Resource allocation through optimized three-phase scheduled VMs by grey wolf optimization and introspection security analysis

Abstract

Even though, cloud computing reduces the operating cost by enabling adaptation of virtual machines, it has suffered in selection of optimal virtual machine due to shortage of resource or resource wastage, sudden changes in requirement so it requires optimal resource allocation. Resource allocation is the process of providing services and storage space to the particular task requested by the users. This is one of the important challenges in cloud computing environment and has variant level of issues like scheduling task, computational performance, reallocation, response time and cost efficiency. In this research work we introduce a three-phase scheduling method based on memory, energy and QOS in order to overcome the above issues which also yield low energy consumption, maximum storage and the high level Quality of Service (QoS). Biggest Memory First and Biggest Access First is introduced with NUMA scheduler and cache scheduler for memory scheduling and the optimal VM resulting from the three phases of scheduling is determined by Grey Wolf Optimization (GWO) algorithm. To carry the security level of optimized VMs, Streamline Security and Introspection security analysis are exhausted for detecting the malware VMs which results the secured and efficient VMs for further resource allocation. Our proposed methodology is implemented using the Cloud Sim tool and the experimental result shows the efficiency of our proposed method in terms of security, time consumption, and cost.

Keywords

Quality of service (QoS)Biggest Access First (BAF)Biggest Memory First (BMF)Grey Wolf Optimization (GWO)streamline security introspection security analysis

1 Introduction

Cloud will operate as the centralized management with unvaried and virtualized resource units. it’s a model for facultative omnipresent, convenient, on-demand network access that is principally centered on business and web-based applications [1]. The 5 indispensable characteristics of cloud are fast physical property, measured service, on demand self-service, omnipresent network access, location-independent resource pooling [2]. There are three delivery models computer code as a Service (SaaS), Platform as a Service (PaaS), Infrastructure as a Service (IaaS), and 4 readying models public cloud, non-public cloud, community cloud, hybrid cloud gift within the Cloud [3]. Cloud computing services are often used for all sorts of users equivalent to people, organizations, government corporations. They will utilize the services for outsourcing their information into the cloud, data processing and for net applications [4]. conjointly the cloud computing offers many advantages like quick readying, pay-for use, lower prices, quantifiability, fast provisioning, larger resiliency, hypervisor protection against network attacks, cheap disaster recovery and information storage solutions [5]. despite the fact that cloud computing yields superior services it ought to be free from attacks, resource wastage, security problems, for healthy atmosphere [6]. The key topics that are oftentimes connected with healthy cloud computing are virtualization, resource management i.e. resource utilization, resource programing, and security that have advantageous action in Quality of Service (QoS), turnout and lower overhead [7].

In Cloud computing, virtualization is one amongst the bottom technologies applicable to the implementation of Cloud computing that improves the supply of the user’s hosted services even just in case of hardware failure [8]. Virtualization is associate approach for sharing common resources of a physical host between multiple VMs. The resources typically embody memory, cupboard space, computer hardware and network devices [9]. Within the method of virtualization the virtual machine (VM) with the whole computer code stack are often migrated to a different server to extend pc resource utilization, efficiency, and quantifiability [10]. The resource allocation is one amongst the most effective theme for the virtual machine administration that is meant to touch upon each resource over-provisioning, and resource under-provisioning [11]. In resource allocation it’s essential that the virtual machine ought to be regular with elastic resources for dynamic provisioning and scaling supported user demands to extend system potency [12]. The primary and foremost factor in VM programing is to scale back the general needed energy. This will be done by introducing energy aware programing and placement algorithms and increased resource management [13].

One of key steps in building energy-aware programing is to scale back memory energy consumption which will give decent memory for every virtual machine that is termed as memory aware programing [14]. Consequently, the allotted memory resources could also be inadequate for large components of the submitted job and unnecessarily increase time interval and value [15]. Conjointly virtual machine adoption and diffusion are vulnerable by unresolved security problems that have an effect on each the resource allocation and also the programing [16]. so as to construct associate economical resource allocation model in cloud services, VM security is a very important issue [17]. By protective each the confidentiality of guest virtual machines and also the cloud design elements the protection are often improved [18]. This could achieved by the protection services equivalent to mistreatment longer key size, strict security access policies, and separate isolations for guarding non-public information, then on [19]. Therefore with the rise of the quantity of Security Service schemes, it’s necessary to assign the cloud resources to maximize the system rewards with the issues of the cloud resource consumption and incomes generated from cloud users [20].

Consequently, Resource utilization through the virtual machine (VM) has been varied because of time-varying workload in the VM, which results in shortage of resource or the resource wastage and there is a need of resource allocation. Resource allocation in VM is the process of providing services and storage space to the particular task requested by the users. This is one of the important challenges in cloud computing environment and has variant level of issues like scheduling task, computational performance, reallocation, response time and cost efficiency. For resource allocation, qualified scheduled and secured VM is necessary. Efficient utilization of resources will reduce cost and the time required for further processing. Several techniques were utilize and they only concentrated on the allocation based on energy and memory wastage that fails in the Quality of Service (QoS) requirements such as Throughput, Bandwidth, CPU speed, stability, Response time, and cost utilization The Dolphin Partner optimization used in the existing research has produced the best solution for moderate number of jobs but when the number of job increases the prediction of best solution for became difficult that may lead to lower convergence. Hence, this scenario reveals the limitation of existing allocation methods in terms of high resource utility rate and processing time.

Performance of optimization of Cloud Computing Environment is about making the components in the cloud to meet the component level requirements and customer expectations. It is aimed at increasing the performance of a cloud service at a minimum cost depending on various constraints.

By considering the above entire things, the paper focused to increase the efficiency of resource allocation in terms of maximum resource utility and minimum resource allocation time with improved QoS parameters by means of detecting the optimized and secured VM. The contribution of the paper for efficient resource allocation by detecting the optimized VM with security given below

For Qualified VM selection, three phase scheduling method based on memory, energy and QOS is carried out.

To determine the optimized VM, grey wolf optimization algorithm is used.

To analyze the security level of VM, streamline combined with introspection security is utilized.

The remainder of this paper is organized as follows. Section 2 presents some of the research works related to our proposed method. Section 3 explains our proposed scheduling and security method. Section 4 presents the simulation results of our proposed method and the comparison with some existing methods followed by the conclusion in section 5.

