A View at Desktop Clouds

Abdulelah Alwabel, Robert Walters and Gary Wills

Electronics and Computer Science School, University of Southampton, Southampton, U.K.

Abstract. Cloud has emerged as a new computing paradigm that promises to

move into computing-as-utility era. Desktop Cloud is a new type of Cloud

computing. It merges two computing models: Cloud computing and volunteer

computing. The aim of Desktop Cloud is to provide Cloud services out of infra-

structure that is not made for this purpose in order to reduce running and

maintenance costs. This paper discusses this new type of Cloud by comparing it

with current Cloud and Desktop Grid models. It, also, presents several research

challenges in Desktop Cloud that require further attention.

1 Introduction

Desktop Clouds represent a new direction of providing Cloud services based on non-

dedicated resources. The resources can be any form of computing devices such as

PCs, laptops …etc. The new type attempts to combine two computing models, name-

ly Cloud computing and Volunteer computing in order to form a Cloud that provides

services for less or no cost. Throughout this paper, Traditional Cloud (TC) refers to a

Cloud that relies on dedicated resources to provide services, whereas Desktop Cloud

(DC) refers to a Cloud that relies on non-dedicated resources. Amazon Cloud, for

instance, is a Traditional Cloud.

The remaining of this paper is organised as follows. First, the paper starts by giv-

ing an overview about the meaning of Desktop Clouds. In addition, the advantages of

Desktop Clouds are presented. A brief comparison study between Desktop Clouds,

Traditional Clouds and Desktop Grids is presented in order to study similarities and

differences. Finally, the paper finishes by discussing research issues in Desktop

Cloud.

2 Desktop Clouds

The success of Desktop Grids motivates the idea of harnessing idle resources to build

Desktop Clouds. Hence, the term Desktop comes from Desktop Grids because both of

Desktop Clouds and Desktop Grids are based on Desktop PCs and laptops etc. Simi-

larly, the term Cloud comes from Cloud as Desktop Cloud aims to provide services

based on the Cloud business model. Several synonyms for Desktop Cloud have been

Alwabel A., Walters R. and Wills G..

A View at Desktop Clouds.

DOI: 10.5220/0004983400550061

In Proceedings of the International Workshop on Emerging Software as a Service and Analytics (ESaaSA-2014), pages 55-61

ISBN: 978-989-758-026-0

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

used, such as Ad-hoc Cloud, Volunteer Clouds and Non-Dedicated Clouds. The liter-

ature shows that very little work has been undertaken in this direction.

Fig. 1. Architecture of desktop clouds.

The overview architecture of Desktop Clouds is depicted in fig. 1. The architec-

ture is consisted of several layers. The users contact the service layer in order to sub-

mit their demands. The physical layer is responsible of managing physical nodes that

are aggregated the resource layer. The virtual layer plays a curtail role in terms of

isolating Clients request from the physical nodes via virtaulisation. Users are assigned

virtual machines that are located in physical machines. Physical machines can be

connected by LAN or WLAN.

Ad-hoc Cloud is the idea of harvesting distributed resources within an organisa-

tion to form a Cloud [1]. Nebula [2, 3] is a project aiming to exploit distributed re-

sources in order to create a volunteer Cloud which offers services free of charge.

Cloud@home [4, 5] is a project representing the @home philosophy in Cloud compu-

ting. The goal of Cloud@home is to form a new model of Cloud computing contrib-

uted to by individual users over the Internet. In addition to that, Cern

has announced

an initiative to move their Desktop Grid project, which is called LHC@home, toward

the Cloud [6]. It is suggested that non-dedicated resources can be exploited by Cloud

providers in case their local infrastructure cannot meet requests by consumers at peak

times [7].

