IJSMER201736 459 | P a g e
Cloud Computing Based New Architecture for Enhancement of Wireless Sensor Network
Prashant Sen ( Phd Research Scholar,V.B.S.P.U (U.P.))
Abstract— Cloud environments are mainly used for storage and processing of data. Cloud computing provides applications, platforms and infrastructure over the internet. Wireless Sensor Network is an very important technology in which sensor are placed in distributed manner to monitor physical and environment changes such as temperature, pressure etc. Combining these two technology helps in easy management of remotely connected sensor nodes and the data generated by these sensor nodes. For security and easy access of data, cloud computing is widely used in distributed/mobile computing environment. Wireless sensor network (WSN) is widely applied in many fields since its emergence. However, the limited resources of a sensor, especially limited battery life, limited bandwidth and limited processing power, are the main challenges for deploying and operating WSNs. This paper proposes a novel architecture based on cloud computing for wireless sensor network, which can improve the performance of WSN. Based on this architecture, a cloud acts as a virtual sink with many sink points that collect sensing data from sensors. Each sink point is responsible for collecting data from the sensors within a zone. Sensing data are stored and processed in distributed manner in cloud. Our simulation results show that the proposed architecture improves the performance of WSN, e.g., reduced packet transmission error rate, decreased number of end-to-end hops, and improved efficiency of energy consumption.
Keywords- wireless sensor network; data centric network; cloud computing; Hadoop
I. INTRODUCTION
Wireless sensor network (WSN) is a self- organizing network consisting of a lot of sensor devices that connect to others through wireless communication channel in multi-hop manner.
Sensors collect environment parameters and transmit sensing data back to a central management node (i.e., a sink node) for further processing. Nowadays, WSN has been applied in
many fields, such as environment monitoring, military, surveillance, disaster rescue and healthcare etc., and it is more widely used in the Internet of things. However, sensors are usually low cost devices equipped with limited resources, e.g., processing power, memory, wireless bandwidth, and battery. The design of WSN must take care of these constrains, especially the battery limitation that determines the lifetime of a sensor and a sensor network. In literature, a number of energy efficient techniques have been proposed. All layer protocols used in WSN are optimized to reduce energy consumption, including MAC layer, network layer [1], transport layer, and cross-layer approaches. Since transmitting operation consumes more energy comparing against sensing and processing operations, some other technologies have been proposed to save energy, e.g., in-network data processing, mobile sink for data collection, and topology reorganization [2]. Cloud computing is considered as a cost efficient way of IT resource provision manner. It offers a number of advantages such as scalability, agility and economy efficiency, in comparison of traditional IT infrastructure. The cloud computing technology invented by Google is suitable for big data storage and processing. The Hadoop, an open source implementation of Google’s cloud by Apache, is widely adopted by many companies.
In this paper, we propose a novel architecture based on cloud computing for wireless sensor network. In this architecture, a cloud acts as a virtual sink and collects sensing data in multiple points. Therefore, the WSN is naturally divided into units which we refer as “zones”. Each node in the cloud is responsible for data collection of sensors in a zone. The sensors in a zone are organized as a local WSN in flat or hierarchy topology, and these local networks are integrated together by the cloud. As a result, the average end-to-end path length of packet transmission could be shortened, and the energy consumption would be reduced. The required bandwidth for
IJSMER201736 460 | P a g e
data transmission is reduced as well. Moreover, data in cloud are stored and processed in distributed manner, thus complicated tasks could be completed timely, which would be preferred for large volume data process. The rest of the paper is organized as follows. In section II, we present literature survey. In section III, we outline our new architecture based on cloud computing. In section IV, we present simulation results and conclude the paper in final section.