2 Related works

Abdul Nasir Khan et al. [21] represented the block-based sharing scheme (BSS) to cut back process, storage, and communication overhead from the mobile device. Additionally, the BSS provides higher security services to the mobile user because of the usage of countersign as a key part for the generation of blocks’ keys. The generated block keys are actively concerned in encryption/decryption method. The countersign was far-famed solely to the mobile user and actively utilize in achieving confidentially and integrity for the mobile users that leads toward a safer security theme. To form the BSS a lot of energy economical for mobile device, there ought to be a block insertion, deletion, and modification operations supported the progressive cryptography thought.

Abdul Nasir Khan et al. [22] explained Cloud-Manager-Based Re-Encryption scheme (CMReS), for the cryptography, decryption, and re-encryption operations by offloaded on trustworthy entity and cloud that improved the resource utilization on the mobile device. The explained CMReS consumed fewer resources on the mobile device whereas acting cryptography, decryption, and re-encryption operations. The trusted-entity had beneath the management of shopper organization and accountable to handle the requests from the users happiness to an equivalent virtual organization. Though the CMReS offloads the cryptography and decipherment operation on trustworthy entity, the involvement of trustworthy entity could have an effect on the quantifiability of system.

Luca Chiaraviglio et al. [23] represented the matter of together managing the upkeep prices and also the electricity consumption in a very CDC. When showing that hanging the facility states of PSs has a sway on each the failure management prices, similarly because the energy consumption. Since the OMEC drawback is NP Hard, thence delineated the MECDC rule that has been designed to showing wisdom leverage the exchange between totally different prices, similarly as taking into consideration their long run impact over time. Moreover, the work fails to realize the various problems, together with the definition and analysis of a lot of complicated failure models to require into consideration the impact on totally different elements, similarly as totally different temperatures of hardware cores.

Mohammad Shojafar et al. [24] represented the important issue in job programing for distribution jobs to the foremost appropriate resources, considering user preferences and needs. For that, the paper explained a hybrid approach known as FUGE, which had supported fuzzy theory and a genetic rule (GA) that aims to perform best load equalization considering execution time and price. To optimize the calculation of the fitness worth and crossover steps any there’s a desire for considering the VM energy consumption as input parameters of the fuzzy system.

Wang, et al. [25] made primary node primarily based design for common tele-health administration on cloud that thought of each memory area and delivery proficiency. They in addition composed a calculation for anticipating and allocating the longer term information measure of all VMs within the tele-health administration. The strategy had balanced each parameter of a Hidden Markov Model (HMM) through gathering the authentic knowledge of the information measure work. at that time they anticipated the longer term information measure utilization of VMs, a programing strategy was utilized and consequently dynamical the information measure to every VM for medical aid administration. The results incontestable that the calculation had given a high-precise accuracy, by leading the allotting module. By and by, the calculation had increased the responsibility of tele health administrations for golf shot away and conveyance patients’ knowledge among Data Centers (DCs). However, those ways failed to take into account the overhead of resource management on a cloud that must schedule and regulate the allocation of network information measure for every knowledge center node once a public emergency event happens.

Today, Cloud computing [32] become an emerging technology which will has a significant impact on IT Infrastructure. Still, Cloud computing is infancy. In the current cloud computing environment there is numerous of application, consist of millions of module, these application serve from large quantity of users and the user request becomes dynamic. So there must be provision that all resources are dynamically made available to satisfy the needs of requesting users [33]. The resource provisioning was done by considering Service Level Agreements (SLA) and with the help of parallel processing using different types of scheduling heuristic [34]. In this paper we realize such various policies for resource provisioning and issues related to them in current cloud computing environment. Keywords: Cloud computing; Scheduling, Service Level Agreements (SLA), Virtualization, Virtual Machines (VM). - Distributed computing is a developing innovation which gives compelling administrations to the customers. It licenses customers to scale here and there their assets utilization relying on their necessities Because of this, under arrangement and over arrangement issues may happen. To defeat this relocation of administration use Our Paper concentrates on conquering this issue by appropriating the asset to various customers through virtualization innovation to upgrade their profits. By utilizing virtualization, it allots datacenter assets powerfully in view of uses requests and this innovation likewise bolsters green innovation by advancing the number of servers being used. We show another approach called “Skewness”, to figure the unevenness in the Multi-level asset usage of a server. By enhancing Skewness, we can join diverse sorts of workloads enough and we can enhance the entire utilization of server assets.

From the above related works it is known to be that providing services and storage space to the particular task requested by the users is one of the important challenges in cloud computing environment and has variant level of issues like scheduling task, computational performance, reallocation, response time and cost efficiency. Several techniques were utilized especially energy aware prioritization and memory aware prioritization were employed that concentrated only on the allocation based on energy and memory wastage but it fails in the Quality of Service (QoS) requirements such as Throughput, Bandwidth, CPU speed, stability, Response time and Utilization cost. This failure reveals that the limitations of existing allocation methods in terms of high resource utility rate and processing time mean while the Dolphin Partner Optimization used in the existing research has produced the best solution for moderate number of jobs but when the number of job increases the prediction of best solution became difficult that may lead to lower convergence.

3 Resource allocation in optimized and secured VM

Cloud computing can be viewed as a dynamically-scalable pool of resources in which virtual machine (VM) scheduling and allocation is essential. For resource allocation, qualified and secured VM is necessary. Therefore, our aim is to achieve optimized VM selection, which can be done by means of following two stages namely Qualified VM selection, Secured VM selection. In Qualified VM selection the number of virtual machines from the cloud is scheduled on three phases that are Memory Based Scheduling, Energy Based Scheduling and QoS Based Scheduling.

In first phase the virtual machines which are satisfied the memory utilization can be scheduled by using two types of scheduler called cache scheduler and NUMA (Non-Uniform Memory Access) scheduler. For improving the computation speed of these schedulers Biggest Access-Memory First mechanism is integrated with the memory based scheduling. The Biggest Access First mechanism helps the NUMA scheduler to prioritize the VM based on their accessing capability similarly the Biggest Memory First supports the cache scheduler to prioritize the VM based on their memory space. Secondly, by using energy based scheduling the energy efficient VMs are prioritized. Finally in QoS based scheduling by means of QOS parameters such as Bandwidth, CPU speed, stability, Response time, Utilization cost and throughput, the VMs are scheduled. The three scheduling scheme will generate three set of VM that are energy efficient, memory efficient and QoS efficient. From these three set of scheduled VM a single optimized VM set selected by means of a new Grey Wolf Optimization. Even though the number of jobs increases, our proposed optimization will generate the best solution, which is the advantage of our proposed framework. The optimization will generate a well-qualified VM set by means of energy, memory and QoS. From the qualified VM the secured VM set is extracted by using streamline security and introspection based security. The streamline security analysis is performed to acquire the VM that have minimum number of failure jobs. Also in order to achieve more secure VM it is essential to build an efficient security scheme that will detect the malicious VM. This can be achieved by a new technique called introspection based security analysis. Virtual machine introspection is a technique used to inspect and analyze the code running on a given virtual machine. By using this incoming information, the process flow of VM is detected. If any flags set the process flow then that VM is recognized as a malicious VM and is rejected. In this way we will get a completely secured and qualified VM for resource allocation. Thus the proposed approach will improve the computation speed, QoS parameters with an end goal of enhanced resource allocation. The process flow for the proposed methodology is given in the following sections and depicted in Fig 1.

