Ibrahim Takouna

Elastic VM for Cloud Resources Provisioning Optimization

Rapid growth of E-Business and frequent changes in websites contents as well as customers’ interest make it difficult to predict workload surge. To maintain a good quality of service (QoS), system administrators must provision enough resources to cope with workload fluctuations considering that resources over-provisioning reduces business profits while under-provisioning degrades performance. In this paper, we present elastic system architecture for dynamic resources management and applications optimization in virtualized environment. In our architecture, we have implemented three controllers for CPU, Memory, and Application. These controllers run in parallel to guarantee efficient resources allocation and optimize application performance on co-hosted VMs dynamically. We evaluated our architecture with extensive experiments and several setups; the results show that considering online optimization of application, with dynamic CPU and Memory allocation, can reduce service level objectives (SLOs) violation and maintain application performance…

Elastic Virtual Machine for Fine-Grained Cloud Resource Provisioning

Elasticity is one of the distinguishing characteristics associated with Cloud computing emergence. It enables cloud resources to auto-scale to cope with workload demand. Multi-instances horizontal scaling is the common scalability architecture in Cloud; however, its current implementation is coarse-grained, while it considers Virtual Machine (VM) as a scaling unit, this implies additional scaling-out overhead and limits it to specific applications. To overcome these limitations, we propose Elastic VM as a fine-grained vertical scaling architecture. Our results proved that Elastic VM architecture implies less consumption of resources, mitigates Service Level Objectives (SLOs) violation, and avoids scaling-up overhead. Furthermore, it scales broader range of applications including databases.

Accurate Mutlicore Processor Power Models for Power-Aware Resource Management

Power management is one of the biggest challenges facing current data centers. As processors consume the dominant amount of power in computer systems, power management of multicore processors is extremely significant. An efficient power model that accurately predict the power consumption of a processor is required to develop effective power management techniques. However, this challenge rises with using virtualization and increasing number of cores in the processors. In this paper, we analyze power consumption of a multicore processor, we develop three statistical CPU-Power models based on the number of active cores and average running frequency using a multiple liner regression. Our models are built upon a virtualized server. The models are validated statistically and experimentally. Statistically, our models cover 97\% of system variations. Furthermore, we test our models with different workloads and three benchmarks. The results show that our models achieve better performance compared to the recently proposed model for power management in virtualized environments. Our models provide highly accurate predictions for un-sampled combinations of frequency and cores, 95\% of the predicted values have less than 7\% error. Thus, we can integrate these models into power management mechanisms for a dynamic configuration of a virtual machine in terms of the number of its virtual-CPUs and the frequency of physical cores to achieve both performance and power constrains.

Communication-Aware and Energy-Efficient Scheduling for Parallel Applications in Virtualized Data Centers

In this paper, we propose Peer VMs Aggregation (PVA) to enable dynamic discovery of communication patterns and reschedule VMs based on the determined communication patterns using VM migration. In the implementation, we consider that communication delays occur at the server (i.e., memory-bus) and at the data center network. To evaluate our approach, we modeled a network and a memory subsystem on CloudSim simulator. We then used NAS Parallel Benchmarks, which consists of six different applications as parallel applications. We thoroughly evaluated our proposed approach measuring several assessment metrics including VMs placement, performance degradation, and network utilization of each link. The results of the simulation show that our proposed approach significantly reduces the total amount of traffic in the network where it reduces the average of the networks utilization by 25%.

Robust Virtual Machine Consolidation for Efficient Energy and Performance in Virtualized Data Centers

Cloud providers use virtualization technologies to provide an isolated execution environment and agile resource provisioning. However, virtualized data centers consume huge amounts of energy, which increases the operational costs. To optimize resource usage and reduce energy consumption of Infrastructure as a Service (IaaS) Cloud, it needs a continuous monitoring and consolidation of VMs using live migration and switching idle hosts to the sleep state. In this paper, we propose a robust consolidation approach to achieve equilibrium between energy and performance. The proposed approach consists of three algorithms: over-utilized host detection, VM selection, and VM placement. Additionally, we implement an adaptive historical window selection algorithm for reducing ineffective VM migration. To validate our approach, we implemented it using Cloud Sim simulator and conducted simulations for different days of a real workload trace of Planet Lab. The results show that our approach reduced the number of power change, the number of migrations, and average SLA violations by 38%, 74.8%, and 31.8%, respectively. Furthermore, it can decrease the energy consumption of network that results from VM migration.

