Bilgisayar Mühendisliği Bölümü Koleksiyonu

Permanent URI for this collectionhttps://hdl.handle.net/20.500.11779/1940

Browse

Filters

2015 - 20207

Settings

Has files: NoPublisher: search.filters.publisher.IEEE

Search Results

Now showing 1 - 7 of 7
  • Conference Object
    Fault-Tolerant Strassen-Like Matrix Multiplication
    (IEEE, 2020) Arslan, Şuayb Şefik; Güney, Osman B.; Oblokulov, Muhtasham
    In this study, we propose a simple method for fault-tolerant Strassen-like matrix multiplications. The proposed method is based on using two distinct Strassen-like algorithms instead of replicating a given one. We have realized that using two different algorithms, new check relations arise resulting in more local computations. These local computations are found using computer aided search. To improve performance, special parity (extra) sub-matrix multiplications (PSMMs) are generated (two of them) at the expense of increasing communication/computation cost of the system. Our preliminary results demonstrate that the proposed method outperforms a Strassen-like algorithm with two copies and secures a very close performance to three copy version using only 2 PSMMs, reducing the total number of compute nodes by around 24% i.e., from 21 to 16.
  • Conference Object
    Citation - WoS: 2
    Citation - Scopus: 2
    Distributed Matrix Multiplication With Mds Array Bp-Xor Codes for Scaling Clusters
    (IEEE, 2019-07-01) Arslan, Şuayb Şefik
    This study presents a novel coded computation technique for distributed matrix-matrix product computation at a massive scale that outperforms well known previous strategies in terms of total execution time. Our method achieves this performance by distributing the encoding operation over the cluster (slave) nodes at the expense of increased master-slave communication. The product computation is performed using MDS array Belief Propagation (BP)-decodable codes based on pure XOR operations. In addition, our scheme is configurable and suited for modern compute node architectures equipped with multiple processing units organized in a hierarchical manner. Assuming the number of backup nodes being sublinear in the size of the product, we shall demonstrate that the proposed scheme achieves order-optimal computation from an end-to-end latency perspective while ensuring acceptable communication requirements that can be addressed by today's high speed network link infrastructures.
  • Conference Object
    Citation - Scopus: 4
    Openmp and Posix Threads Implementation of Jerasure 2.0
    (IEEE, 2017-06-01) Arslan, Şuayb Şefik; Le, Hoa; Landman, Joseph; Goker, Turguy
    In shared memory multiprocessor architectures, threads can be used to implement parallelism. POSIX threads (pthreads) is a low-level bare-bones programming interface for working with threads. Therefore, we have extremely fine-grained control over thread management (create/join/etc), mutexes, and so on. On the other hand, openMP, as a shared-memory standard, is much higher level and portable interface which makes it easier to use multi-threading capability and obtain satisfactory performance improvements. Since pthreads is more flexible, it helps programmers gain more control on performance optimizations. Jerasure 2.0 erasure coding library has encoding/decoding engines which comprise independent "for" loop iterations and hence possess huge potential for multi-threaded processing. In this short paper, we investigate multi-threaded implementations of encoder/decoder pair of Jerasure 2.0 using two different technologies: OpenMP and pthreads. We constrain our changes to a minimum possible and compare the pure encoding/decoding performance with respect to each other as well as against that of the original single-threaded version by running them on two different server systems.
  • Conference Object
    Citation - WoS: 3
    Citation - Scopus: 3
    A Joint Dedupe-Fountain Coded Archival Storage
    (IEEE, 2017-05-01) Arslan, Şuayb Şefik; Göker, Turguy; Wideman, Rod
    An erasure-coded archival file storage system is presented using a chunk-based deduplication mechanism and fountain codes for space/time efficient operation. Unlike traditional archival storage, this proposal considers the deduplication operation together with correction coding in order to provide a reliable storage solution. The building blocks of deduplication and fountain coding processes are judiciously interleaved to present two novel ideas, reducing memory footprint with weaker hashing and dealing with the increased collisions using correction coding, and applying unequal error protection to deduplicated chunks for increased availability. The combination of these two novel ideas made the performance of the proposed system stand out. For example, it is shown to outperform one of the replication-based as well as RAID data protection schemes. The proposed system also addresses some of the fundamental challenges of today's low-cost deduplicated data storage systems such as hash collisions, disk bottleneck and RAM overflow problems, securing savings up to 90% regular RAM use.
  • Conference Object
    Corner Detection by Local Zernike Moments
    (IEEE, 2015) Ozbulak, Gokhan; Gökmen, Muhittin
    In this paper, our corner-based interest point detector, Robust Local Zernike Moment based Features (R-LZMF), which was proved to be scale, rotation and translation-invariant, is investigated for invariance against affine transformation, lighting and blurring. Furthermore, R-LZMF's corner detection capability with Zernike moments of order 4 is theoretically explained in detail. Experimental results on the Inria Dataset show that R-LZMF outperforms SIFT, CenSurE, ORB, BRISK and competes with SURF in terms of repeatability for images under affine transformation and photometric deformation such as lighting and blurring.
  • Conference Object
    Citation - WoS: 2
    Citation - Scopus: 5
    Mojette Transform Based Ldpc Erasure Correction Codes for Distributed Storage Systems
    (IEEE, 2017-05-01) Arslan, Şuayb Şefik; Normand, Nicolas; Parrein, Benoit
    Mojette Transform (MT) based erasure correction coding possesses extremely efficient encoding/decoding algorithms and demonstrate promising burst erasure recovery performance. MT codes are based on discrete geometry and provide redundancy through creating projections. Projections are made of smaller data structures called bins and are generated from a two dimensional convex-shaped data. For exact data recovery, only a subset of projections are needed by the decoder. We realize that the discrete geometry definition of MT erasure codes corresponds to creating structured/deterministic generator matrices. In this study, we show an alternative Low Density Parity Check (LDPC) code construction methodology through investigating parity check matrices of MT codes which shows sparseness as the blocklength of the code gets large. In a distributed storage setting, we also quantify the repair bandwidth and show that this novel interpretation can be used to facilitate bin-level local repairs.
  • Conference Object
    Citation - WoS: 1
    Citation - Scopus: 1
    Security Analysis of Revocable and Bipartite Biotokens
    (IEEE, 2015-10-01) Sarıer, Neyire Deniz; Saner, Neyire Deniz
    In this paper, we analyze the security of bipartite biotokens that release a secret key hidden in the biotoken by using biometrics. We show that the biotoken encoding of 80/112/128-bit symmetric encryption keys are vulnerable to brute force attacks, whose complexity is lower than cryptographic security. Also, we present the weaknesses in the design of revocable biotokens that form the basis for bipartite biotokens. Finally, we propose countermeasures to prevent these attacks and discuss the employment of other efficient cryptographic techniques that possess provable security guarantees.