Fig. 1

Architecture of proposed method.

3.1 Three-phase scheduling

In three-phase scheduling, the qualified VMs are selected based on the energy level consumption, maximum memory storage, and QoS metrics level. From these scheduling the VMs are given to the next level for optimization and the following section explains about the techniques used in the scheduling that is a scheme for viewing the runtime state of a virtual machine (VM) to cope with the following scenarios such as performance consideration during sudden change in requirement and the semantic gap problem (i.e., understanding the low-level information available through VMI (Virtual Machine Introspector) which enables the vendor to allocate the optimal virtual machine not only based on memory, Qos and cost but also based on the level of security.

3.1.1 Energy aware scheduling

The impact of energy consumption in VM is dependent on various factors such as utilization of CPU, memory power, disk and network interfaces. In order to be achieving efficient energy, VMs should be ideally start and halt quickly and without utilizing too much of energy usage. Considering the above factors, the CPU utilizes more energy than other factors. In our proposed work we focus on CPU utilization as an important concern for the VMs in cloud computing environment. CPU utilization is directly proportional to the overall system node. i.e) when the CPU utilization increases the energy consumption level by the VMs will increase. The energy model for the VMs can be calculated as follows: $E_{vm} (u) = E_{idle} \cdot E_{max} + (1 - E_{idle}) \cdot E_{max} . E_{cpu}$ (1)

Where, E_idle is the fraction of energy consumed by the idle server, E_max is the maximum energy consumed when the server is fully utilized, E_cpu denotes the CPU utilization in VM.

The average of an idle server consumes approximately 70% of the energy consumed by the server running at the full CPU speed. When the server is fully utilized, the maximum consumption of energy is set to 250 W which is usual value for the modern servers. The above equation (1) can be simplified as $E_{vm} (u) = E_{max} (E_{idle} + (1 - E_{idle}) \cdot E_{cpu})$ (2)

Where E_idle is 0.7. The use of the CPU may change after some time due to the workload variability. Thus the CPU utilization of function of time and can be represented byE_cpu (t). Therefore, the total energy consumption by the VMs can be defined as an integral of the energy consumption function over a period of time can be expressed as follows $E = \int_{t_{0}}^{t_{1}} E_{vm} (u (t)) dt$ (3)

The result of an above equation produces the energy consumption of VMs in cloud computing environment. After calculation of energy model for each VM some number of VMs will be select based on efficient energy level. Moreover the edge model of the bandwidth between the VMs on the cloudlet need to be discussed here for selecting the VM in accordance with the bandwidth constraint. Consequently, recent advanced technologies like Software-Defined Networking (SDN) provide the potential to implement adaptive bandwidth, which allows us to change the bandwidth on each link connected to a cloudlet node, as long as the total bandwidth of the links connected to a node is below the node’s maximum bandwidth capacity. This adaptive bandwidth offers more flexible network resources allocation. It relieves us from the constraint of the link’s bandwidth, which allows us to only focus on the constraint of each node’s maximum bandwidth capacity. To be specific, we only need to consider whether or not the bandwidth allocation exceeds the node’s maximum bandwidth capacity when we allocate bandwidth resources to VMs. After this the VMs having efficient memory are selected which is illustrated in subsequent section.

3.1.2 Memory aware scheduling VMs

The memory efficient VMs are selected based on the cache scheduler and NUMA (Non-uniform Memory Access) scheduler. The above two schedulers examine the efficient VMS with high access and capability of maximum storage in cloud computing environment. Cache behavior in cloud computing swaps VMs with minimum LLC (Last Level Cache) misses by monitoring all VMs. For the average memory access latencies, the NUMA policy will be used to improve the performance of cloud computing environment. Based on the VM status information collected from all the nodes, the cloud scheduler makes global scheduling decisions as well as local scheduling for reducing overall LLC misses. But the NUMA scheduler considers the global scheduling only and migrate VMs to different sockets to reduce the overall LLC misses. The above two schedulers’ leads the complexity by calculating the LLC misses of all VMs in a core. To eliminate this problem of both cache behavior and NUMA scheduler, BMF (Biggest Memory First) and BAF (Biggest Access First) techniques are introduced which will reduce the complexity that means to improve the computation speed of cache and NUMA schedulers Biggest Access-Memory First mechanism is integrated with the memory based scheduling. The Biggest Access First mechanism helps the NUMA scheduler to prioritize the VM based on their accessing capability similarly the Biggest Memory First supports the cache scheduler to prioritize the VM based on their memory space. Secondly by using energy based scheduling the energy efficient VMs are prioritized. Finally in QoS based scheduling by means of QOS parameters such as Bandwidth, CPU speed, stability, Response time, Utilization cost and throughput, the VMs are scheduled. The following notations are used in the techniques such as BMF and BAF.

Consider the set of all VMs denoted as {V} and V_i means the VM from the set {V} is executed on the core processor i.The number of core processor presents in the system is denoted as {C}, where C ≥ 2, and it has the number of VMs. The VMs are scheduled in the queue based on rank level whereas rank value can be assigned by sharing the amount of CPU time. Each VM has the set of memory nodes which is called as access set and denoted by S (v) where v corresponding VM. In the time of t, the currently used set of memory nodes i.e) access set by VMs on every processor is denoted as A (t). A (t) is the union of the access sets of virtual machines that are run on the processor cores.