2. LITERATURE SURVEY
Peter et. al. [3] discusses the idea of combining wireless sensor networks and cloud computing from the existing approach. The paper proposes wireless sensor network by virtual sensor in the cloud. The idea is to store the data on both the real sensor and virtual sensor. The paper proposes architecture to realize distributed shared memory in WSNs with the help of a middleware called tiny DMS. Tongrang Fan et. al. [4] proposes sensor data storage solutions based on the Hadoop cloud computing framework. Due to the rapid growth of sensor data storage and processing, traditional storage systems are not able to meet the data access requirements. By contrast to the private cloud system storage and traditional storage model, the characteristics and advantages of private cloud system storage are analyzed. The designed storage solutions were performed by Map Reduce programming model, and the experimental results indicated that the new cloud storage solution had higher data access performance. Geoffrey et. al. [5] discusses the characteristics of distributed cloud computing infrastructure for collaboration sensor-centric applications on the Future Grid. The paper mainly focuses on the performance, scalability and reliability at the network level using standard network performance tools. Sajjade t. al. [6]
proposes a new framework for Wireless Sensor Network integration with Cloud Computing model with a possibility of an existing Wireless Sensor Network getting connected to the proposed framework. The integration controller unit of the proposed framework integrates the sensor network and cloud computing technology which offers reliability, availability and extensibility. Khandakar et. al. [7] discusses novel
and attractive solutions for information gathering across transportation, business, health-care, industrial automation, and environmental monitoring. The next generation of WSN will benefit when sensor data is added to blogs, virtual communities, and social network applications.
The proposed architecture is based on cloud computing for wireless sensor network, which can improve the performance of wireless sensor network. Based on this architecture, a cloud acts as a virtual sink with many sink points that collect sensing data from sensors. Each sink point is responsible for collecting data from the sensors within a zone. Sensing data are stored and processed in distributed manner in cloud. Rajeev Piyareet. al. [8] discusses the Wireless Sensor Network applications in important areas such as healthcare, military, critical infrastructure monitoring, environment monitoring, and manufacturing. Due to some of the limitations of WSNs in terms of memory, energy, computation, communication, and scalability, efficient management of the large number of WSNs data in these areas is an important issue. Cloud computing provides huge computing, storage, and software services in a scalable and virtualized manner at low cost. This paper proposes architecture for integrating Wireless Sensor Networks with the Cloud. REST based Web service is used to monitoring e-health care service and smart environment. Data can be accessed from anywhere due to using of IP in REST based Web service. Sudarsan et. al. [9] presents a Secured Wireless Sensor Network- integrated Cloud Computing for Life Care. This is used to monitors human health, activities, and shares information among doctors, care-givers, clinics, and pharmacies in the Cloud so that users can have better care with low cost. It incorporates various technologies with novel ideas including;
sensor networks, Cloud computing security, and activities recognition. Ahmed Lounis et. al. [10]
focuses on the major issue of data management in Wireless Sensor Networks. The huge amount of data generated and collected by medical sensor networks introduces several challenges for the existing architectures. This paper proposes an innovative architecture for collecting and accessing large amount of data generated by medical sensor networks. This architecture aims
IJSMER201736 461 | P a g e
at resolving the challenges and makes easy information sharing between healthcare professionals. Giancarlo Fortino et. al. [11]
presents Body Cloud, a system architecture based on Cloud Computing for the management and monitoring of body sensor data streams. It system includes scalability and flexibility of resources. An important application is monitoring human condition using distributed sensor nodes. A network of body sensors in a group of people generates large amounts of data that should be stored and processed. Cloud computing can provide a powerful, scalable storage and processing infrastructure to perform both online and offline analysis and mining of body sensor data stream.
David Tracey et. al. [12] discusses about Wireless Sensor Network and cloud computing. Wireless Sensor Networks (WSNs) consist of large number of applications and services to interact with the physical world. This paper uses cloud services and big data approaches to store data and to analyse data. With the cloud and the big data approach, scalability and availability have increased, which will be needed for many of the devices in the Internet of Things (IoT). The paper also proposes an architecture to integrate information of WSNs and their services and it help in flowing of data from sensor through to services. RazviDoomunet. al. [13] defined a solution for the global attack problem on the overall data transmissions and the existing approaches do not provide sufficient privacy when the attacker can visualize the transmissions and infer contextual information. The Paper proposes
SECLOUD: Source and Destination Seclusion using Clouds to confuse the true source/destination nodes and make them indistinguishable among a group of neighbor nodes which works well even under network- wide traffic visualization by a global attacker.
Rosangelaet. al. [14] present the main aspects of a prototype to test Hadoop framework applied to process climate related data sets in a cloud computing infrastructure. Hadoop frame work is used to handle big data normally in unstructured forms, such as text, sensor data, audio, video, log
files, and more. Climate and sensor network data are applied to assist the decision making process in agricultural production.
