THE IETF SHOULD CREATE AN INTERCLOUD RFC
The Interfaces in the IETF Cloud Reference Framework should be Standardized
Jan Sipke van der Veen
TNO, Groningen, The Netherlands
Robert J. Meijer
University of Amsterdam, Amsterdam, The Netherlands
Keywords:
Cloud computing, Standardization.
Abstract:
This paper discusses the current state of cloud computing and shows that it is comparable to the state of
networks before the internet. Clouds need to be connected more to make it easier for users to switch between
providers and at the same time make it easier for providers to supply ”infinite resources”. At the moment the
big cloud providers do not feel the need to standardize this intercloud and there is no authorative body that
sets the standard. This paper argues that it is necessary for the IETF to take its current effort to create a cloud
reference framework one step further and standardize the interfaces between the functions and layers as well.
1 INTRODUCTION
The concept of computing as a utility has been around
since the 1960’s (Parkhill, 1966), but it took until the
mid 2000’s for cloud computing to turn that concept
into a reality. The past years have shown a rapid up-
take in its usage and a lot of providers have joined the
market (Armbrust et al., 2010). Although their ser-
vices are similar in concept, the differences between
them make it difficult for users to switch from one
provider to another. We argue that standardization of
a reference framework alone is not enough to realize
the intercloud. The interfaces between cloud func-
tions and layers should be standardized as well.
This paper will discuss the currently disconnected
clouds in section 2, describe the potential of the inter-
cloud in section 3 and make the case for the need of
standardization of the intercloud by the IETF in sec-
tion 4. Finally it presents two cases where standard-
ization of the intercloud would be very beneficial in
section 5.
2 DISCONNECTED CLOUDS
Cloud computing is still a relatively young phe-
nomenon. Amazon played a key role in the develop-
ment of cloud computing when it launched Amazon
Web Services (AWS) and specifically Elastic Com-
pute Cloud (EC2). Several providers followed in the
next years to offer network based IT services in a pay
per use fashion. These services can be grouped into
one of three service models (Mell and Grance, 2009):
Infrastructure as a Service (IaaS), Platform as a Ser-
vice (PaaS) and Software as a Service (SaaS).
Because of technical differences between these of-
ferings it is hard to switch from one provider to an-
other. This is true even for similar services within
the same service model. For example, transferring an
e-mail archive from one provider to the next is impos-
sible if not all providers allow a protocol like IMAP
to extract all current messages from the system.
Cloud computing is often attributed ”infinite re-
sources”, at least in the eyes of the consumer. How-
ever, this apparent infinity of resources comes at a
price to the provider. He must ensure that enough
resources are available when demand increases unex-
pectedly. A relatively high percentage of resources is
therefore unused most of the time in small clouds. In
bigger clouds this percentage can be lower, but is still
limited by the size of a single cloud.
The state of cloud computing looks a lot like
the state of networks before the internet. Networks
used to be disconnected and connectivity between
networks could only be achieved with networks of the
same type. Internet changed all this by defining layers
192
Sipke van der Veen J. and J. Meijer R..
THE IETF SHOULD CREATE AN INTERCLOUD RFC - The Interfaces in the IETF Cloud Reference Framework should be Standardized.
DOI: 10.5220/0003437101920195
In Proceedings of the 1st International Conference on Cloud Computing and Services Science (CLOSER-2011), pages 192-195
ISBN: 978-989-8425-52-2
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
on top of the disconnected networks, see left side of
figure 1. Even though the separate networks still use
different technologies, e.g. ethernet and roken ring,
the layers on top ensure that communication between
them is possible.
Network 1
Internet Layer
Network n
Transport Layer
Application Layer
Cloud 1
Intercloud Layer
Cloud n
Resource Layer
Application Layer
Figure 1: Current internet RFC (left) and possible inter-
cloud RFC (right).
3 CONNECTED CLOUDS
The tranformation of all these separate clouds into a
connected intercloud would bring great benefits to its
users as well as its providers.
The most important benefit for users would be the
ease of switching between providers. Taking the ex-
ample of the internet again, it is currently easy to
switch from one network vendor to the next. It does
not even matter if the network is provided via a fixed
line, a mobile connection or via satellite, each and ev-
ery machine using IP can communicate with all other
machines using IP.
The intercloud would benefit providers as well.
It allows them to share their resources in such a
way that the percentage of unused resources, wait-
ing for unexpected demand, can be much lower per
provider. Currently, a provider attracting mostly UK
customers probably needs more resources during UK
office hours. To meet this demand, a lot of resources
are unused during the night. In the intercloud sce-
nario, a provider with mainly US or Australian cus-
tomers could use these resources during their office
hours.
At the moment the big cloud computing providers
do not feel the need for standardization. They mainly
see the drawbacks of a standard which allows cus-
tomers an easier move to their competitors. The
smaller providers and especially the users would like
to see such a standard, but the current efforts in this
direction are not good enough. The (proposed) stan-
dards are either limited in scope or the authors are not
seen as authorative. In our opinion the IETF would
make the ideal candidate to describe such an impor-
tant cloud standard. The right side of figure 1 shows
how such an intercloud RFC may look quite similar
to the existing internet RFC.