The cache scheduler in cloud computing migrates VMs to minimize the LLC misses in cloud system in which the LLC misses is stated as, where fails to attempt the data content from the cache or memory. In cache scheduler there are two phases namely local phase and global phase and the local phase first compute the LLC misses of both physical machines (PMs) and VMs. Then gathering the list of LLC misses for all VMs in PM_i, i, (i = 1, 2,. . . n), where n denotes the number of physical nodes. The sorted list of VMs can be produced by the LLC misses, in which the higher LLC misses values are assigned as lower priority values. In global phase the physical nodes with maximum and minimum LLC miss values are identified and then find the maximum and minimum LLC misses for VMs from the both PMs. Then check whether the difference between the LLC of two physical nodes is greater than threshold value. If it is greater means swap the VMs with maximum and minimum LLC misses. For calculating the LLC misses of all VMs the selective VMs are given as input to the cache scheduler. So the complexity of cache scheduler is reduced for that, the Biggest Memory First (BMF) technique is proposed in this paper and the technique is discussed in the following section. The fundamental goal of this memory-aware virtual machine programming is to reduce the quantity of memory nodes within the standby mode. Within the following sections, we have a tendency to chiefly concentrate on this issue. Once making a virtual machine hardware, one in every of the vital issues is to confirm fairness among virtual machines. In our memory power management design, we have a tendency to insert memory-aware programming options into the hardware of rather than creating a replacement hardware from scratch. Thus, our hardware selects a virtual machine among the virtual machines that have credits and reside within the run queue. This quickly reorders the execution sequence of virtual machines within the run queue, and thus, ensures that the memory-aware hardware provides constant level of fairness.

Biggest memory first (BMF)

Biggest memory first (BMF) schedules virtual machines supported the recognition of individual memory nodes. The recognition of a definite memory node is outlined because the range of virtual machines that use the memory node within the entire system. The upper the recognition of a memory node is, the additional virtual machines use it. BMF in cache scheduler operated based on the popularity of individual memory nodes of VMs and the popularity of each memory node can be stated as the number of virtual machines that use the memory node in the system [31]. If the popularity of memory node is high means the particular memory node is used by the more number of VMs. So the memory node of VMs produced by the algorithm has maximum capability of storing the contents in cloud computing environment. By the above process the efficient VMs are selected with the biggest memory node and the BMF algorithm uses the following two global variables: α Denotes the set of memory nodes with high popularity, β Used for selecting the memory node for α.

Algorithm 1
Algorithm for BMF

1. If (Q_i =empty) OR begins the scheduling then

2. β← 1 //select the memory node forα

3. α← null

4. end if

5. begin

6. α← α ∪ the β^th popular memory node in access set

7. λ ← null

8. for ^{n ←1,|Q_i|} do

9.        v← the n^h VM in Q_i

10. if

         S (v) ⊂ αAND|S (λ) | < |S (v) |then

11.                          λ ← v

12. end if

13. End for

14. End

In the result of BMF, the α set contains the high popularity nodes. The α set is extended, when the memory node with high popularity is inserted. By the BMF algorithm the efficient memory of VMs can be selected. These VMs are given as input to the cache scheduler for performing the global and local phase scheduling to calculate LLC misses. From the group of VMs with maximum and minimum LLC, select the VMs with less LLC misses,

$\begin{matrix} V_{c_{1}} = \sum_{i = 1}^{n} V_{i} (V_{i} < min LLC) / / VMs with \\ minimum LLC in cache \end{matrix}$ (4)

After the scheduling by cache scheduler NUMA scheduling is performed which is explained as follows. The NUMA scheduler considers the NUMA affinity for improving the memory access latencies with minimum LLC misses of cloud computing environment. In NUMA the only scheduling as global scheduling in which checking the difference of highest and smallest LLC misses of different system is greater than the threshold value. Instead of selecting the efficient VMs all the VMs chosen from the different system for calculating LLC misses, due to the above process the time complexity will increase and fail to choose the efficient VMs for resource allocation in cloud computing environment. So the above problem can be eliminated by the BAF technique for reducing the complexity in NUMA scheduler. In the scheduling iteration, the scheduler selects two sockets, within the Cloud procedure, with the most important and smallest numbers of LLC misses of VMs are selected and the two VMs also swap. For improving the access latencies in NUMA the selective VMs are chosen for calculating LLC misses, for that BAF technique will be introduced and it will be demonstrated in the next section.

Biggest Access First (BAF)

To improve the NUMA scheduler for selecting efficient VMs, BAF method is employed, in which the VMs with biggest access set is chosen. In BAF, the VMs are selected from the run queue system and checks for each VM access set. If the access set of VM is overlapped with the A (t), then the VMs with biggest access set is chosen. The following notations are used in the BAF algorithm which is explained as follows: Q_i denotes the ordered set of VMs waiting in run queue of the ith processor core. λ denotes the next VM to run as a result of scheduling decision. The result of BAF technique produces the VMs with biggest access set, and then the selected VMs are given as input to the NUMA scheduler for the LLC scheduling. Similar to Cache scheduler the VMs with minimum LLC are selected in the NUMA but the difference of both is get the maximum storage and biggest access respectively.

$\begin{matrix} V_{c_{2}} = \sum_{i = 1}^{n} V_{i} (V_{i} < min NUMA) / / VMs with \\ minimum LLC in NUMA \end{matrix}$ (5)

From the above Equations (4), (5) the VMs with minimum LLC miss are chosen and the memory aware scheduling produces the VMs with maximum storage and highest level access for the efficient resource utilization. The BAF algorithm is explained as follows:

Algorithm for BAF

1. λ ← empty

2. for n ← 1, |Q_i| do

3. v← the n^h VM in Q_i

4. if S (v) ⊂ A (t) AND|S (λ) | < |S (v) |then

5. λ ← v

6. end if

7. end for

8. if λ =empty then

9. shared ← 0

10. for n ← 1, |Q_i| do

11. v← the n^h VM in Q_i

if |S (v) ∩ A (t) | > shared then

13. λ ← v

14. shared ←S (v) ∩ A (t)

15. end if

16. end for

17. end if

The VMs with minimum LLC misses are selected from cache scheduler and NUMA scheduler and further they are integrated to generate memory efficient VM which can be expressed as follows $M_{i} = V_{c_{1}} + V_{c_{2}}$ (6)

Where V_{c
₁} -VMs selected from cache scheduler with minimum LLC

V_{c
₂} - VMs selected from NUMA scheduler with minimum LLC

After the second phase of scheduling, QoS scheduling is employed for the efficient selection of VM that is demonstrated in the succeeding section.

