An efficient priority-driven congestion control algorithm for data center networks

The explosive growth of GPU data centers is driven by the surging demand for machine learning, artificial intelligence, and high-performance computing applications, with the pressure on the power grid intensifying. Meanwhile, as renewable energy becomes more integrated into the grid, the carbon intensity of grid electricity varies at different times. This variability presents an opportunity for data centers to align their workloads with periods of high renewable energy generation and low carbon intensity, thereby reducing their operational carbon emissions. This paper proposes a multi-objective scheduling strategy to minimize carbon emissions and maximize quality of service (QoS) by identifying the Pareto front. The proposed strategy employs an integrated reinforcement learning model to dynamically integrate action strategies for multiple objectives. Results from a case study based on a real-world, large-scale data center illustrate the method’s effectiveness. This effectiveness arises from its dual-objective job scheduling strategy, which simultaneously reduces carbon emissions from electricity consumption and maintains a high quality of service in data centers. The method aims to promote the low-carbon operation of data centers, provide decision support, and facilitate coordination between data centers and power systems in real-world applications.

SummarySharing the network infrastructure, the performance of emerging distributed applications and services in data centers is directly impacted by the network. As these applications are becoming more and more demanding, it is challenging to satisfy their requirements of low latency, high throughput, and low packet loss rate simultaneously. Prior approaches typically resort to flow control or scheduling mechanisms, prioritizing flows according to their demands. However, none of the methods can solely satisfy the various demands of data center applications. Addressing this challenge, we propose tasch, a preference aware flow scheduling mechanism equipped in the software network edge (ie, end‐host networking). This mechanism utilizes multiple separate queues for flows with different preferences, which guarantees low packet delay for latency‐sensitive flows and provides bandwidth guarantees for throughput‐sensitive flows. A coordinating algorithm is presented to share the network resource among multiple queues with pareto‐optimality. tasch is implemented as a thin and plugable kernel module in Linux based hypervisors, which lies between the complicated physical network and tenants VMs. Subsequently, based on the flow traces of real‐world applications, extensive experiments were conducted to verify the effectiveness of network management mechanism.

Cloud computing (CC) could be a massive distributed computing driven by business, during which the services and resources are area unit delivered on request to external consumer via the Web. The distributed computing environment comprises of physical servers, virtual machines, data centers, and load balancers which are appended in an efficient way. With the increasing size of a number of physical servers and utilization of cloud services in data centers (DC), the power consumption is a critical and challenging research problem. Minimizing the operational cost and power in a DC becomes essential for cloud service provider (CSP). To resolve this problem, we introduced a novel approach that leads to nominal operational cost and power consumption in DCs. We propose a multi-objective modified differential evolution algorithm for first placement of virtual machine (VM) in the physical hosts and optimize the power consumption during resource allocation using live migration. The experimental results reveal that our proposed method is significantly better against state-of-the-art techniques in terms of limited power consumption and SLA for any given workload.

TCP is widely used in the communications between servers in data centers. However, TCP has scaling limit that its ability to provide fair sharing at bottleneck link is reduced as the number of competing flows increases. Such limit is common in current data center networks and causes unfairness among TCP flows on a small time scale. The unfairness would add unpredictable delay to short flows, which would affect the user experiences of soft real-time applications. In this paper, we propose a simple and cost-effective method — shrinking MTU — to improve the fairness among TCP flows. Through model analysis and extensive simulations, the fairness on a small time scale is improved and the completion time of short flows is reduced by decreasing MTU size.

Research Problem. Aiming at the problems such as fast update and iteration speed, high development cost, and isolated enterprise data in hydrogen equipment manufacturing industry, a low code programming system based on enterprise data center and business center was proposed and designed. Research Method. This paper designs the data center service and business center architecture of hydrogen equipment manufacturing industry. The data center collects, mines, and integrates massive data into unified standard structured data and stores them. The business center provides architectural support for the establishment of enterprise resource management system. With the data center service as the system data interaction center and the business center as the model to build the system architecture, the hydrogen equipment manufacturing industry resource management system is constructed through the steps of data interaction, process modeling, and UI editing. Research Results. According to the specific requirements of hydrogen equipment manufacturing industry, the business process of hydrogen enterprise resource management system is constructed, and the construction and development of low code system are studied and discussed by integrating data center and business center. Conclusions. Low code is not just a tool to increase productivity, it has the potential to be a catalyst for technology and culture to promote enterprise innovation and business agility, and ERP systems for hydrogen equipment manufacturing based on low coding technology are needed for the industry of the future. This article contributes to the hydrogen equipment manufacturing industry of low code programming system, users can according to their own needs, through the combination of components, quickly generate a complete management system, update system iteration, greatly reduce the development cost, solve the problem of enterprise data isolation, reduce the development cost, and realize the seamless integration of modern software development practice, so as to promote enterprise innovation and business agility.

An algorithm for network and data-aware placement of multi-tier applications in cloud data centers

Multipath protection for data center services in OpenFlow-based software defined elastic optical networks

Cloud computing refers to internet software as a service, hardware and servers in data centers, which run the applications. Cloud computing is highly scalable and provides virtualized resources as a subscription to users. Use of clouds in education provides great opportunity for students to improve achieved learning outcomes, and learners can gain significant advantages in the new learning environment. Cloud adaption can also help to significantly reduce cost of infrastructure, software and human resources. In this article service oriented cloud computing architecture is used to transfer E-learning into the cloud. These architectures cover challenges of e-learning such as scalability, application development, efficient use of resources, saving expense, and security.

