• Laser & Optoelectronics Progress
  • Vol. 60, Issue 17, 1706003 (2023)
Zhongjun Ma, Fengqing Liu*, and Yuxing Chen
Author Affiliations
  • School of Electronic and Optical Engineering, School of Flexible Electronics (Future Technology), Nanjing University of Posts and Telecommunications, Nanjing 210023, Jiangsu , China
  • show less
    DOI: 10.3788/LOP222310 Cite this Article Set citation alerts
    Zhongjun Ma, Fengqing Liu, Yuxing Chen. Research on Placement Algorithm of Flexible Virtual Machine in Elastic Optical Data Center[J]. Laser & Optoelectronics Progress, 2023, 60(17): 1706003 Copy Citation Text show less
    Placement of K-Means virtual machine algorithm. (a) Pipeline model virtual data center request after hose model transformation; (b) clustering result of K-means algorithm; (c) virtual machine placement to physical server in data center fat-tree topology
    Fig. 1. Placement of K-Means virtual machine algorithm. (a) Pipeline model virtual data center request after hose model transformation; (b) clustering result of K-means algorithm; (c) virtual machine placement to physical server in data center fat-tree topology
    Placement of the VT-VMPA algorithm. (a) Pipe model virtual data center request after the hose model transformation; (b) result of VT-VMPA algorithm clustering; (c) virtual machine is placed to the physical server in the data center fat-tree topology
    Fig. 2. Placement of the VT-VMPA algorithm. (a) Pipe model virtual data center request after the hose model transformation; (b) result of VT-VMPA algorithm clustering; (c) virtual machine is placed to the physical server in the data center fat-tree topology
    Average network bandwidth consumption of the three algorithms
    Fig. 3. Average network bandwidth consumption of the three algorithms
    Blocking rate of three algorithms with different number of virtual machines
    Fig. 4. Blocking rate of three algorithms with different number of virtual machines
    Average revenue of the three algorithms under different number of virtual machines
    Fig. 5. Average revenue of the three algorithms under different number of virtual machines
    Blocking rate of three algorithms under different service intensities
    Fig. 6. Blocking rate of three algorithms under different service intensities
    Average network bandwidth consumption of algorithms under different service intensities
    Fig. 7. Average network bandwidth consumption of algorithms under different service intensities
    Time revenue of the three algorithms under different service intensities
    Fig. 8. Time revenue of the three algorithms under different service intensities
    1:for 虚拟数据中心请求中的虚拟机集合V do
    2:根据虚拟机集合中虚拟机与其他虚拟机的业务量,降序排序形成虚拟机集合VM
    3:end for
    4:从集合VM中选取核心虚拟机Vcore,构成超级节点SuperV=Vcore,则剩余的虚拟机集VM=VM-Vcore
    5:找出Fat-Tree数据中心中满足VcoreCPU计算资源的物理节点,形成候选的服务器SselectVcore
    6:for 每一个物理服务器SiSselectVcore do
    7:计算权重Wweight,Si,并将服务器和服务器对应的权重以键值对的形式存入字典
    8:end for
    9:从字典中选择权值最小的物理服务器Smin用来承载核心虚拟机Vcore
    10:将已使用的物理服务器Smin添加到集合Ψ 中,将已映射的虚拟机Vcore加入Ψ集合中
    11:获取虚拟数据中心请求第一个子集合可合并的最大虚拟机数Umax以及当前服务器消耗的CPU资源量CpuTotal(Vcore)

    12:将虚拟请求中与Vcore相连但未做映射的虚拟机按照与Vcore的通信量降序得到VMsorted

    13:for 每一个虚拟机vjVMsorted do

    14:if RCPU,Vj+CpuTotal(Vcore)<PCPU,Smin and len(SuperV)<Umax then

    15:将该虚拟机映射到物理服务器Smin,修改CpuTotal(Vcore)=CpuTotal(Vcore)+RCPU,vj,在集合VM中删除vjSuperV=SuperVvj,将虚拟机vj加入到Ψ