Recently, new types of WSN, such as wireless multimedia sensor network (WMSN) and wireless sensor and actor network (WSAN), are emerging. In WMSN, more powerful sensors, for example, image sensor, audio sensor, and video sensor, etc., are employed to capture rich information of monitoring area. In WSAN, the actor, which is usually a robust and can move around, is introduced into WSN. During its movement, it is capable of communicating with scalar sensors and reacts to environment changes.
In these new sensor networks, the resource challenges (especially bandwidth and energy) are even severer since more data are captured and needed to be processed and transmitted. Data compression, efficient image data coding, cross- layer communication protocols, smart activity scheduling, and energy-efficient routing techniques are typical ways to alleviate problems faced[15].
Cloud Computing enables the possibility of wide deployment of rich services in mobile devices.
Mobile Cloud Computing (MCC) refers to an infrastructure and mechanisms for providing computation and data storage capabilities to mobile devices based on cloud computing.
MCC is seen as the future of mobile [16].
III.CLOUD COMPUTING BASED ARCHITECTURE
A.Architecture
The architecture is shown in Fig 1. A number of special nodes (sink points) distributed across the WSN area constitute a cloud. We refer this type node as “cloud node”, which is equipped with more resources. Furthermore, a cloud node acts as a sink for sensors nearby.
Therefore, the architecture is naturally cluster based. In order to differentiate from the term
“cluster” in traditional WSN, we refer a cluster as a “zone” in this paper. These sensors in a zone are organized as independent local WSNs (LWSN),
IJSMER201736 462 | P a g e
and all LWSN are integrated together by the cloud. The cloud can be viewed as a “virtual sink” for the whole sensor network.
B. Organizing Sensors in a Zone
We present a different way of heterogeneous sensors organization, as shown in Fig 2. The homogeneous sensors form a WSN (a different type sensor could be used as a software is deployed in the cloud. After processing data from all sensors, the cloud can designate sensors to perform properly through the accessed sink point.
If the size of a zone is reasonable, the scheduling commands from the sink point could reach the destination sensor timely. This way of organization is different from the single-tier clustered architecture of WMSN. First, the independent WSN is built in a zone of which the size is smaller, instead of the whole sensor network. Second, the cloud can control activities of a sensor timely based on data collected from the whole WSN.
Figure 1. Cloud computing based WSN architecture
C. Organizing Cloud
As considering the organization of nodes to a cloud, the bandwidth is still one of the most important factors, especially when they communicate through wireless channel. Hadoop is adapted to the infrastructure built on smart phones, which has the similar requirements as here. In fact, Hadoop aims for big data storage and process, and data movement reduction is one of its design goals. One of its design criteria is
“moving computation is easier than moving data.”
We suggest the cloud is organized in the Hadoop
way. Both of Hadoop’s storage system (Hadoop Distributed File System, HDFS) and data processing system (Map-Reduce framework) have a master/slave architecture, in which the master is responsible for storing file meta data or scheduling jobs, and slaves are responsible for storing file content or task execution (task is one piece of split job). The two masters could be deployed either in the same physical node or in different physical nodes. In its storage system, when a client accesses a file, it firstly contacts the master to retrieve the file’s meta data (file location, for example), then it directly retrieves the content from the slaves which store a part data of the file. In its data processing system, the code used is submitted to the master, and then the master distributes the code to the slaves which store the processed data, or to those slaves near by the slaves storing the processed data.
As a result, the data is processed almost locally and the data movement (thus bandwidth requirement) is reduced. However, data movement cannot be avoided in distributed processing system. Hadoop supports data compression mechanism to reduce bandwidth cost as much as possible. Therefore, a master node should be introduced to gateway) and send their data to the sink. Each type sensors form a logical independent (and physical overlapped) WSN, and these WSNs access the same sink. Data processing build the cloud. The master node usually connects to the Internet and is the access point of the system.
It should be clear in mind that Hadoop cannot be adopted directly in the proposed architecture, just as in Hyrax. The configuration parameters, even the code, should be optimized. In summary, in the architecture a cloud node is not only a slave of the Hadoop cloud, but also the sink point of a zone. It receives data sent by sensors in the zone, and then acts as a client to store these data into Hadoop. By optimized the Hadoop code, data movement can be avoided, and the client just creates or updates file meta data in master. Data processing requirement and code are sent to the master, and the master schedules the job (the requirement) to run on appropriate slaves. Fig 3 shows an
IJSMER201736 463 | P a g e
example that cloud nodes communicate with others by Ad-hoc manner to form the virtual sink cloud increased as the number of sink points increases. The delays and hops are of the packets successfully reaching a sink point. In addition, these delays greater than 0.2 second are filtered to make the illustration be clear. The percent of filtered delays is less than 10%.
