Easing the Deployment and Management of Cloud Federated Networks
Across Virtualised Clusters
Ivan Andrade Casta
˜
neda
1
, Ignacio Blanquer
1 a
and Carlos de Alfonso
2 b
1
Institute for the Instrumentation for Molecular Imaging - I3M, Universitat Polit
`
ecnica de Val
`
encia, Valencia, Spain
2
SENESCYT - Secretar
´
ıa de Educaci
´
on Superior, Ciencia, Tecnolog
´
ıa e Innovaci
´
on, Ecuador
Keywords:
Cloud Federation, Overlay Networks, Virtual Machine Migration, Federated Virtual Private Network.
Abstract:
Cloud federation, in the last years, has grown rapidly in literature because to its multiple advantages to co-
ordinate and re-use different services across multiple sites, different geographic locations or cloud providers.
This article presents a federated network architecture and focuses on the multi-tenant overlay networks cre-
ation across different sites, as well as, the inter-site migration of virtual machines, introducing a framework
that allows us to manage our federate cloud environment in three scenarios: distribution of Virtual Machines,
resource management, and network management.
1 INTRODUCTION
Nowadays, big data centers networks provide redun-
dancy and ensure reliability to their customers in case
of a site failure, also has to minimize cost while deal-
ing with high peaks of demands. Thus, cloud feder-
ation has grown rapidly in the literature and became
a powerful paradigm for cloud providers and for sci-
entific computing, as its facilities the interoperability
of different cloud computing environments, offering
access to different resources that are allocated in dif-
ferent geographic locations or cloud providers.
Despite containers are sometimes treated as vir-
tual machines, they are conceptually different objects.
Containers are groups of processes that run on envi-
ronments isolated at the level of a filesystem, devices,
process namespaces or resource allocations. Contain-
ers are mainly implemented through operative system
calls and therefore do not boot a separate operating
system. They offer other advantages as packaging ap-
plications dependencies, image preparation and dis-
tribution, and memory footprint. However, container
isolation is reduced with respect to Virtual Machine
environment, especially in a multi-tenancy environ-
ment.
The network and resource management require-
ments are demanding, with the emerging need for a
middleware layer to provide a general orchestration
a
https://orcid.org/0000-0003-1692-8922
b
https://orcid.org/0000-0002-2378-021X
of resources, giving the right connectivity between the
resources across the federation, as well as a complete
view of the entire network from any point of the cloud
federated environment.
Such a middleware layer will imply requirements
as Network re-programming (such as SDN, which
makes possible to separate the Control Plane from
the Data Plane (Kaur et al., 2016)), a centralized ser-
vice that provides a reliable storage data, discovering
and configuring services (in our case Consul (Con-
sul, 2019), EC managed (EC-Managed, 2019) or etcd
(Coreos, 2019)), coherent IP address assignment sup-
porting WAN live VM migration, user authentication,
and authorization, etc.
The framework that we present in this paper, im-
plements a solution that performs resource federation
at the network layer, based in Software Defined Net-
working (SDN) technology, creating virtual networks
on demand across multiple data centers including the
VM live migration over WAN in a federated environ-
ment.
In this article, we are going to test and analyze
the benefits of the framework in 3 different scenarios:
distribution of VMs, resource management, and net-
work management. The rest of the paper is structured
as follows. First, Section II describes an overview of
related works in cloud federation, network federation
and live migration. In Section III, we define the fed-
erated network topology used, based on three differ-
ent data centers. Section IV describes more in detail
the lower-level processes for each scenario that our
Castañeda, I., Blanquer, I. and de Alfonso, C.
Easing the Deployment and Management of Cloud Federated Networks Across Virtualised Clusters.
DOI: 10.5220/0007877406010608
In Proceedings of the 9th International Conference on Cloud Computing and Services Science (CLOSER 2019), pages 601-608
ISBN: 978-989-758-365-0
Copyright
c
2019 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
601
model offers for both, network and resources manage-
ment. The preliminary results are presented in sec-
tion V. The article concludes in Section VI, where we
discuss the proposed approach, summarizes the work
and suggests future works
2 RELATED WORKS
2.1 Cloud Federation
Cloud federation is to interconnect two or more cloud
computing environments of the same or different
service provider, to interact between different types
of typologies that can share resources to optimize
the quality of services and reduce provisioning costs
(Moreno-Vozmediano et al., 2012). In the literature,
cloud federation can be defined as centralized or de-
centralized (Melhem et al., 2017), where the main dif-
ferent is that in a decentralized federation the clouds
communicate and negotiate directly with each other.
