data and the separate management of data locators. It enables any CSD to execute vendor-neutral operations on canonical tuples with separate indexes, regardless of host databases. To eliminate excessive data access, we prescreen the indexes before offloading; thus, host-side prescreening can obviate the need for running costly version searching in CSDs and boost analytics. We implemented our prototype for PostgreSQL and MyRocks, demonstrating that AIDE supports efficient ISP for two databases using the same FPGA logic. Evaluation results show that AIDE improves query latency up to 42× on PostgreSQL and 34× on MyRocks.

Cooperative regenerating codes are a kind of erasure codes, which are optimal in terms of minimizing the repair bandwidth. An (n, k, d, r)-cooperative regenerating code has n storage nodes, where k arbitrary nodes are enough to reconstruct original data, and r failed nodes can be repaired cooperatively with the help of d arbitrary surviving nodes. In the regenerating-code framework, there exists a tradeoff between the storage capacity of each node α and the repair bandwidth y, but the problem of specifying the optimal storage-bandwidth tradeoff of the exact-repair cooperative regenerating codes remains open. A key contribution of this paper is that an outer bound on the storage-bandwidth tradeoff of exact-repair linear cooperative regenerating codes is proposed. This result can be regarded as a generalization of the outer bound proposed by Prakash et al., which specifies the optimal tradeoff of exact-repair regenerating codes for the case of d = k = n - 1. The proposed outer bound suggests the (α, γ) pairs that no exact-repair codes can achieve but only functional-repair codes can. By observing the size of the set of such (α, γ) pairs, the performance of the proposed outer bound is evaluated under various parameter settings.

(n, k, d, r)-cooperative regenerating codes are a kind of erasure codes, where r failed nodes can be repaired cooperatively with the help of arbitrary d surviving nodes. In this paper, we consider exact-repair cooperative regenerating codes whose parameters satisfy k = d = n - r. There exists a tradeoff between the storage capacity of each node α and the repair bandwidth γ in regenerating codes, but the optimal storage-bandwidth tradeoff of the exact-repair cooperative regenerating codes has not been fully specified. We propose an outer bound on the storage-bandwidth tradeoff for the case of k = d = n - r. This result can be regarded as a generalization of the outer bound proposed by Prakash et al. that specifies the optimal tradeoff of exact-repair regenerating codes for the case of k = d = n - 1. Although the proposed outer bound is not always tighter than the cutset bound of the functional repair model, in the cases where n is large and r is small, the proposed outer bound suggests the region on the α-γ plane that no exact-repair codes can achieve, but functional-repair codes can.

The storage-bandwidth tradeoff of linear and exact-repair regenerating codes is considered. An outer bound on the storage-bandwidth tradeoff is proposed for arbitrary (n, k, d) case. n, k, and d denote typical code parameters of regenerating codes. The proposed outer bound becomes tighter than other existing outer bounds as k gets closer to n. In addition, for the case of d = n − 1, it is shown that the proposed outer bound asymptotically meets the inner bound proposed by Tian et al., which implies the proposed bound converges to the optimal storage-bandwidth tradeoff of exact-repair linear regenerating codes in high rates (k/n ≅ 1).

Due to considerable increases in user mobility and frame length through aggregation, the wireless channel remains no longer time-invariant during the (aggregated) frame transmission time. However, the existing IEEE 802.11 standards still define the channel estimation to be performed only once at the preamble for coherent OFDM receivers, and the same channel information to be used throughout the entire (aggregated) frame processing. Our experimental results reveal that this baseline channel estimation approach seriously deteriorates the WiFi performance, especially for pedestrian mobile users and the recently adopted frame aggregation scheme. In this paper, we propose Channel-Aware Symbol Error Reduction (ChASER), a new practical channel estimation and tracking scheme for WiFi receivers. ChASER utilizes the re-encoding and re-modulation of the received data symbol to keep up with the wireless channel dynamics at the granularity of OFDM symbols. Our extensive, trace-driven link-level simulation shows significant performance gains over a wide range of channel conditions based on the real wireless channel traces collected by the off-the-shelf WiFi device. In addition, the feasibility of its low-complexity and standard compliance is demonstrated by Microsoft's Software Radio (Sora) prototype implementation and experimentation. To our knowledge, ChASER is the first IEEE 802.11n-compatible channel tracking algorithm since other approaches addressing the time-varying channel conditions over a single (aggregated) frame duration require costly modifications of the IEEE 802.11n standard.

Spectrum sensing is a critical component in cognitive radio. Meanwhile, robust principal component analysis (rPCA) can decompose a matrix into low-rank and sparse matrices. In general, the covariance matrix of a correlated signal is low-rank and the covariance matrix of white noise is diagonal, which can be regarded as sparse. This fact implies that rPCA can be used as a powerful tool for spectrum sensing. A novel spectrum sensing technique which utilizes the characteristics of covariance matrices and rPCA is proposed in this paper. The proposed scheme is also compared to existing schemes based on sample covariance matrices by simulations.

The cognitive radio (CR) system in which multiple secondary users (SU) search for spectrum opportunities generated by the absence of primary users (PU) is considered in this paper. The occupancy of a CR channel is modeled as a Markov chain, and it is assumed that the Markov chain has only two states: idle or busy. Since parameters of the Markov chain are unknown to SUs \emph{a priori} and the states transit independently of the sensing and utilization of SUs, this problem can be considered as a kind of RMAB (restless multi-armed bandit) problem. We propose an efficient channel allocation algorithm for SUs, which is constructed through combination of multiple single-user MAB policies. When a performance of the proposed algorithm is measured by regret which is defined as the total reward difference from the ideal Bayesian policy in which the stationary probability is known to SUs, the order of regret growth of the proposed algorithm seems to be negatively decreasing, giving a better performance than any other existing policy under the 2-state Markov chain case. In order to estimate the performance of the proposed algorithm appropriately, we introduce a new definition of the regret, which uses a belief vector based Bayesian policy as the ideal policy. We observe experimentally that the order of the newly defined regret in the proposed algorithm is similar to logarithmic order under certain conditions.

To reduce temporal redundancy, conventional video coding standards, such as H.264/AVC, use block based motion estimation (ME) and compensation algorithm. However, the coding performance of H.264/AVC significantly drops when characteristics of video contents are changed due to the motion. In this paper, we proposed a pre-filtering algorithm with locally adaptive filter set for the reference frame to minimize the prediction error in the ME process. With image statistics, we adaptively construct a pre-filter set and apply the pre-filter set to the reference frame locally. The proposed algorithm is more efficient in sequences that have high resolution or fast motion. By manipulating the reference frame similar to the current frame, the coding performance is significantly improved. Compared with H.264/AVC, the proposed algorithm achieves bit-rate reduction up to 15.28%.

We intuitively derive DC and AC constraints to model local block complexities using ordered Hadamard transform from pixel based gradient method. For lossless motion estimation (ME), using (1), we obtain the optimized search order in the matching error calculation by descending order of local constraints using sum of these two constraints. LBC(k) = ACSum(k) + |DCMB- DCLB(k)|.

A general discrete minimization algorithm that can be implemented by highly parallel neural networks is developed. It can be applied to the energy functions that can be expressed as arbitrary types of polynomial functions of the state variables. The algorithm can be operated in a synchronous way.