4 STANDARDIZATION
Standardization of cloud computing would need to
address several aspects of cloud technology, but also
some business cooperation between its providers.
4.1 Interoperability
One of the first areas of standardization should be
the interoperability between different clouds. We will
distinguish between the three service models again,
because they differ very much in the level and type of
standardization that can take place.
4.1.1 Software as a Service
Standardization in this service model should address
the input and output format(s) of the user data, e.g.
how the messages in an e-mail service should look
like. It should also define the protocol(s) for trans-
mission of this data, e.g. which steps are necessary to
transfer the messages out of the e-mail service. In the
general case this is hard to achieve, because there are
so many different (possible) services. But for specific
types of services it is quite possible to define certain
requirements on these formats and protocols. Several
standards are therefore needed in this service model,
one for each service type.
4.1.2 Platform as a Service
Standardization in this service model should address
the programming APIs. Ideally, this standard would
be language independent or at least easy to use in
different languages. Of course there will always be
a need for providers to distinguish themselves from
their competitors. This can be achieved by providing
a common set of calls that are mandatory and a set
of optional or extendable calls that should be clearly
marked as such. This leaves the user free to choose
between interoperability and extra functionality that
is specific to a certain provider. In this case providers
can still make their own decisions about the program-
maging languages they offer. But if they do offer a
specific one, the user can be sure that his applica-
tion, using only mandatory calls or supported optional
calls, will work with this provider.
4.1.3 Infrastructure as a Service
Standardization in this service model should address
the virtual machine representation. One part of this is
the representation of the state of the virtual machine,
e.g. the harddisk and the memory contents of a paused
virtual machine. Another part is the representation of
THE IETF SHOULD CREATE AN INTERCLOUD RFC - The Interfaces in the IETF Cloud Reference Framework should
be Standardized
193
metadata about the virtual machine, e.g. the amount
of memory and the number and type of processors it
contains. Standardization should also address the in-
teraction with the virtual machines from a client com-
puter. This interaction, also with the cloud as a whole,
is especially important for management tools. To be
able to make informed decisions on the number of ac-
quired virtual machines and their providers, it is also
necessary to standardize the cost reporting of the dif-
ferent virtual machines at offer.
4.2 Federation
A second area of standardization should be the feder-
ation between different clouds.
4.2.1 Identity and Security
To create the illusion of a single cloud to its users,
there is the need for a shared sense of ”who is who”.
This can be achieved by standardizing on a federated
identity management method. Currently authentica-
tion is mostly achieved with preshared passwords, but
can also be more secure with public and private keys.
It is vital for the intercloud to have a common set of
authentication methods, perhaps with restrictions on
the capabilities if a less secure form of authentication
is used.
Creating a secure connection between servers in
the cloud and servers located elsewhere is a big issue.
This issue needs to be resolved for the intercloud to
work properly, because it depends on a much more
fluid landscape of servers that need to communicate
in a secure fashion.
4.2.2 Accounting and Billing
In the intercloud scenario it would make sense for
a user to subscribe to a single cloud and be able to
use resources of other clouds without being aware of
the technical and business differences between these
clouds. Much in line with a need for a federated iden-
tity management method therefore is the need for fed-
erated accounting and billing methods. This would
allow the cloud providers a way to keep track of the
billable activities of the user, even if these activities
take place in other clouds.
4.3 IETF
The IETF is currently working on or has worked on
several draft standards involving cloud computing:
A6: The Automated Audit, Assertion, Assess-
ment and Assurance API (Hoff et al., 2010)
LISP: Locator / Identifier Separation Protocol
(Farinacci et al., 2009)
These two draft standards focus on specific parts
of cloud computing. The first draft standard provides
an open, extensible and secure interface that allows
cloud computing providers to expose audit, assertion,
assessment and assurance information for IaaS, PaaS
and SaaS services. A client would then be able to in-
terrogate the service and verify compliance with local
policy before making use of it, e.g. by checking the
geographical location of the servers. The second draft
standard describes a simple, incremental, network-
based protocol to implement separation of internet ad-
dresses into endpoint identifiers and routing locators.
This would alleviate some of the scaling issues cur-
rently visible in network routing.
The IETF is also currently working on reference
frameworks that can be used to guide the standardiza-
tion of cloud computing:
CSF: Cloud Security Framework (Karavettil et al.,
2010)
CRF: Cloud Reference Framework (Khasnabish
et al., 2010)
The first one deals with the security aspects
of cloud computing and establishes security stan-
dards, policies, procedures and guidelines for cloud
providers. This framework would enable cloud
providers and cloud users to practice safe security
techniques for their applications and intracloud and
intercloud information exchange. The second frame-
work addresses cloud computing in a more general
way. This Cloud Reference Framework involves ba-
sic functions or layers to support the general require-
ments of cloud applications and services. It divides
itself into four horizontal layers and one stacked ver-
tical layer to support configuration management, reg-
istry, logging and auditing, security management and
service level agreement (SLA) management, see fig-
ure 2.
