节能型虚拟网络嵌入策略

173 浏览量更新于2024-08-29 收藏 2.43MB PDF 举报

"Energy Aware Virtual Network Embedding - IEEE/ACM TRANSACTION ON NETWORKING, VOL.22, NO.5, OCTOBER 2014 - SenSu, Zhongbao Zhang, Alex X. Liu, Xiang Cheng, Yiwen Wang, and Xinchao Zhao" 这篇研究论文“Energy-Aware Virtual Network Embedding”探讨了在网络虚拟化环境中如何实现节能的虚拟网络映射（Virtual Network Embedding，VNE）。VNE是将用户请求的虚拟网络映射到由互联网服务提供商维护的共享物理基础设施网络的关键功能。以往的工作主要关注于最大化服务提供商的收入，而没有考虑满足这些请求时的能源成本。随着能源成本占到物理网络运营成本的一半以上，对于基础设施提供者来说，在尝试接纳更多的虚拟网络请求时，降低能源消耗变得至关重要。在论文中，作者首次提出了一个能源成本模型，并将能源意识型的虚拟网络嵌入问题表述为一个整数线性规划问题。这一建模方法旨在优化能源效率，同时确保网络服务的质量和可靠性。为了有效地解决这个能源意识型VNE问题，论文提出了两种算法：一种基于启发式的算法和另一种近似算法。启发式算法通常能在较短的时间内提供接近最优解的解决方案，而近似算法则能够在保证性能的同时，减少计算复杂性，适应大规模网络环境的需求。这两种算法的设计目标是通过智能地分配和调度物理网络资源，如计算能力、存储和带宽，来最小化能源消耗。它们考虑了虚拟网络的拓扑结构、流量需求以及物理网络的物理特性，如设备的能效比，从而实现更节能的网络操作。此外，论文可能还深入分析了这些算法在不同工作负载和网络条件下的性能，通过模拟实验评估了它们的效率和节能效果。这有助于理解在实际部署中，如何根据具体场景选择合适的嵌入策略。 "Energy-Aware Virtual Network Embedding"论文为网络虚拟化领域带来了新的视角，强调了能源效率的重要性，并提供了实用的解决方案，以期在满足日益增长的虚拟网络需求的同时，降低网络运营的环境影响和经济成本。这项工作对云服务提供商、数据中心管理者以及网络优化研究人员具有很高的参考价值。

SU et al.: ENERGY-AWARE VIRTUAL NETWORK EMBEDDING 1609

Fig. 1. Example of VN embedding. (a) VN request. (b) Substrate network.

Fig. 2. Hourly electricity price of ﬁve locations in one week.

,where denotes t he set of routers and

denotes the set of hosts. Similar ly, the substrate links can also

be classiﬁed into two categories: the backbone link and the local

link, denoted by

and , respectively. Fig. 1(b) shows an

SN exam ple where circles and rectangles denote router and host

nodes, respectively.

For router nodes, we consider the following th ree attributes.

The ﬁrst attribute is the number of virtual rou ters t hat t he

physical router can suppo rt fo r deploying and runn ing different

personalized network protocols. In Fig. 1(b), the numbers that

are in parentheses and b eside each circle are such num bers.

The second attribute is the location that the router is located as

routers are generally geo grap hically distribu ted. For example,

in Fig. 1(b),

may be located in Los Angeles, CA, USA;