Figure 2 Organizing sensors in a zone
IV.SIMULATIONS
We perform simulations to verify the performance of WSN in the proposed architecture. Since the performance of the cloud is dependent on the algorithms used in data processing and the communication way hired, we focus only on the performance of packet transmission of sensors. We employ the popular simulator NS2. The scenarios are that sensors are distributed in regular way (grid) in a 200mX200m area with each sensor distanced from others in 10m in both X and Y direction, and there is(are) 1, 4 or 8 sink points deployed uniformly in the area. All nodes constitute an Ad-hoc network communicating by IEEE 802.15.4 . The range of radio frequency is 15m and the route protocol is AODV. CBR or Poisson traffic generator, which starts randomly in the first 10 seconds of simulation, is deployed in sensors, and each sends packet to the nearest sink point. The packet length is 70 bytes. The CBR generator generates a packet every 5 second, and the average arrival rate of Poisson generator is 5 second as well. All simulations last for 200 seconds.
Figure 3. The organization of a cloud, each node in cloud acts as a cloud node and a sink point.
Cloud node reaches the master in Ad-hoc manner.
Table 1 and Table 2 show the statistics. The last column represents the total number of sending operations of traffic packets divides the number of packets successfully reaches a sink point, and it indicates the efficiency of sending operation thus the energy consumption. Other columns are self explained. It is clear that as the number of sink points increases, the transmission performance of the network is improved greatly.
The performance improvement can further be supported by Fig 4 and Fig 5, which are distribution of end-to-end delays and hops, respectively.
TABLE I. STATISTICS OF CBR TRAFFIC SIMULATIONS
# of sink poin
ts
# of genera
ted packet
s
# of sendin
g operat
ions
# of successf
ul sending packets
succes s ratio
avg.
hops per succe ss packe
avg.
delay per success
packet
（sec）
avg. # of sending operatio ns per success packet 1 5336 44444 3588 67.24
%
8.03 0.2861 12.39
4 4424 24010 3558 80.43
%
5.10 0.2054 6.75
8 10571 54552 10394 97.90
%
4.95 0.1056 5.25
IJSMER201736 464 | P a g e
TABLE II.STATISTICS OF POISSON TRACFFIC SIMULATIONS
Figure 4. Delay distribution of simulations
Figure 5. Hop Distribution of Simulations
5. DISCUSSION
To identify advantages and drawbacks of the proposed combination of wireless sensor networks and cloud computing environments we discuss it in terms of i) development issues , ii) scalability, iii) interoperability, iv), security, v) law, vi) business issues. The development of such systems has the advantage, that several programming languages and different kind of devices can be used and easily interconnected. It provides a lot of flexibility in terms of programming languages and devices, but this come with the drawback of heterogeneity, which quite often leads to more complexity. The scalability of this approach seems to be unlimited, since wireless sensor networks operate independently and in a mesh network, and are
# of sink poin
ts
# of gener ated packet
s
# of sendin
g operat
ions
# of successf
ul sending packets
succes s ratio
avg.
hops per succes
s packe
t
avg.
delay per success packet
（sec）
avg. # of sending operatio ns per success
packet
1 16711 247622 10873 65.06
%
13.02 0.0892 14.82
4 13569 94566 12674 93.40
%
6.40 0.0399 7.46
8 12168 63837 11499 94.50
%
4.87 0.0297 5.55
IJSMER201736 465 | P a g e
connected to the cloud computing environment through a scalable number of wireless sensor network communication hubs.
The cloud computing environment itself offers a scalable infrastructure, which makes it very attractive. The main bottleneck in this architecture is the remote communication between the wireless sensor network communication hubs and the cloud computing environment. The local computation within the cloud computing environment is usually fast, but the remote communication between the hub and the cloud computing environment is much slower, which has to be considered in advance and further investigated in the future [17].