Application / Service Layer
Resource Control Layer
Resource Abstraction & Virtualization Layer
Execution
Node
Execution
Node
Execution
Node
Physical Resource Layer
Cloud
Management
Manager
Node
Figure 2: Cloud Reference Framework (draft).
The Cloud Reference Framework seems very
promising as it provides a common understanding of
the different cloud computing functions and layers. In
our opinion it should form the basis for another stan-
CLOSER 2011 - International Conference on Cloud Computing and Services Science
194
dard that describes the interfaces between the func-
tions and layers described in the framework.
5 CASES
This section discusses two cases where standardiza-
tion of the intercloud would be very beneficial.
5.1 Early Warning Systems
Early Warning Systems (EWS) provide a technolog-
ical alternative to human survaillance on critical in-
frastructures like dikes and bridges. These systems
take sensor data from the monitored infrastructure,
analyse the data and report to other information sys-
tems or people about the state of the infrastructure,
see figure 3.
During normal operation of an EWS, i.e. there is
nothing wrong with the infrastructure, not much com-
puting capacity is needed to run the analysis. How-
ever, if things are about to go wrong, e.g. the EWS
thinks a dike section will break soon, a lot more com-
puting capacity is needed. This capacity is needed
for instance to simulate the water flowing through the
dike section and into the urban area behind it.
An EWS is therefore a very good candidate for
cloud computing. It would be very inefficient to have
this much computing capacity standing by just in case
something goes wrong with the monitored infrastruc-
ture. However, there is a risk in using a single cloud
provider for such an important IT system as an EWS,
which may warrant the extra cost involved in set-
ting up dedicated computing capacity. The intercloud
would solve this dilemma, because it allows for very
fast switching of providers if something goes wrong
with one of them.
ŶŝƚŽƌ ŶĂůLJƐĞ
sŝƐƵĂůŝƐĞ
ůĞƌƚ
^ŝŵƵůĂƚĞůĐĂŶŽ
ŝŬĞ
ƌŝĚŐĞ
Figure 3: Early Warning System (EWS).
5.2 Dynamic CDNs
Content Delivery Networks (CDN) are widely used to
distribute content to users. In a CDN, content is dis-
tributed to several nodes owned by the CDN operator.
If a user requests this content, he is redirected to a
node that is able to provide the content in a fast and
efficient way, e.g. the one that is close to him and not
at maximum capacity yet.
A CDN could be distributed even more if the
nodes do not have to be owned by the CDN opera-
tor itself. If the CDN node is actually a virtual ma-
chine that can be easily transferred from one cloud
to another, the node could be placed even closer to
the potential users. For this scenario to work well, it
is necessary to have many potential locations for the
nodes. The intercloud would be able to provide this.
6 CONCLUSIONS
This paper discussed the current state of cloud com-
puting and showed that it is comparable to the state
of networks before the internet. Clouds need to be
connected more to make it easier for users to switch
between providers and at the same time make it eas-
ier for providers to supply ”infinite resources”. Two
cases were presented that would benefit from the in-
tercloud: early warning systems and dynamic content
delivery networks.
At the moment the big cloud providers do not feel
the need to standardize this intercloud and there is no
authorative body that sets the standard. It is necessary
for the IETF to take its current effort to create a cloud
reference framework one step further and standard-
ize the interfaces between the functions and layers as
well.
REFERENCES
Armbrust, M., Fox, A., Griffith, R., Joseph, A. D., Katz,
R., Konwinski, A., Lee, G., Patterson, D., Rabkin, A.,
Stoica, I., and Zaharia, M. (2010). A view of cloud
computing. Communications of the ACM.
Farinacci, D., Fuller, V., Meyer, D., and Lewis, D.
(2009). Locator/id separation protocol (lisp).
http://tools.ietf.org/html/draft-farinacci-lisp-12.
Hoff, C., Johnston, S., Reese, G., and Sapiro, B. (2010).
Automated audit, assertion, assessment, and as-
surance api (a6). http://tools.ietf.org/id/draft-hoff-
cloudaudit-00.html.
Karavettil, S., Khasnabish, B., So, N., Golovinsky,
G., and Yu, M. (2010). Cloud security framework.
http://trac.tools.ietf.org/area/app/trac/attachment/wiki/
Clouds/Karavettil-et-al-CSF-Proposal-10Dec2010.
pdf.
Khasnabish, B., Chu, J., Ma, S., Meng, Y., So, N.,
and Unbehagen, P. (2010). Cloud reference frame-
work. http://www.ietf.org/id/draft-khasnabish-cloud-
reference-framework-00.txt.
Mell, P. and Grance, T. (2009). The nist definition of cloud
computing. http://csrc.nist.gov/groups/SNS/cloud-
computing/cloud-def-v15.doc.
Parkhill, D. (1966). The Challenge of the Computer Utility.
Addison-Wesley Publishing Company.
THE IETF SHOULD CREATE AN INTERCLOUD RFC - The Interfaces in the IETF Cloud Reference Framework should
be Standardized
195