in Chicago, IL, USA; in New York, NY, USA; and

in New Jersey, USA. We use a two-dimensional coordin ate

to denote the location of nod e . The third

attribute is electricity price. The power markets of different lo-

cations are managed by different independen t system operators

(ISOs), and the ISOs are under competitive electricity mark et

structure. Therefore, different locations often have different

electricity prices. Even for the sam e location, the electricity

price m ay vary frequently over time [9], [12]. Fig. 2 show s the

hourly electricity price of the ﬁrst week of Septem ber 2011

available at [14] for ﬁve regions in the day-ahead market, in-

cluding the Eastern Hub of PJM (Pennsylvania–Maryland–New

Jersey), NP-15 Hub of CA I SO (Cali for nia), Capital Hub of

NYISO (New York), Mass Hub of ISO-NE (New England),

and Illinois Hub of MISO (Midwest). We observe that the elec-

tricity price varies over both location and time. To characterize

the spot price dynamics, in this paper, we use a discrete-time

model, which has a time window (e.g., an hour) of interest

.Weuse to denote the electricity price

for node

at time-slot , to denote th e arriving time of a VN

request, and

to denote the duration of the VN request being

served in the SN. Thus, the expiration time

of the VN request

For host nod es, we consider three attributes: CPU speed,

memory size, and storage capacity. In Fig. 1 (b), the triple

beside each host node denotes the values of these attributes.

For links, we consider bandwidth. In Fig. 1(b), the number

beside each link is the bandwidth. Note that t he modeling,

analysis, and algorithms in this paper can be easily extended

to incorporate o ther attributes, such as latency and jitter con-

straints, as well.

A VN is represented as a weighted graph

Here,

,where and denote

the set of virtual router and host nodes, respectively; and

,where denotes the set of virtual links

between any two virtual routers and

denotes the set of

virtual links between host nodes and routers. Fig. 1(a) shows

the VN requested by a user on the SN in Fig. 1(b).

We now formally deﬁne th e VN embedding problem. Given

a VN request

with a set of virtual n odes

and a set of virtual links ,andanSN

with a set of physical nodes and a set of

physical links

,embed on ,which

means to ﬁ nd two one-to-one mappin gs:

and . Here,

is a one-to-one mapping from to a s ubset of .Th

mapping includes tw o submappings,

and ,where

is from to a subset of ,and is from

to a subset of . For each virtual router node a

nd the

physical node

that it maps to, satisﬁes

both virtual router quan tity and location requirements of

For each virtual host node

and the physical n

ode

that i t maps to, satisﬁes the node requirements on

CPU speed, mem ory size, and storage capacity of

. Here,

is from to a subset of , which denotes a

ll loop-free

paths composed by the physical links in

. This mappin g

also includes two subm app ings,

and ,where

is from to a subset of , denoting a ll lo

op-free paths

between any two routers, and

is from to a subset of

, denoting all loop-free paths between hosts and routers.

For each virtual link

and the ph ys

ical path

that it

maps to, the bandwidth of each physical link in

is no

less than the bandw idth requirement of

. Take the em bed-

ding in Fig. 1 as an example. The no

de mapping solution is

and the link m apping solution is

Note that we focus on consideringoneISPinthispaper.

When multiple IS Ps collabora

te to provided VN services, as

each ISP knows only the characteristics of his own infrastruc-

ture, there are many technical challenges to address this issue.

The readers can refer to [15

] for more discussion on this topic.

B. Energy Co st Modeling

1) Node Energy Cost: Let

denote the additional

power consumption for mapp ing a virtual router

asubstraterouter

,and denote the additional

power for mapp ing a virtual host node

to a substrate

剩余13页未读，继续阅读

执念高

粉丝: 10
资源: 952

节能型虚拟网络嵌入策略

节能虚拟网络嵌入：30%能耗降幅与收入增长

深度学习驱动的青光眼筛查：Disc-aware Ensemble Network

云网络中多层应用的高可用性虚拟网络嵌入策略

PACE: Policy-Aware Application Cloud Embedding

v23 final.rar_Aware_WSN matlab_energy aware_energy-aware_routing

Cost-Aware Multi-Domain Virtual Data Center Embedding

An energy aware resource allocation for cognitive radio OFDMA systems

A Context-aware Attention Network for Interactive Question Answering

Disc-aware Ensemble Network for Glaucoma Screening from Fundus Image.pdf

QoE and Energy Aware Resource Allocation in Small Cell Networks with Power Selection, Load Management and Channel Allocation

最新资源