3.1.3 QoS aware scheduling

The QoS (Quality of Service) metrics has the important role in selecting efficient VMs from the cloud computing environment and also for the resource utilization. The VM will provide the services to the client request with the efficient cost based resources as well as the tasks can be done within the particular period of time. Because of time varying workloads in VMs the efficiency will reduce, so this is a challenge to meet the efficient resource utilization of VMs by considering the QoS factors. In this paper the QoS metrics such as Bandwidth, CPU speed, stability, Response time, Utilization cost and throughput is evaluated for the better performance.

Bandwidth

Bandwidth (BW) is stated as the data transfer rate, the amount of data that can transfer from host to another. However, Bandwidth is the capacity for the VM and it is particularly important for $\underline{I / O}$ devices. For example the client send the job request for 254 kpbs but the service provider has the capable of give the response with 128 kpbs means the bandwidth metric become fail to meet the requirements of client.

B W_{u t i l} (V M_{i}) = (A c t u a l u s e d B W i n V M_{i}) / T o t a l B W * 100

(7)

CPU speed

The CPU Speed (C) is the measure of processor speed. i.e) the number of jobs could be done per second. For efficient VMs it has the capable doing the more number of works based on the service request.

$C = \frac{No . of jobs}{Time (in seconds)}$ (8)

Stability

The stability (S) of VMs represents status of current task and the status can be assigned by the string code. The status of the current task may be in the form of ready (r), execution (e), success (s) or in the queue (q). The string value for the status is expressed as follows: $\begin{matrix} \begin{matrix} Re ady (r) = {start} \\ Execution (e) = (run} \\ Success (s) = {yes} \\ Fail (f) = {NO} \end{matrix} \end{matrix}$

If the status is Ready means it is represented by the string code “start”, Execute means “run”, Success means “yes” and fail means “NO”. Response time

Response Time (RT) is stated as time taken for sending the job request and arrival for response from the source. In general, the cloud’s response time depends upon the average load of the machine. The average load is directly proportional to the response time, i.e., increase in average load delays response to the client. By executing more number of virtual instances, more number of users tries to access the resource therefore, the load average increase and for better performance, this parameter should be reduced. It is usually denoted as milliseconds (ms) and can be represented by the following: $RT = AT - ST$ (9)

Where, AT is the arrival time for the job, ST is the sending time of job request.

Utilization cost

The Average Utilization Cost U_c of CPU or memory for VMs in cloud calculated as $U_{c} = \sum_{i = 1}^{I} \frac{C (t)}{I}$ (10)

The utilization rate of processor or memory at the time slot t is calculated as, $C (t) = \frac{{RT}_{n} a_{nt}}{TR}, i = {1, 2, \dots I}$ (11) where, RT -Number of CPU or the size of the memory requested by task, TR -Total CPU or memory in the cloud, If task runs in thei^tth time slot, then a_ni = 1 ; otherwise, a_ni = 0.

Throughput

Throughput (T_p) is the number of transmitted data per unit of time in the system, therefore higher throughput comes along better system load balancing situation. Let the number of tasks (n) and given to the number of machines (m) that are run on the cloud provider and the execution time is denoted as T (n, m), then the throughput equation is as follows:

$T_{p} = \frac{n}{T (n, m)}$ (12)

The QoS for all VMs can be calculated by the above metrics and it is expressed as follows:

$\begin{matrix} {QoS}_{i} = BW + C + S + RT + U_{c} + T_{p}, \\ i = 1, 2, \dots n \end{matrix}$ (13)

Where n is the number of VMs present in the cloud system, S represents the stability of VMs. The maximum QoS of the VMs are the efficient one for resource utilization. In the phase of scheduling the three sets of VMs are selected based on the low level energy consumption, maximum memory utilization and the high level of throughput.

3.2 Grey wolf optimization

The Grey Wolf Optimization algorithm produces the efficient VMs from the energy, memory and QoS based scheduling in the previous phase. In this algorithm the grey wolf searching for the prey and finished this work by hunting and the number of species with social hierarchy follow the strict manner for finding the best prey. The topmost level α is responsible for making decisions about hunting, sleeping place, time to walk and so on and these are the leaders of the pack. The second level is β, these are the subordinate wolves, which help the α in decision making or other activities. The third level is δ wolves have to submit to α and β. In the GWO algorithm the three best candidate solutions are namely α, β andω these can be taken as the best VMs from each set such as energy, memory and QoS set respectively. The model of the wolves’ encircling process is as follows: $\vec{X} (t + 1) = X_{p} (t) - \vec{A} \cdot \vec{D}$ (14) $\vec{D} = | \vec{C} \cdot {\vec{X}}_{p} (t) - \vec{X} (t) |$ (15)

Where $\vec{A}, \vec{C}$ are coefficients vectors, ${\vec{X}}_{p}$ is the prey position, $\overset{0.5 em ⇀}{X}$ is the grey wolf position t indicates the current iteration. $\vec{C} = 2 . {\vec{r}}_{2}$ , $\vec{A} = 2 \vec{a} \cdot {\vec{r}}_{1} - \vec{a}$ . Where components of $\vec{a}$ are linearly decreased from 2 to 0 over the course of iterations and ${\vec{r}}_{1}, {\vec{r}}_{2}$ are random vectors in [0, 1]. The readjusting the position with respect to α, β and δ by the following model,

$\begin{matrix} - 12 pt & {\vec{D}}_{α} = | {\vec{C}}_{1} \cdot {\vec{X}}_{α} - \vec{X} |, {\vec{D}}_{β} = | {\vec{C}}_{2} . {\vec{X}}_{β} - \vec{X} |, \\ - 12 pt & {\vec{D}}_{δ} = | {\vec{C}}_{3} . . {\vec{X}}_{3} - \vec{X} | \end{matrix}$ (16)

$\begin{matrix} - 22 pt & {\vec{X}}_{1} = {\vec{X}}_{α} - {\vec{A}}_{1} \cdot ({\vec{D}}_{α}), {\vec{X}}_{2} = {\vec{X}}_{β} - {\vec{A}}_{2} . ({\vec{D}}_{β}), \\ - 22 pt & {\vec{X}}_{3} = {\vec{X}}_{δ} - ({\vec{A}}_{3} . ({\vec{D}}_{δ})) \end{matrix}$ (17)

Renew the position of the current task is given by (17) $\vec{X} (t + 1) = \frac{{\vec{X}}_{1} + {\vec{X}}_{2} + {\vec{X}}_{3}}{3}$ (18)

GWO is updating the parameter $\vec{a}$ and is linearly updated in each iteration to range from 2 to 0. $\vec{a} = 2 - t \cdot \frac{2}{{Max}_{iter}}$ (19)