2.1 Scenario

Suppose a group of universities wishes to benefit from its computing resources to

form a Cloud. The resources range from PCs to servers etc, each of them is called a

the European Organization for Nuclear Research

Cloud node. A node can join the Cloud when it becomes idle. This scenario is moti-

vated by Condor [8]. Users in Desktop Cloud submit their request to acquire services

with requirements as it is stated in the service level agreement between a client and

the Cloud interface. The requests are processed in the virtualisation layer on top of

Cloud physical nodes. The virtualisation isolates the guest operating system from the

host physical machine. The isolation improves security and prevents unauthorised

access between two parties.

Another scenario that can be considered is a universal Desktop Cloud which al-

lows people to contribute their own computing resources to be used by Cloud clients

[9]. This example can be considered as public Desktop Cloud. The people are asked

to contribute their machines in order to form a Desktop Cloud. People can be stimu-

lated to participate in DC to serve science within research communities.

Table 1. Desktop Cloud vs. Traditional Cloud.

Feature Desktop Cloud Traditional Cloud

Elasticity √ √

Virtualisation √ √

Idle Resources √ X

Ease of Use √ √

3 Desktop Clouds vs. Traditional Clouds

This section clarifies Desktop Clouds further by comparing them with related areas:

Traditional Clouds, Grids and Desktop Grids. There are basic differences between

Desktop Cloud and Traditional Clouds as it is shown in Table 1. Elasticity means that

a consumer can acquire computing services. Then the user can scale up or down ac-

cording to their needs. Both Traditional Cloud and Desktop Cloud rely heavily on

virtualisation. The ease of use principle means that users can use a specific service

without making a lot of changes to their work. Both Traditional Clouds and Desktop

Clouds let their users harness services without making significant changes to their

code. However, there are several differences between Desktop and Traditional

Clouds. Firstly, the infrastructure of DC is made of resources that are non-dedicated,

i.e. not dedicated to be part of Cloud infrastructure. In the contrary, the infrastructure

of TC is made of a huge number of dedicated computing resources. Secondly, the

resources in DC are highly distributed across the globe, whereas they are limited in

TC to several locations in data centres. Furthermore, nodes in DC are highly volatile

due to the fact that they can be down unexpectedly without prior notice. Resource

high volatility has negative impact on availability and performance [10]. Although,

resources in both TC and DC are heterogeneous, they are even more heterogeneous

and dispersed in DC. Virtualisation plays a key role in Desktop Clouds which makes

it different from other large scale systems.

Desktop Clouds have some advantages over Traditional Clouds. Firstly, Tradi-

tional Clouds have a negative impact on the environment since their data centres

consume massive amounts of electricity [11]. The second advantage is cost effective-

ness of Desktop Clouds for both Cloud services providers and their consumers. For

service provider, there is no need to build new data centres to meet the increasing

demands of the future. Consumers will get their services at lower prices, if not free

compared to Traditional Clouds. Also, Desktop Cloud helps in reducing energy con-

sumption since it utilises already-running undedicated resources which would other-

wise remain idle. Some studies show that the average percentage of local resources

being idle within an organisation is about 80% [12]. Furthermore, Traditional Clouds

are formed from a limited number of data centres located around the globe. There-

fore, they are inefficient in terms of data mobility and pay little attention to the loca-

tion of clients [3]. Finally, Traditional Clouds are centralised, which leads to the po-

tential that there could be a single point of failure issue if a Cloud service provider

goes out of the business. In contrast, Desktop Clouds manage and offer services in a

decentralised manner.

Table 2. Desktop Cloud vs. Traditional Cloud.

Feature Desktop Cloud Desktop Grid

Elasticity √ X

Virtualisation √ X

Idle Resources √ √

Ease of Use √ X

4 Desktop Cloud vs. Desktop Grid

Desktop Clouds can be confused with other distributed systems, especially Desktop

Grids. Table 2 shows a comparison between DC and Desktop Grid. Both Desktop

Clouds and desktop Grids serve the same goal that is exploiting idle computing re-

sources denoted by the public. However, Desktop Grid cannot offer services to clients

in elasticity way as in Desktop Clouds. Elasticity means that users can require more

computing resources in short term. Users, in Desktop Grids, are expected to know in-

depth details about the middleware used in order to be able to harness the offered

services [13]. Desktop Grids do not employ virtualisation to isolate users from the

actual machines. People who wish to contribute their computing machines need to

install a specific software in order to join a Desktop Grid.