Meanwhile in a centralized federation, such as Inter-
cloud (Elmroth et al., 2009), Contrail (Carlini et al.,
2012), Dynamic Cloud Collaboration (Hassan et al.,
2010), and Federated Cloud Management (Marosi
et al., 2011); central entities act as intermediates for
performing tasks such as resource registration, mar-
keting, searching or allocation.
This work is focused on a decentralized federation
environment, such as, The European Grid Infrastruc-
ture (EGI) Federated Cloud (Newhouse and Brewer,
2012) that is a federation of publicly funded comput-
ing, storage, and data resource centers across Europe
to provide access to high-throughput computing re-
sources through the EGI Core Infrastructure Platform.
The idea behind it is to offer a cloud infrastruc-
ture consisting of a pool of resources and services
provided by both public and private partners and pre-
sented as a single system. The problem is that the fed-
erated resources demanded by the tenants are imple-
mented at the level of the cloud management middle-
ware, making it more complicated to federate clouds
that run non-supported middleware or even particular
versions of supported middleware that are not com-
patible with the EGI model.
Another model of cloud federation is Reservoir
(Rochwerger et al., 2009), a European project in-
terconnecting computing grids, creating a peer-to-
peer(P2P) cloud federation to lease its resources to
other members of the federation.
OpenStack (OpenStack, 2019), one of the most
used and popular platforms to build private clouds,
where a cloud installation can be configured as a ser-
vice provider and as an identity provider.
Used as a service provider, a local cloud will
trust users that present proof of authentication
from a trusted identity provider; such as SAML
or OpenID to give authentication and authoriza-
tion to local clouds.
When is used as an entity provider, the local cloud
can be used to emit authentication tokens that can
be taken to a trusted remote service provider.
Fogbow (Brasileiro et al., 2016) is a middleware
for cloud federation that has been developed in the
context of the EUBrazilCC project, co-funded by
the European Commission and the Brazilian Min-
istry of Science, Technology, Innovation, and Com-
munication (MCTIC). It has the same concept as
EGI Federated Cloud, but implements federation at a
higher level, being more flexible to interact with cloud
providers that are not compatible with the EGI model.
In summary, the actual model of cloud federation
are mature in Authentication and Authorization In-
frastructures (AAI), virtual machine repositories fed-
eration, IaaS interfaces or use of distributed entities
for directories or brokering, as we presented in our
previous work (Andrade et al., 2017) . But, in Feder-
ation of cloud networks is still an issue in cloud com-
puting.
2.2 Networking Federation
Software-Defined Networking (SDN) makes possible
to separate the Control Plane from the Data Plane giv-
ing us more control to re-programming the network
on demand. There are various solutions that provide a
tool for cloud network management, such as, Onos,
Nox/Pox, OpenDaylight, Beacon, RYU, Floodlight
and many more (Khondoker et al., 2014) allowing the
traffic orchestration through the OpenFlow protocol
(McKeown et al., 2008).
Network Functions Virtualisation (NFV) enable
the virtualization of dedicated network components,
like routers, firewalls, load balancers, and other
components that need dedicated hardware equipment
and give them to users as services (Rodriguez and
Guillemin, 2016). Also NFV is considerate as a key
concept in container-based computing to create over-
lay network, such as, Flannel (Flannel, 2019), Weave
(Weave, 2019), Calico (Calico, 2019), Romana (Ro-
mana, 2019); used in well-known platforms for vir-
tualization and resource management,such as Mesos
(Mesos, 2019) and Kubernetes (Kubernetes, 2019).
These two technologies allow cloud providers to
provide the necessary federated networking capabil-
ities, having better control and performance over the
networks, also allow the automation of resource man-
IWFCC 2019 - Special Session on Federation in Cloud and Container Infrastructures
602
agement to improve scalability, quality of services
and reduce costs.
2.3 Live Migration
The main advantage of live migration is that give us
flexibility, allowing us to move one virtual machine
from one physical host to another without disrupt-
ing its normal operation, where memory, storage, and
network connectivity are transferred. In a WAN net-
work, the live migration faces more challenges than
in a LAN-based live VM migration, such as the la-
tency, limited bandwidth, having a shared storage sys-
tem between the sites that can be a bottleneck, correct
IP addresses allocation after the migration.
Solutions such as Silvera et al. (Silvera et al.,
2009), they propose to have central agents on both
sites to avoid changing the virtual machine IP address
in each migration. These agents are responsible to en-
sure the connectivity using Proxy-ARP and IP-in-IP
tunnels to forward the traffic between them.