Based on this technique the three set of VMs are selected based upon the energy level consumption, memory capability and efficient QoS and these VMs are assigned as three best solution X_{α
₁}, X_{α
₂}, X_{α
₃}. Grey wolf position, prey positions are taken as virtual machine position and optimal solutions respectively. From the Equations (3), (4), (5) and (12), the fitness function of GWO is calculated as

$F (V M_{i}) = {\begin{matrix} \min \int_{t 0}^{t 1} E_{v_{i}} (u (t)) d t) + \max [M = \sum_{j = 1}^{n} V_{j} (V_{j} < \min L L C) + \sum_{k = 1}^{n} V_{k} (V_{k} < \min N U M A)] \\ + \max (Q_{s} = B W + \frac{R T_{n} a_{n i}}{T R} + \frac{n}{T (n, m)} + (A T - S T) + T_{p} \end{matrix}$

Algorithm for GWO

Initialize the set of Virtual Machines VM_i = (i = 1, 2, …, n)

Initialize a, A and C

For all VM_i do

Calculate fitness function

End for

Get the first three best VMs from each set as

While (t < Max number of iterations)

For three best VMs

Update the position of current VMs using (17)

End for

Update a, A and C

For all VM_i do

Calculate fitness function using (19)

End for

Update α, β and δ

t = t + 1

End while

Return X_α

The GWO algorithm produces the result of optimal set of VMs with low energy consumption, maximum memory storage and maximum QoS. These optimal set VMs are given as input to the security analysis for producing the secured VMs.

3.3 Introspection based security analysis

VMI (virtual machine introspect) inspects the VM memory and disk from the outside without intrusively injecting agents. Thus, one of its main benefits is to protect the VM monitoring tool from being compromised in the event of a successful security attack. To that end, the VM monitoring tool is placed outside of the target VM but in a trusted VM. The guest VM’s internal behavior is then inferred by using the VM state information obtained at the hardware level.

For efficient resource allocation the VMs have to be more secure and it is essential to build an efficient security scheme that will detect the malicious VM. The streamline security analysis checks the security for each VM from the optimization section and detects the VMs, if any attack will be present. In the energy and QoS based VM it is easy to detect the failure and success jobs thereby inspecting the security is good enough. In case of the memory based $F (V M_{i}) = {\min (\int_{t_{0}}^{t_{1}} E_{υ i} (u (t)) d t) + \max [M = \sum_{j = 1}^{n} V_{j} (V_{j} < \min L L C) + \sum_{k = 1}^{n} V_{k} (V_{k} < \min N U M A)] + \max (Q_{s} = B W + \frac{R T_{n} a_{n i}}{T R} + \frac{n}{T (n, m)} + (A T - S T) + T_{p})$ (20)

VMs using NUMA scheduler, the failure and success job by the provider’s resources such as CPU, memory, containing processors and local memory, energy is indicated by the System Resource Allocation Table (SRAT), which holds the malicious VMs and application shared by the third parties. In the streamline security level checks the parameters of each VM based on both processor and memory level, if any VMs are fails to satisfy the parameter level means the VM become the malicious one. Some of the parameters considered in the level of streamline security in processor level as Total number of job requests, completed/successful jobs and failure jobs and in memory level remaining memory is consider. These parameters are expressed in streamline security level is expressed as following, $SSL = \frac{(1 - RJ) \times RJ + (CJ)^{2}}{RJ (FJ + CJ + R_{mem})}$ (21) where,

SSL - Streamline Securiy Level, RJ - Re quest Jobs, CJ - Comlpeted Jobs, FJ - Failure Jobs, R_mem - Remaining memory

From the above equation (20) calculating the SSL for each VMs of result from the optimization algorithm then check whether all VMs are satisfy the above parameters, if it is satisfied the VMs are in the security level else the VMs are neglected. To increase the security of VMs then the next level of security can be achieved by a new technique called introspection based security analysis. The result of previous security level VMs are givens as input to the introspection security level, which is used to inspect and analyze the code running on a given virtual machine. This security analysis, inspects and monitoring the VMs from outside by checking the process flow of each VM. The process flow of each VM is monitored by the flag, if any malicious attack is present in the VM the flag value is set and reset. The malicious VMs are eliminated from this level and the remaining VMs are considered as secured for resource allocation.

4 Results and discussions

In our proposed method the qualified and secured VMs are selected based on scheduling, optimization and security policies. For qualified VMs are chosen according to the low level energy consumption, maximum storage utilization and high level of QoS metrics in three-phase scheduling. From these scheduling the VMs are given to the next level for optimization that is, grey wolf optimization for the qualified VMs. For the security level streamline security and introspection security is added for detecting the malware VMs. The unsecured VMs are neglected then the qualified and secured VMs are obtained. The following section describes the experimental setup, proposed results and performance evaluation. The proposed technique is implemented in Java Cloudsim platform and it supports the modeling for virtualized machine on the processor. In the proposed system following resource allocation strategies are considered such as Total number of virtual machines, Energy aware scheduling, memory aware scheduling, Qos based virtual machines, qualified virtual machines, secured virtual machines.

4.1 Simulation result

In proposed method the total number of virtual machines as 100. For energy consumption, memory aware and QoS three set VMs obtained by the three stage scheduling methods. These VMs are given as input to the grey wolf optimization algorithm for selecting the qualified set VMs based on the fitness function and the qualified set is pass to the security level. Using the security analysis techniques, the secured VMs are fulfilled. The below table shows the resource allocation strategies with number of VMs chosen based on the scheduling, optimization and the security level. Table 1 shows efficient VMs selection criteria for resource allocation.

Table 1
Efficient VMs selection for resource allocation

Resource allocation strategies Number of VMs

Total number of VMs 100

Energy aware scheduling 7

Memory aware scheduling 6

QoS based VMs 7

Qualified VMs 3

Secured VMs 1

Resource allocation strategies	Number of VMs
Total number of VMs	100
Energy aware scheduling	7
Memory aware scheduling	6
QoS based VMs	7
Qualified VMs	3
Secured VMs	1

In the energy aware scheduling from the total number of VMs, the efficient VMs are chosen according to the low level energy consumption. In our proposed method seven VMs 1, 24, 20, 70, 12, 28, 61 are selected for the resource utilization with respect to the low level energy consumption and the below Fig. 2 graph shows the seven VMs with respective energy consumption level.

Fig. 2

VMs with energy level.