5 Research Challenges

This section discusses several research issues that need further attention. Some of

these challenges are inherited from Cloud computing while others are driven by the

nature of used resources being highly volatile.

5.1 Security

Security is one of the major concerns that prevent organisations from moving to the

cloud [14]. Ristenpart et al. show that an attacker can uncover the actual location of a

particular virtual machine (VM) [15]. Then, a cross-VM side channel attack can re-

veal critical information about the targeted VM by placing a malicious VM on the

same physical machine.

More worries arise in DCs where both consumers and contributors are from the

public. Therefore, security can be a major issue in this context. In addition to the

previous threats presented in the cloud, both consumers and contributors take on risk

themselves when they join a DC. A contributor can put his own data at risk by allow-

ing access to a virtual image located in his machine. Likewise, consumers are vulner-

able to malicious contributors. Nodes in DC are more likely to be vulnerable to out-

side attacks due to weaknesses in local antivirus software and firewalls.

Virtualisation can be vital in order to isolate the host completely from guest oper-

ating systems and, thus, prevent any unwanted access from either party. Trust mecha-

nisms can be employed in this matter. For example, a DC can maintain a behaviour

table which contains information about both consumers and contributors. The table

can be used to decide which parties are trustworthy enough to join the cloud. Fur-

thermore, VCCs should rely on autonomous mechanisms such as sandbox or certifi-

cation in order to prevent various attacks from participants [16].

5.2 Resource Management

Resources in volunteer clouds are highly heterogeneous, therefore, managing them

can be considered problematic. Virtualisation plays a key role in Desktop Clouds

because it virtualises contributed resources and delivers them to users as VMs. DCs

face a challenge of developing a resource allocation mechanism that is able to: a)

manage non-dedicated, heterogeneous resources, b) deliver a virtualized machine to

upper level in DCs and c) work closely with users’ tasks in order to find most suitable

nodes for each request.

It has been pointed out that lacking central management in DCs cause a major is-

sue in terms of reliability and state maintenance in case of failures [17]. The infra-

structure of DC is consisted of nodes that are highly volatile. Therefore, fault recov-

ery mechanisms are crucial in order to improve reliability in this environment [10]. In

addition, volunteer clouds require means to interact with other clouds for data migra-

tion or to gain extra computing resources [4].

5.3 Quality of Service

VCCs are expected to offer services at a low level of reliability and availability due to

the fact that they depend on unreliable volunteered resources which can join or leave

the cloud without prior knowledge for various reasons [7]. Availability of individual

nodes is considered a primary issue in VCCs [10]. For example, it is estimated that

resource unavailability can reach up to 50% in volunteer projects [18]. Availability of

each individual node can affect the service quality of VCC. Andrzejak et al. [19]

propose a technique to predict the availability of a group of high volatility resources.

6 Conclusion

This paper has introduced Desktop Clouds as being a new type of Cloud computing.

Desktop Cloud aims at providing services based on Cloud business model on top on

infrastructure that is not made for this purpose. The success of Desktop Grids projects

has simulated the idea of applying the same concept within Cloud computing. How-

ever, the paper has presented several research issues that need further attention.

References

1. Kirby, G., Dearle, A., Macdonald, A., Fernandes, A.: An Approach to Ad hoc Cloud

Computing. Arxiv Prepr. arXiv1002.4738. (2010).

2. Chandra, A., Weissman, J.: Nebulas: Using distributed voluntary resources to build clouds.

Proceedings of the 2009 conference on Hot topics in cloud computing. pp. 2–2. USENIX

Association (2009).

3. Weissman, J.B., Sundarrajan, P., Gupta, A., Ryden, M., Nair, R., Chandra, A.: Early

experience with the distributed nebula cloud. Proceedings of the fourth international

workshop on Data-intensive distributed computing. pp. 17–26. ACM (2011).