The interoperability between different kinds of devices, services, and programming platforms is organized on the level of web services interfaces, which provides a good level of abstraction and is a proven concept. But it must be noted, that the necessary interoperability measures usually lead to some overhead, which may have an impact on the overall system performance. Several measures are taken to ensure security of the sensed, processed and stored data. The data store and processed in cloud computing infrastructure are protected by authentication and authorization measures, and can if needed be encrypted. But Foster [18] mentioned, that currently, the security model for Clouds seems to be relatively simpler and less secure than the security model adopted by Grids. The sensor data transferred from the wireless sensor network communication hub to the cloud computing infrastructure can be secured by TLS/SSL, which is a proven technology. Last but not least can the data communication within the wireless sensor network also be secured with a 128-bit AES encryption. Since motes in a wireless sensor network have limited energy resources (i.e. are battery powered), and they sometimes don`t process in some cases mission critical data, it`s possible to omit the encryption overhead if necessary.
6. CONCLUSION
We have presented a novel way of combining wireless sensor networks with cloud computing services. The wireless sensor networks are used
to sense and collect environmental data. Since wireless sensor networks are limited in their processing power, battery life and communication speed, cloud computing usually offers the necessary storage capacity and processing power for long term observations, analysis and use in different kind of environments and projects, since the basic infrastructure remains the same. The presented concept and implementation must now be further evaluated in different use cases to measure its advantages and shortcomings. In this paper, we propose a new architecture for wireless sensor network based on the cloud computing platform. In the architecture, a number of specific nodes are carefully distributed across the WSN area, and they form a cloud computing platform by Hadoop. The cloud acts as a virtual sink and has multiple sink points which could collect sensing data from WSN. Therefore, the WSN is naturally divided into a number of zones. The sensors in a zone could be organized in flat or hierarchy way as in traditional WSN.
We also propose a new way of sensor organization in a zone: the homogeneous sensors form a WSN, while heterogeneous sensors form logical independent but physically overlapped WSNs. Sensors could be managed by the cloud through sink point accessed. All WSNs in zones are integrated together by the cloud. Sensing data in cloud are stored and processed in distributed manner. Simulations results show that the transmission performance of a WSN built in such architecture is improved greatly. The ratio of packets successfully reaching the sink is increased, the number of end-to-end hops is reduced, and the end-to-end delay is shortened.
The average number of sending operations for a packet being sent successfully is reduced as well. This indicates that the efficiency of sending operation (thus the energy consumption) is improved.
However, the WSN in this architecture is a structured network in which some nodes (at least the cloud nodes) are needed to be carefully managed. Further work should investigate the integration of predefined complex filters and different classifiers and define a suitable ontology
IJSMER201736 466 | P a g e
for semantic reasoning, so that data analysis can be done within the cloud computing environment.
REFERENCES
[1] G. H. Raghunandan and B. N. Lakshmi, "A comparative analysis of routing techniques for wireless sensor networks," in 2011 National Conference on Innovations in Emerging Technology (NCOIET), 2011, pp. 17-22.
[2] B. Ridong, "Topological optimization based on small world network model in wireless sensor network," in 2011 2nd International Conference on Control, Instrumentation and Automation (ICCIA),
2011, pp. 254-257.
[3] Peter Langendoerfer, Krzysztof Piotrowski, Manuel Diaz, Bartolome Rubio, “Distributed Shared Memory as an Approach for Integrating WSNs and Cloud Computing”, 5th International Conference on New Technologies, Mobility and Security, Istanbul, 7-10 May 2012, pp 1-6, DOI:
10.1109/NTMS.2012.6208723.
[4]Tongrang Fan, Xuan Zhang, FengGao, “Cloud Storage Solution for WSN Based on Internet Innovation Union”, International Journal of Database Theory and Application, Vol 6, No. 3, June 2013, pp 49-58.
[5]Geoffrey C Fox, Alex Ho, Eddy Chan Anabas,
“Measured Characteristics of FutureGrid Clouds for Scalable Collaborative Sensor-Centric Grid Applications”, International Conference on collaboration Technologies and systems, Philadelphia, 23-27 May 2011, pp 151-160, DOI:
10.1109/CTS.2011.5928677.
[6] SajjadHussain Shah, FazleKabeer Khan, Wajid Ali, Jamshed Khan, “A New Framework to Integrate Wireless Sensor Networks with Cloud Computing”, IEEE Aerospace Conference, Big Sky, MT, USA, 2-9 Mar 2013, pp 1-6, Print
ISBN: 978-1-4673-1812-9, DOI:
10.1109/AERO.2013.6497359.