The below Fig. 3 show VMs 11, 46, 47, 67, 68, 70 with biggest memory from the total number of VMs based on memory aware scheduling. These VMs are having the maximum available storage than remaining VMs and these are the efficient for resource utilization.

Fig. 3

VMs with biggest memory.

QoS metrics for efficient resource allocation the VMs are having the high level of QoS such as throughput, stability, response time, utilization cost and bandwidth. For proposed technique from the total number of VMs, where 7 VMs are chosen according to the QoS aware scheduling and the below Table 2 represent the VMs with QoS metrics.

Table 2

Selective VMs with QoS metrics

VMs Id	Throughput (ms)	Stability (status)	Response Time (ms)	Utilization Cost	Bandwidth
4	719	Success	428.01	0.640	87
5	689	Success	438.01	0.8	92
16	719	Success	490.18	0.608	81
8	600	Success	493.83	0.31	88
1	659	Success	498.86	0.9	75
11	629	Success	514.79	0.6	87
2	599	Success	535.27	0.632	82

In GWO the input is taken as the VM of the above scheduling method, then calculate the fitness values of VMs. According to the fitness function with maximum value the optimized sets are chosen and the below Figs. 4–6 shows the optimized set VMs as 1, 11, 70.

Fig. 4

Energy level for optimized set VMs.

Fig. 5

Memory aware for optimized set VMs.

Fig. 6

QoS level for optimized set VMs.

In Fig. 7, from the optimized set VMs the security level is applied for detecting the malicious attacks by introspection level and the streamline security. In the level of streamline security the minimum number of failure jobs VMs is chosen with high SSL values then the secured set VMs are chosen by the process flow of VMs with flag values for identifying the attacks VMs. From the optimized set the qualified and secured VM is chosen and then the above graph shows the VM 1 has high security compared to other optimized VMs numbered 11 and 70. The below section discuss about the performance evaluation of proposed method with other existing techniques.

Fig. 7

Secured percentage level for optimized VMs.

4.2 Performance evaluation and comparison

Comparison of existing methods and the proposed method techniques are evaluated by the QoS metrics such as throughput, Utilization cost, bandwidth (using the equations in section 3). Below figure shows the graphical representation of comparisons.

Using the three stage scheduling with the grey wolf algorithm, the optimized VM is selected and security is inspected by the streamline and introspection analysis to provide the secured and qualified VM. In the Fig. 8, the comparison for the throughput for the proposed methodology has been proved that the proposed approach finds the qualified and Secured VM for resource allocation hence that it produces the higher throughput values of 780. Compared to the low throughput for the FCFS, RWS, SLPSO, SHC, the proposed methodology achieves high. Since the proposed system has the ability to adapt to any changes due to incorporation of grey wolf optimization. Fig. 9 shows the comparison of utilization cost between the existing and proposed method in that case following existing methods has been taken into account FCFS, RWS, SLP, SHC and proved that proposed system has obtained optimal utilization cost. Fig. 10 shows the comparison of bandwidth between the existing and proposed method in that case following existing methods has been taken into account FCFS, RWS, SLP, SHC and proved that proposed system has achieved better bandwidth.

Fig. 8

Comparison of Throughput between existing and proposed method.

Fig. 9

Comparison of Utilization cost between existing and proposed method.

Fig. 10

Comparison of Bandwidth between existing and proposed method.

In addition to the utilization rate, is calculated using the equation (10) to prove the effectiveness of our proposed method. The comparison of utilization cost of our proposed method with respect to CPU and memory is tabulated as below.

The Table 3 represents memory utilization rates for our proposed method, SLPSO, SPSO, FCFS, SHC and RWS. It can be seen that our proposed method can achieve a higher resource utilization ratio than the SLPSO, SPSO, FCFS, SHC and RWS.

Table 3

Comparison of utilization rate

Utilization rate	Proposed	DPO	SLPSO, [26]	SPSO [27]	FCFS [28]	SHC [29]	RWS [30]
Average CPU Utilization rate	0.921	0.905	0.829	0.826	0.30	0.313	0.20
Average memory Utilization rate	0.859	0.810	0.774	0.771	0.75	0.766	0.822

The Fig. 11 illustrates that the average CPU and memory utilization rates of our proposed method, SLPSO, SPSO, FCFS, SHC and RWS. The figure clearly explains that the utilization rate of our proposed method is higher than the SLPSO, SPSO, FCFS, SHC, RWS. The average CPU utilization of our proposed method is 0.921 and average memory utilization of our proposed method is 0.859. For the security level and qualified VMs the other existing methods such as First Come First Scheduling (FCFS), self-adaptive learning particle swarm optimization (SLPSO), Roulette Wheel Selection (RWS), Stochastic Hill Climbing (SHC), the proposed level has the secured and qualified VMs. The proposed work fulfilled the qualified with low level energy consumption, maximum storage and high level of QoS then the GWO optimization produces the optimized set compare than other existing methods since the proposed methodology enables optimal memory utilization by incorporating grey wolf optimization algorithm.

Fig. 11

Comparison of Average CPU and Memory utilization rates.

5 Conclusion

For the better resource allocation, a qualified and secured Virtual machine is selected by the proposed three stage scheduling based optimization with streamline security and introspection analysis, which detects the optimized VM based on the memory, QOS and Energy by the three stage scheduling, and the GW optimization. For the security inspection, the streamline security with the introspection analysis carried that finds the qualified and secured VM for the resource allocation with the average memory and CPU utilization of 0.859 and 0.921. To save the memory energy further, the work extended to support virtual machine migration and devise various migration policies. Even though optimization techniques have increased the performance of the cloud computing in selecting the optimal virtual machine, optimization techniques themself have had limitations such as computational complexity and time complexity in achieving convergence.

References

Zissis

, Lekkas

, Addressing Cloud computing security issues, Future Generation Computer Systems28(3) (2012), 583–592.

Xing

, Zhan

, Virtualization and Cloud computing, In Future Wireless Networks and Information Systems, Springer, (2012), 305–312.

Rimal

B.P.

, Jukan

, Katsaros

, Goeleven

, Architectural requirements for Cloud computing systems: an enterprise Cloud approach, Journal of Grid Computing9(1) (2011), 3–26.

Paquette

, Jaeger

P.T.

, Wilson

S.C.

, Identifying the security risks associated with governmental use of Cloud computing, Government Information Quarterly27(3) (2010), 245–253.

Subashini

, Kavitha

, A survey on security issues in service delivery models of Cloud computing, Journal of Network and Computer Applications34(1) (2011), 1–11.

