Oddlab: fault-tolerant aware load-balancing framework for data center networks

annals of telecommunications - annales des télécommunications ◽

10.1007/s12243-021-00898-0 ◽

2021 ◽

Author(s):

Aymen Hasan Alawadi ◽

Sándor Molnár

Keyword(s):

Data Center ◽

Traffic Engineering ◽

Fault Tolerant ◽

Weak Link ◽

Rooted Tree ◽

Shortest Paths ◽

Data Center Networks ◽

Link Failures ◽

Link Utilization ◽

Network Utilization

AbstractData center networks (DCNs) act as critical infrastructures for emerging technologies. In general, a DCN involves a multi-rooted tree with various shortest paths of equal length from end to end. The DCN fabric must be maintained and monitored to guarantee high availability and better QoS. Traditional traffic engineering (TE) methods frequently reroute large flows based on the shortest and least-congested paths to maintain high service availability. This procedure results in a weak link utilization with frequent packet reordering. Moreover, DCN link failures are typical problems. State-of-the-art approaches address such challenges by modifying the network components (switches or hosts) to discover and avoid broken connections. This study proposes Oddlab (Odds labels), a novel deployable TE method to guarantee the QoS of multi-rooted data center (DC) traffic in symmetric and asymmetric modes. Oddlab creatively builds a heuristic model for efficient flow scheduling and faulty link detection by exclusively using the gathered statistics from the DCN data plane, such as residual bandwidth and the number of installed elephant flows. Besides, the proposed method is implemented in an SDN-based DCN without altering the network components. Our findings indicate that Oddlab can minimize the flow completion time, maximize bisection bandwidth, improve network utilization, and recognize faulty links with sufficient accuracy to improve DC productivity.

Download Full-text

HTPC: heterogeneous traffic-aware partition coding for random packet spraying in data center networks

Journal of Cloud Computing Advances Systems and Applications ◽

10.1186/s13677-021-00248-4 ◽

2021 ◽

Vol 10 (1) ◽

Author(s):

Jiawei Huang ◽

Shiqi Wang ◽

Shuping Li ◽

Shaojun Zou ◽

Jinbin Hu ◽

...

Keyword(s):

Data Center ◽

Large Scale ◽

Network Performance ◽

Rooted Tree ◽

Heterogeneous Traffic ◽

Data Center Networks ◽

Packet Reordering ◽

Traffic Characteristics ◽

Network Utilization ◽

The Impact

AbstractModern data center networks typically adopt multi-rooted tree topologies such leaf-spine and fat-tree to provide high bisection bandwidth. Load balancing is critical to achieve low latency and high throughput. Although the per-packet schemes such as Random Packet Spraying (RPS) can achieve high network utilization and near-optimal tail latency in symmetric topologies, they are prone to cause significant packet reordering and degrade the network performance. Moreover, some coding-based schemes are proposed to alleviate the problem of packet reordering and loss. Unfortunately, these schemes ignore the traffic characteristics of data center network and cannot achieve good network performance. In this paper, we propose a Heterogeneous Traffic-aware Partition Coding named HTPC to eliminate the impact of packet reordering and improve the performance of short and long flows. HTPC smoothly adjusts the number of redundant packets based on the multi-path congestion information and the traffic characteristics so that the tailing probability of short flows and the timeout probability of long flows can be reduced. Through a series of large-scale NS2 simulations, we demonstrate that HTPC reduces average flow completion time by up to 60% compared with the state-of-the-art mechanisms.

Download Full-text