[7] Khandakar EntenamUnayes Ahmed, Mark A
Gregory, “Integrating Wireless Sensor Networks with Cloud Computing”, 7th International Conference on Mobile Ad-hoc and Sensor Networks (MSN), Beijing, 16-18 Dec 2011, pp 364-366, Print ISBN: 978-1-4577-2178-6, DOI:
10.1109/MSN.2011.86
[8] Rajeev Piyare, Sun Park, Se YeongMaeng, Seung Chan Oh, Sang Gil Choi, Ho Su Choi, Seong Ro Lee, “Integrating Wireless Sensor Network into Cloud Services for Real-time Data Collection, International conference on ICT Convergence [ICTC], 14-16 Oct 2013, Jeju, pp752-756, DOI : 10.1109/ICTC.2013.6675470.
[9] Sudarsan Rao Vemuri, Dr. N Satyanarayana, V Lakshmi Prasanna, “Generic Integrated Secured WSN-Cloud Computing For U-Life Care”, International Journal of Engineering Science and Advanced Technology, Vol 2, Issue 4, pp 897-907, Aug 2012, ISSN: 2250-3676.
[10] Ahmed Lounis, AbdelkrimHadjidj, AbdelmadjidBouabdallah, YacineChallal,
“Secure and Scalable Cloud-based Architecture for e-Health Wireless sensor networks”, 21st International Conference on Computer Communications and Networks (ICCCN), Munich, 30 July-2 Aug 2012, pp 1-7, Print ISBN:
978-1-4673-1543-2, DOI:
10.1109/ICCCN.2012.6289252.
[11] Giancarlo Fortino, MukaddimPathan, Giuseppe Di Fatta, “Body Cloud: Integration of Cloud Computing and Body Sensor Networks”, IEEE 4th International Conference on Cloud computing Technology and Science, Taipei, 3-6 Dec 2012, pp 851-856, DOI:
10.1109/CloudCom.2012.6427537.
[12] David Tracey, Cormac Sreenan, “A Holistic Architecture for the Internet of Things, Sensing Services and Big Data”, 13th IEEE/ACM International Symposium on Cluster, Cloud, and Grid Computing, Delft, 13-16 May 2013, pp 546- 553, Print ISBN: 978-1-4673-6465-2, DOI:
10.1109/CCGrid.2013.100.
[13] RazviDoomun, ThaierHayajneh, Prashant Krishnamurthy, David Tipper, “SECLOUD:
IJSMER201736 467 | P a g e
Source and Destination Seclusion Using Clouds for Wireless Ad hoc Networks”, IEEE Symposium on Computers and Communications, Sousse, 5-8 July 2009, pp 361-367, Print ISBN:
978-1-4244-4672-8, DOI:
10.1109/ISCC.2009.5202367.
[14] Rosangela de Fátima Pereira, Marcelo Risse de Andrade, ArturCarvalhoZucchi, Karen Langona, Walter Akio Goya, Nelson Mimura Gonzalez, Tereza Cristina MeloBrito de Carvalho, Jan-Erik Mangs, AzimehSefidcon,
“Distributed processing from large scale sensor network using Hadoop”, IEEE International Congress on Big Data, Santa Clara, CA, 27 June- 2 July 2013, pp 417-418, Print ISBN: 978-0- 7695-5006-0, DOI:
10.1109/BigData.Congress.2013.64.
[15] C. Chen, B. Tian, Y. Li, and Q. Yao, "Data aggregation technologies of wireless multimedia sensor networks: a survey," in 2010 IEEE International Conference on Vehicular Electronics and Safety (ICVES), 2010, pp. 83- 88.
[16] H. T. Dinh, C. Lee, D. Niyato, P. Wang, "A survey of moible cloud computing –architecture, applications and approaches" in Wireless Communication and Mobile Computing, 2011.
[17] F. Douglas, A. Mirno C. Loosley, High- Performance Client/Server.: John Wiley and Sons Inc., 1997.
[18] Ian Foster, Yong Zhao, Ioan Raicu, and Shiyong Lu, "Cloud Computing and Grid Computing 360-Degree Compared," Proc. Grid Computing Environments Workshop GCE '08, pp.