Hwang

, Li

, Trusted Cloud computing with secure resources and data coloring, IEEE Internet Computing14(5) (2010), 14–22.

Manvi

S.S.

, Shyam

G.K.

, Resource management for Infrastructure as a Service (IaaS) in Cloud computing: A survey, Journal of Network and Computer Applications41 (2014), 424–440.

Ibrahim

A.T.H.

, Ibrar

, Nor

B.A.

, Salimah

, Abdullah

, Samee

U.K.

, The rise of big data on Cloud computing: Review and open research issues, Information Systems47 (2015), 9–115.

Beloglazov

, Abawajy

, Buyya

, Energy-aware resource allocation heuristics for efficient management of data centers for Cloud computing, Future Generation Computer Systems28(5) (2012), 755–768.

10.

Khan

A.N.

, Kiah

M.M.

, Khan

S.U.

, Madani

S.A.

, Towards secure mobile Cloud computing: A survey, Future Generation Computer Systems29(5) (2013), 1278–1299.

11.

Islam

, Keung

, Lee

, Liu

, Empirical prediction models for adaptive resource provisioning in the cloud, Future Generation Computer Systems28(1) (2012), 155–162.

12.

Rong

, Nguyen

S.T.

, Jaatun

M.G.

, Beyond lightning: A survey on security challenges in Cloud computing, Computers & Electrical Engineering39(1) (2013), 47–54.

13.

Jennings

, Stadler

, Resource management in clouds: Survey and research challenges, Journal of Network and Systems Management23(3) (2015), 567–619.

14.

Jang

J.W.

, Jeon

, Kim

H.S.

, Jo

, Kim

J.S.

, Maeng

, Energy reduction in consolidated servers through memory-aware virtual machine scheduling, IEEE Transactions on Computers60(4) (2011), 552–564.

15.

Warneke

, Kao

, Exploiting dynamic resource allocation for efficient parallel data processing in the cloud, IEEE transactions on Parallel and Distributed Systems22(6) (2011), 985–997.

16.

Lombardi

, Di Pietro

, Secure virtualization for Cloud computing, Journal of Network and Computer Applications34(4) (2011), 1113–1122.

17.

Sharkh

M.A.

, Jammal

, Shami

, Ouda

, Resource allocation in a network-based cloud computing environment: design challenges, IEEE Communications Magazine51(11) (2013), 46–52.

18.

, Raghunathan

, Jha

N.K.

, A trusted virtual machine in an untrusted management environment, IEEE Transactions on Services Computing5(4) (2012), 472–483.

19.

Rahimi

M.R.

, Ren

, Liu

C.H.

, Vasilakos

A.V.

, Venkatasubramanian

, Mobile cloud computing: A survey, state of art and future directions, Mobile Networks and Applications19(2) (2014), 133–143.

20.

Yousafzai

, Gani

, Noor

R.M.

, Sookhak

, Talebian

, Shiraz

, Khan

M.K.

, Cloud resource allocation schemes: review, taxonomy, and opportunities, Knowledge and Information Systems50(2) (2016), 347–381.

21.

Khan , Nasir

Abdul

, et al., (BSS: block-based sharing scheme for secure data storage services in mobile cloud environment, The Journal of Supercomputing70(2) (2014), 946–976.

22.

Khan , Nasir

Abdul

, et al., A cloud-manager-based re-encryption scheme for mobile users in cloud environment: a hybrid approach, Journal of Grid Computing13(4) (2015), 651–675.

23.

Chiaraviglio , Luca , et al., An Approach to Balance Maintenance Costs and Electricity Consumption in Cloud Data Centers, IEEE Transactions on Sustainable Computing (2018).

24.

Shojafar , Mohammad , et al., FUGE: A joint meta-heuristic approach to cloud job scheduling algorithm using fuzzy theory and a genetic method, Cluster Computing18(2) (2015), 829–844.

25.

Wang

, Qiu

, Guo

, High reliable real-time bandwidth scheduling for virtual machines with hidden Markov predicting in tele health platform, Future Generation Computer Systems49 (2015), 68–76.

26.

Zuo

, Zhang

, Tan

, Self-adaptive learning PSO-based deadline constrained task scheduling for hybrid IaaS Cloud, IEEE Transactions on Automation Science and Engineering11(2) (2014), 564–573.

27.

Sridhar

, Hybrid genetic swarm scheduling for cloud computing, Global Journal of Computer Science and Technology15(3) (2015).

28.

Madhumathi

, Radhakrishnan

, A resource allocation strategy in cloud using Roulette Wheel Selection method, International Conference on in Green Computing and Internet of Things (ICGCIoT), (2015), 341–345.

29.

C.C.

, Wang

, An SLA-aware load balancing scheme for cloud data centers, International Conference on In Information Networking (ICOIN), (2014), 58–63.

30.

Shahapure

N.H.

, Jayarekha

, Load balancing with optimal cost scheduling algorithm, International Conference on In Computation of Power, Energy, Information and Communication (ICCPEIC), (2014), 24–31.

31.

Jang , Jae-Wan , et al., Energy reduction in consolidated servers through memory-aware virtual machine scheduling, IEEE Transactions on Computers60(4) (2011), 552–564.

32.

Aggarwal, Ritu. Resource Provisioning and Resource Allocation in Cloud Computing Environment, International Journal of Scientific Research in Computer Science, Engineering and Information Technology, (2018).

33.

Arivudainambi

and Dhanya

, Scheduling optimized secured virtual machine using cuckoo search and flow analyzer, Journal of computational and Theoretical Nanoscience, 14(16), (2017).

34.

Arivudainambi

and Dhanya

, Three phase optimization for qualified and secured VMs for resource allocation, International Journal of Enterprise and Network Management. Inderscience publisher, 9(3/4) (2018), 273–293.

Resource allocation through optimized three-phase scheduled VMs by grey wolf optimization and introspection security analysis

Abstract

Keywords

1 Introduction

2 Related works

3 Resource allocation in optimized and secured VM

3.1.1 Energy aware scheduling

4.1 Simulation result

Table 1 Efficient VMs selection for resource allocation Resource allocation strategies Number of VMs Total number of VMs 100 Energy aware scheduling 7 Memory aware scheduling 6 QoS based VMs 7 Qualified VMs 3 Secured VMs 1

References

Table 1
Efficient VMs selection for resource allocation

Resource allocation strategies Number of VMs

Total number of VMs 100

Energy aware scheduling 7

Memory aware scheduling 6

QoS based VMs 7

Qualified VMs 3

Secured VMs 1