Peningkatan Utilisasi Jaringan Distributed Storage System Menggunakan Kombinasi Server dan Link Load Balancing

<p><em>Distributed Storage System </em>(DSS) memiliki sejumlah perangkat server penyimpanan yang terhubung dengan banyak perangkat <em>switch</em> untuk meningkatkan utilisasi jaringan. DSS harus memperhatikan keseimbangan beban pada sisi server penyimpanan dan trafik data pada semua jalur yang terhubung. Jika beban pada sisi server penyimpanan dan trafik data tidak seimbang, maka dapat menyebabkan <em>bottleneck network</em> yang menurunkan utilisasi jaringan. Kombinasi <em>server</em> dan <em>link</em> <em>load balancing</em> adalah solusi yang tepat untuk menyeimbangkan beban pada sisi server penyimpanan dan trafik data. Penelitian ini mengusulkan metode kombinasi algoritme <em>least connection</em> sebagai metode <em>server</em><em>-load balancing</em> dan algoritme <em>global first fit</em> sebagai metode <em>link</em><em> load balancing</em>. Algoritme <em>global first fit</em> merupakan salah satu dari algoritme <em>load balancing</em> <em>hedera</em> yang bertujuan untuk menyeimbangkan trafik data berukuran besar (10% dari <em>bandwidth</em>), sehingga terhindar dari permasalahan <em>bottleneck network</em>. Algoritme <em>least connection</em> merupakan salah satu algoritme <em>server</em><em> load</em><em> balancing</em> yang menggunakan jumlah total koneksi dari server untuk menentukan prioritas server. Hasil evaluasi kombinasi metode tersebut didapatkan peningkatan pada rata-rata <em>throughput </em>sebesar 77,9% dibanding hasil metode <em>Equal Cost Multi Path </em>(ECMP) dan <em>Round robin </em>(RR). Peningkatan pada rata-rata penggunaan <em>bandwidth</em> sebesar 65,2% dibanding hasil metode ECMP dan RR. Hasil Penggunaan <em>CPU</em> dan <em>memory</em> pada <em>server</em> di metode kombinasi ini juga terjadi penurunan beban <em>CPU</em> sebesar 34,29% dan penurunan beban penggunaan <em>memory</em> sebesar 9,8% dibanding metode ECMP dan RR. Dari hasil evaluasi, penerapan metode kombinasi metode <em>server</em> dan <em>link load balancing</em> berhasil meningkatkan utilisasi jaringan dan juga mengurangi beban server.</p><p> </p><p><em><strong>Abstract</strong></em></p><p class="Judul2"><em>Distributed Storage System (DSS) has a number of storage server devices that are connected to multiple switch devices to increase network utilization. DSS must pay attention to the balance of the load on the storage server side and data traffic on all connected lines. If the load on the storage server side and data traffic is not balanced, it can cause a network bottleneck that reduces network utilization. The combination of server and link-load balancing is the right solution to balance the load on the server side of storage and data traffic. This study proposes a combination of the least connection algorithm as a server-load balancing method and the global first fit algorithm as a link-load balancing method. The global first fit algorithm is one of Hedera's load balancing algorithms which aims to balance large data traffic (10% of bandwidth), so as to avoid network bottleneck problems. Least connection algorithm is one of the server balancing algorithms that uses the total number of connections from the server to determine server priority. The results of the evaluation of the combination of these methods showed an increase in the average throughput of 77.9% compared to the results of the Equal Cost Multi Path (ECMP) and Round robin (RR) methods. The increase in the average bandwidth usage is 65.2% compared to the results of the ECMP and RR methods. The results of CPU and memory usage on the server in this combination method also decreased CPU load by 34.29% and a decrease in memory usage load by 9.8% compared to the ECMP and RR methods. From the evaluation results, the application of a combination of the server method and the link load balancing method has succeeded in increasing network utilization and also reducing server load.</em></p><p><em><strong><br /></strong></em></p>

A New Method for Reconstructing Data on a Single Failure Node in the Distributed Storage System Based on the MSR Code

As a storage method for a distributed storage system, an erasure code can save storage space and repair the data of failed nodes. However, most studies that discuss the repair of fault nodes in the erasure code mode only focus on the condition that the bandwidth of heterogeneous links restricts the repair rate but ignore the condition that the storage node is heterogeneous, the cost of repair traffic in the repair process, and the influence of the failure of secondary nodes on the repair process. An optimal repair strategy based on the minimum storage regenerative (MSR) code and a hybrid genetic algorithm is proposed for single-node fault scenarios to solve the above problems. In this work, the single-node data repair problem is modeled as an optimization problem of an optimal Steiner tree with constraints considering heterogeneous link bandwidth and heterogeneous node processing capacity and takes repair traffic and repair delay as optimization objectives. After that, a hybrid genetic algorithm is designed to solve the problem. The experimental results show that under the same scales used in the MSR code cases, our approach has good robustness and its repair delay decreases by 10% and 55% compared with the conventional tree repair topology and star repair topology, respectively; the repair flow increases by 10% compared with the star topology, and the flow rate of the conventional tree repair topology decreases by 40%.

Minimum-storage regenerating codes resistant to special adversary

Introduction: In order to deal with temporarily unavailable nodes in a distributed storage system, engineers apply special classes of erasure correction codes. These codes allow you to repair a temporarily unavailable node by downloading small amounts of data from the remaining ones. However, this creates safety threats in the presence of an eavesdropper. Purpose: Introducing a new mathematical model in which the eavesdropper has a limited access to all nodes in the system, and developing codes resistant to it. Methods: Information-theoretic arguments, and mixing information symbols with random ones by systematic Reed — Solomon code. Results: We introduced a new mathematical model of an eavesdropper with limited access to all nodes in a distributed storage system. Note that the proposed eavesdropper is passive, being unable to change the accessed data. We found parameters for optimal regenerating codes resistant to such adversary, and provided a technique to ensure necessary resistance. As a result, we obtained the construction of optimal minimum storage regenerating codes resistant to such adversary. Practical relevance: The proposed constructions can provide resistance to adversary while ensuring effective data repair.