Web search constitutes an important class of dataintensive online services in data centers. Optimizing search systems for energy efficiency, timely response and high search quality (i.e., how relevant the returned results are to a search query), however, is very challenging, as a search system involves a distributed architecture with hundreds of thousands of index serving nodes (ISNs) that return searching results to an aggregator through multiple interdependent retrieval stages in a partition-aggregate fashion. In this paper, we discover through experiments two important characteristics that can affect the system performance: (1) response time and energy consumption are greatly impacted by a small fraction of queries with long processing times; (2) the quality contribution of the ISN is independent of the query processing time. Based on our observation, we propose TailCut, which judiciously discards long query executions and enables ISN-aggregator coordination to minimize energy consumption subject to latency and quality constraints. Our experimental results show that TailCut can achieve up to 39% power saving, while satisfying the tail latency and quality constraint.

The migration of computation and storage to cloud data centers has imposed new communication requirements and a need for fundamental changes in infrastructure. The cloud paradigm means that significant network communication will occur between servers in data centers. Furthermore, the migration of virtual machines and containers across physical servers means that in addition to traditional requirements of efficiency, scalability, and easy manageability, future data center networks must also support efficient mobility. To solve the problem, the paper proposes a radical redesign for data center networks, using a new architectural approach that we call DCnet. The DCnet approach changes the addressing and routing schemes at layers 2 and 3 completely. The new addressing mechanism allows an organization to assign addresses that span multiple data centers and permits VM mobility across data centers. Despite radical changes in addressing and routing, the DCnet architecture retains basics from Ethernet and IP, allowing existing hardware building blocks to be reused. The paper uses IPv6 in examples, but the new addressing model and address mobility approach can be applied to IPv4. By assigning each addressable entity a unique ID, and using the ID as the host suffix in an IP address, DCnet allows legacy operating systems and applications to be used without changes. The paper reports a simulation study that uses mininet and SDN to demonstrate the feasibility of the DCnet approach.

The cost effective deployment of applications into cloud has resulted in significant increase of cloud based services. This has in turn led to large number of data centers in delivering such services at scale, offering myriad of user experiences and minimal downtime. Such commitments of providing differentiated services at scale, invites the necessity to manage energy and performance of constituting nodes in data centers without impacting service level agreements (SLAs). Ensuring energy efficiency in these data centers is a major problem in cloud computing. Many optimization policies like workload consolidation, machine placement etc. helps in containing the energy requirement of servers in data centers. In this paper, we introduce a data-driven prognostic neural network based framework that will consider power consumed by all the components in server beyond incoming request load and effectively forecast it at any point in future.

Integration of "Traditional" Data Centre Services with GIS Technology

In this paper we propose a monitoring system of a data center that is able to infer when the data center is getting into an anomalous behavior by analyzing the power consumption at each server and the data center network traffic. The monitoring system is non-intrusive in the sense that there is no need to install software on the data center servers. The monitoring architecture embeds two Elman Recurrent Networks (RNNs) to predict power consumed by each data center component starting from data center network traffic and viceversa. Results obtained along six mounts of experiments, within a data center, show that the architecture is able to classify anomalous system behaviors and normal ones by analyzing the error between the actual values of power consumption and network traffic and the ones inferred by the two RNNs.

Data center containers are regarded as the basic units to build mega data centers. In practice, heterogeneity exists among data center containers, because of technical innovation and vendor diversity. In this paper, we propose uFix, a scalable, flexible and modularized network architecture to interconnect heterogeneous data center containers. The inter-container connection rule in uFix is designed in such a way that it can flexibly scale to a huge number of servers with stable server/switch hardware settings. uFix allows modularized and fault-tolerant routing by completely decoupling inter-container routing from intra-container routing. We implement a software-based uFix stack on the Linux platform. Simulation and experiment results show that uFix enjoys high network capacity, gracefully handles server/switch failures, and brings light-weight CPU overhead onto data center servers.

In order to reduce power and energy costs, giant cloud providers now mix online and batch jobs on the same cluster. Although the co-allocation of such jobs improves machine utilization, it challenges the data center scheduler and workload assignment in terms of quality of service, fault tolerance, and failure recovery, especially for latency critical online services. In this paper, we explore various characteristics of co-allocated online services and batch jobs from a production cluster containing 1.3k servers in Alibaba Cloud. From the trace data, we find the following: 1) For batch jobs with multiple tasks and instances, 50.8% failed tasks wait and halted after a very long time interval when their first and the only one instance fails. This wastes much time and resources as the remaining instances are running for an impossible successful termination. 2) For online services jobs, they are clustered in 25 categories according to their requested CPU, memory, and disk resources. Such clustering can help the co-allocation of online services jobs with batch jobs. 3) Servers are clustered into seven groups by CPU utilization, memory utilization, and their correlations. Machines with a strong correlation between CPU and memory utilization provides an opportunity for job co-allocation and resource utilization estimation. 4) The MTBF (mean time between failures) of instances are in the interval [400, 800] seconds while the average completion time of the 99th percentile is 1003 seconds. We also compare the cumulative distribution functions of jobs and servers and explain the differences and opportunities for workload assignment between them. Our findings and insights presented in this paper can help the community and data center operators better understand the workload characteristics, improve resource utilization, and failure recovery design.