Analysis and Simulation of HPC Applications in Virtualized Data Centers

Cloud computing is a consolidation environment that hosts heterogeneous applications. Currently, clouds use virtualization technologies to provide an isolated execution environment and agile resource provisioning. Furthermore, live migration is widely exploited to achieve energy saving. However, consolidation and live migration can cause resource contention and utilization overhead that influence application performance. In this paper, we implement a network model and a memory subsystem model into CloudSim simulator. Then, we present a thorough analysis of the High Performance Computing NPB suite including memory bandwidth demand, communication patterns between processes, and migration overhead. To study the performance of the NPB Benchmark suite, we simulate the behaviour of these benchmarks including multi-thread communication via shared memory, multi-process communication via network. Thus, this allows implementing efficient VMs scheduling and resource provisioning policies. The results show that using CPU utilization as a trigger to perform management action such as VM migration is misleading and might aggravate application performance. Furthermore, we present a power model including the utilization of CPU, memory-bus, and network. This model provides a precise estimation of power consumption.

Dynamic scalability and contention prediction in public infrastructure using Internet application profiling

Recently, the advance of cloud computing services has attracted many customers to host their Internet applications in the cloud. Infrastructure as a Service (IaaS) is on top of these services where it gives more control over the provisioned resources. The control is based on online monitoring of specific metrics (e.g., CPU, Memory, and Network). Despite the fact that these metrics guide resources provisioning, the lack of understanding application behavior can lead to wrong decisions. Moreover, current monitored metrics alone do not help in resources contention prediction, which is very common in shared infrastructures like IaaS. Nevertheless, the architecture of Internet applications, as multi-tier systems, makes contention prediction more complex while its influence can migrate from one tier to another. In this paper, we propose a pro-active global controller not only for dynamic resources provisioning, but also for predicting and eliminating contentions in multi-tier applications. Our technique combines monitored metrics, which are provided by current IaaS providers, with models that are built depending on the Internet applications profiling. The fitness of the monitored metrics to the application model is used for contention prediction. We examined our technique using RUBiS benchmark. The results express the efficiency of the developed algorithms in maintaining Internet applications performance even in shared infrastructures.

Routing schemes for network recovery under link and node failures

Routing schemes combined with link-state detection mechanisms can be used to recover connection paths or network connectivity under the cases of link and node failures. Some schemes have large computation and state-update overhead as re-routing is performed throughout the network. Moreover, as soon as new routes are found, some link may become congested with flows under recovery. This congestion may add extra recovery delays and even further link or node failures. In this paper, we propose proactive routing recovery schemes that perform rerouting on links affected by the failure, therefore, minimizing the computation overhead. Congestion avoidance is also achieved in these schemes by calculating the distribution of re-routed traffic in a proactive fashion. We compare our proposed schemes with the open shortest path first (OSPF) scheme and show that our schemes can provide higher utilization of links and nodes for large networks in post-recovery. We show simulation results under link and node failures.

SecPlace: A Security-Aware Placement Model for Multi-tenant SaaS Environments

Software-as-a-Service (SaaS) is emerging as a new software delivery model, where the application and its associated data are hosted in the cloud. Due to the nature of SaaS and the cloud in general, where the data and the computation are beyond the control of the user, data privacy and security becomes a vital factor in this new paradigm. In multi-tenant SaaS applications, the tenants (i.e., Companies) become concerned about the confidentiality of their data since several tenants are consolidated onto a shared infrastructure (i.e., Databases). Consequently, two main questions raise. First, how to prohibit a tenant from accessing others data? Second, how to avoid the security threats from co-located competing tenants? In this paper, we address the second question. We present Sec Place, a resource allocation model designed to increase the level of security for tenants sharing the same infrastructure. Sec Place avoids hosting competing companies on the same database instance. We minimize the risk of co-resident tenants by preventing any two tenants of the same business type to be hosted on the same database server. Sec Place utilizes the usage of tenant subscription data, such as business type and tenant size and place the tenant accordingly. We conduct extensive experiments to validate our approach. The results show that our approach is practical, achieves its goal, and have a moderate complexity.

Coordinating VMs' Memory Demand Heterogeneity and Memory DVFS for Energy-Efficient VMs Consolidation

We propose memory-aware VM consolidation to achieve energy-efficiency of a data center while enhancing performance of VMs. Consolidation without awareness of the memory-access demand can cause inefficient resource utilization and degrade system performance. In this chapter, we propose efficient consolidation of VMs based on the memory-access demand of these VMs to improve overall system performance. The proposed algorithm, Memory-bus Load Balancing (MLB), is executed by the Global Migration Manger (GMM). We evaluated our algorithm using several simulation setups and several performance metrics, such as performance degradation, VM placement, memory-bus utilization of each server, and energy consumption. The results showed that we could achieve balance in memory-bus utilization of servers and improve performance of the system compared to the CPU-based consolidation approach. Furthermore, we investigated the effectiveness of using the memory DVFS mechanism to achieve efficient energy consumption.