4. Cunsolo, V.D., Distefano, S., Puliafito, A., Scarpa, M.: Volunteer computing and desktop

cloud: The cloud@ home paradigm. Network Computing and Applications, 2009. NCA

2009. Eighth IEEE International Symposium on. pp. 134–139. IEEE (2009).

5. Cunsolo, V., Distefano, S.: From volunteer to cloud computing: cloud@ home. Conf.

Comput. Front. 103–104 (2010).

6. Harutyunyan, a, Blomer, J., Buncic, P., Charalampidis, I., Grey, F., Karneyeu, a, Larsen,

D., Lombraña González, D., Lisec, J., Segal, B., Skands, P.: CernVM Co-Pilot: an

Extensible Framework for Building Scalable Computing Infrastructures on the Cloud. J.

Phys. Conf. Ser. 396, 032054 (2012).

7. Andrzejak, A., Kondo, D., Anderson, D.P.: Exploiting non-dedicated resources for cloud

computing. 2010 IEEE Netw. Oper. Manag. Symp. - NOMS 2010. 341–348 (2010).

8. Litzkow, M.J., Livny, M., Mutka, M.W.: Condor-a hunter of idle workstations. [1988]

Proceedings. 8th Int. Conf. Distrib. 104–111 (1988).

9. Cunsolo, V., Distefano, S., Puliafito, A., Scarp, M.: Cloud@ home: Bridging the gap

between volunteer and cloud computing. ICIC’09 Proc. 5th Int. Conf. Emerg. Intell.

Comput. Technol. Appl. 2009. (2009).

10. Marosi, A., Kovács, J., Kacsuk, P.: Towards a volunteer cloud system. Futur. Gener.

Comput. Syst. (2012).

11. Gupta, A., Awasthi, L.K.L.: Peer enterprises: A viable alternative to Cloud computing?

Internet Multimedia Services Architecture and Applications (IMSAA), 2009 IEEE

International Conference on. pp. 1–6. IEEE (2009).

12. Arpaci, R.H., Dusseau, A.C., Vahdat, A.M., Liu, L.T., Anderson, T.E., Patterson, D.A.:

The Interaction of Parallel and Sequential Workloads on a Network of Workstations. ACM

(1995).

13. Foster, I., Zhao, Y., Raicu, I., Lu, S.: Cloud computing and grid computing 360-degree

compared. Grid Computing Environments Workshop, 2008. GCE’08. pp. 1–10. IEEE

(2008).

14. Dillon, T., Wu, C., Chang, E.: Cloud computing: Issues and challenges. 2010 24th IEEE

International Conference on Advanced Information Networking and Applications. pp. 27–

33. IEEE (2010).

15. Ristenpart, T., Tromer, E., Savage, S., Shacham, H.: Hey, you, get off of my cloud:

exploring information leakage in third-party compute clouds. Proceedings of the 16th ACM

conference on Computer and communications security. pp. 199–212. ACM (2009).

16. Cao, B.Q., Li, B., Xia, Q.M.: A Service-Oriented Qos-Assured and Multi-Agent Cloud

Computing Architecture. Cloud Comput. 644–649 (2009).

17. Endo, P., Palhares, A. de A., Pereira, N.N., Goncalves, G.E., Sadok, D., Kelner, J.,

Melander, B., Mangs, J.-E.: Resource allocation for distributed cloud: concepts and

research challenges. Network, IEEE. 25, 42–46 (2011).

18. Kondo, D., Taufer, M., Brooks, C.: Characterizing and evaluating desktop grids: An

empirical study. Int. Parallel Distrib. Process. Symp. 2004. 00, (2004).

19. Andrzejak, A., Kondo, D., Anderson, D.P.: Ensuring collective availability in volatile

resource pools via forecasting. Manag. Large-Scale Serv. 149–161 (2008).