    16:else

    17:退出第一个子集合中虚拟机的合并过程

    18:P*等于集合VM中剩余的虚拟机

    19:while len(P*)0 do

    20:置为P**

    21:for 每一个虚拟机vkP* do

    22:找出物理网络中满足vkCPU计算资源的物理节点,形成候选的服务器Sselectvk

    23:for 每一个物理服务器SkSselectvk do

    24:计算权重Wweight,Sk,存入字典中

    25:end for

    26:从字典中选择权值最小的物理服务器Smin*用来承载虚拟机vk

    27:将使用的物理服务器Smin*添加到集合Ψ,将已映射的虚拟机vk加入集合Ψ

    28:获取集合P*的CPU资源消耗CpuTotal(P*)

    29:if CpuTotal(P*)<PCPU,Smin* then

    30:将P*中的每一个虚拟机映射到服务器Smin*中,并将其加入到集合Ψ 退出虚拟机映射

    31:else

    32:while CpuTotal(P*)>PCPU,Smin* do

    33:从P*中删除对外务量最小的Vmin,并将其加入到集合P**

    34:for 每一个虚拟机vxP* do

    35:将P*中的每一个虚拟机映射到服务器Smin*中,将其加入到集合Ψ

    36:退出本次子集合中虚拟机的合并过程

    37:P*=P**

    38:if 虚拟数据中心请求中的所有虚拟机都映射成功 then

    39:return 映射成功,记录映射关系

    40:else

    41:return 映射失败

    Table 0. [in Chinese]
    1:将每一个子集合虚拟为超级节点,并根据管道模型虚拟数据中心的虚拟机间的业务关系转化为超级节点间的业务关系,存入字典中
    2:将字典中的元素按照超级节点间业务量的大小降序排序
    3:for超级节点间的每一个业务量do
    4:根据虚拟机映射阶段的结果找到Fat-Tree数据中心网络中对应的服务器,采用KSP 算法计算两服务器间的K条最短路
    5:从K条最短路中选择连续频谱隙资源最多的路径并判断其是否满足超级节点间的业务量所需的带宽资源,若满足,则将该路径赋值给最终选择的路径SelectPc,链路映射成功,否则链路失败
    6:end for
    7:if超级节点间的所有业务量映射成功 then
    8:return虚拟数据中心映射成功,分配物理网络资源,记录虚拟数据中心
    9:else
    10:return 虚拟数据中心映射失败
    Table 0. [in Chinese]
    步骤1:确定Vcore,构成当前超级节点SuperV=Vcore,将P*为空
    步骤2:将剩下的虚拟机集合V=V-Vcore,再为核心虚拟机寻找合适物理服务器SVcore
    步骤3:将与Vcore相连但没有映射的虚拟机按照与Vcore通信量降序得到VMsorted
    步骤4:将VMsorted中虚拟机合并到超级节点SuperV,确保超级节点的总CPU资源需求小于服务器SVcore的剩余CPU计算容量、总的虚拟机数量小于Umax
    步骤5:将剩余未映射的虚拟机放入P*,将P*作为当前超级节点,为其寻找物理服务器SP*
    步骤6:判断P*能否放入服务器SP*中,若能则虚拟机集合放置完毕。不能则需要不断从P*中选择对外业务量最小的Vmin放入P**中,直到P*能整体放入SP*
    步骤7:将P**赋值给P*,重复步骤5和6,直到所有虚拟机都完成映射
    Table 0. [in Chinese]
    Variable nameMeaning
    MTotal number of virtual machines in virtual center requests
    VsiThe number of virtual machines deployed on the server Si
    fx,yTraffic between virtual machines x and y
    mvj,mvkMap function,returns vjvk,mapped physical server
    bmvj,mvkCall a function that returns the hop distance between the physical server mvj and mvk to be mapped by the virtual machine to vj and vk
    nvnrtThe set of virtual network requests that have been successfully mapped over a period of time
    TnvnrDuration of virtual network request nvnr
    NnvnrvVirtual node set for virtual network requests nvnr
    bS,fmv2A function call that returns the hop distance between two physical nodes
    BS,fmv2A calling function that returns the number of bits that can be encoded per symbol,determined by the distance between physical nodes
    VVirtual machine collection
    nvnrMap the successful set of virtual network requests
    vBecause VMs are mapped to virtual nodes,VMs and virtual nodes can be represented
    Table 1. List of variable meanings
    Zhongjun Ma, Fengqing Liu, Yuxing Chen. Research on Placement Algorithm of Flexible Virtual Machine in Elastic Optical Data Center[J]. Laser & Optoelectronics Progress, 2023, 60(17): 1706003
    Download Citation