LayerMover (Zhang et al., 2018) use a duplication
technique in three-layer image structure, in order to
optimize the performance.
Bradford et al. (Bradford et al., 2007) propose a
solution based on DNS-resolutions through IP tunnel-
ing. Improving the network performance because the
VM machine maintains its canonical names, and the
new IP address is registered with the named host.
A unified virtual network to provide a complete
view of the LAN keeping a single IP address, for
instance, is presented by wood et al. (Wood et al.,
2015). Where the idea is to combine virtual Private
networks (VPNs), layer 3, and virtual private LANs
service (VPLS), layer 2, to provide end-to-end rout-
ing across multiple networks and bridge LANs at dif-
ferent locations.
TCP connection before the migration to reduce the
total size of the dataset is proposed by Kuribayashi et
al. (Kuribayashi, 2013). Eliminating the redundancy
on block level and more coarse-grained than compres-
sion.
3 ARCHITECTURE
The architecture that has been deployed to recreate
a federated network environment is shown in Figure
1. Where we have some resources deployed across
three different sites reaching each other through a pri-
vate overlay network, having two different simulated
tenants with full isolation between them and with the
underlying network with the same IP segment address
(172.16.0.0/16).
Figure 1: Multi-site Federated Network.
This model of cloud federation architecture enables
the access to resources that are not allocated in the
same site, such as the Router as a Service - RaaS (to
reach to the outside world), the dynamic IP allocation
(via DHCPaaS) or the IPfloater tool to serve floating
IPs.
It includes a centralized service-based for dy-
namic infrastructure, CONSUL (Consul, 2019). It is a
distributed service mesh to connect, secure, and con-
figure services across any runtime platform and public
or private cloud.
Features such as service and node discovery
mechanisms to incorporate new Cloud servers, ensure
that services are always working and scalable when
facing peak demand, health checks with the purpose
of providing detailed monitoring and control of ser-
vices and nodes. Therefore, we can build a sophis-
ticated level of awareness into your applications and
services.
The network and resource management are also
managed by OpenVswitch among with libvirtd,
where the first one is a multilayer software switch,
that provides the connectivity among the nodes; as a
virtual boundary switch. Enabling massive network
automation through programmatic extension and pro-
viding logical end-points to the resources (Open-
Vswitch, 2019). On the other hand, libvirt is used
as a wrapper to handle VM and LXC deployed across
the federation (Libvirt-project, 2019).
Floodlight (Floodlight-Project, 2019) is an SDN
controller designed to work with the growing num-
ber of switches, routers, virtual switches, and access
points that support the OpenFlow standard; which
is an open standard to orchestrate traffic flows in an
SDN environment.
To manage the WAN-based live VM migration
and reduce problems such as network performance,
data migration performance, the continued connectiv-
ity, and IP address reassignment; we decide to use
NFS (NFS, 2019) as file sharing protocol in combi-
nation with OFS (OFS, 2019), also Known as Of-
Easing the Deployment and Management of Cloud Federated Networks Across Virtualised Clusters
603
fline FileSytem. It is is a tool that makes it possi-
ble to extend every filesystem with offline capabili-
ties. In other words, we are able to work in offline
mode avoiding the RPCs calls made by NFS that can
be deadly to performance in WAN network.
In summary, the framework uses these technolo-
gies to grant elasticity, reliability, isolation, scalabil-
ity, monitoring, deployment, releasing and control of
services and nodes in our federated cloud network.
4 USES CASES
According to the cloud architecture presented in sec-
tion 3, we have defined three main uses cases: distri-
butions of VMs, resource management, and network
management. Inside of these three scenarios, we de-
rive several uses cases that describe lower-level pro-
cesses for each scenario which are described in this
section briefly.
4.1 Network Management
Combining Libvirt, Consul, OpenVswitch and the
Floodlight controller, we can manage our federate
network built across the sites. It allows us to create
overlay networks, assign the right VLANs for each
tenant, manage the complete process of creating and
deleting end-points for each tenant (Linux Bridges),
define firewall rules and modifies network character-
istics such as routing, NAT capacity, firewall rules,
and IP assignment.
The set of network management calls that we have
defined is the next:
Fed net create: It creates a new federated net-
work from scratch from either of the sites, it de-
fines the network configuration (name, UUIDs,
VLANs, end-points for the resources, bound-
ary switches creation, overlay network connec-
tion, IPs assignment, assigns the corresponding
SDN controller and applies the flow rules from
fed
net sdn controller use case), avoiding the du-
plication between tenants. The network defini-
tion, in most of the process, is carried out by
configuration files and replicated in all the sites
thanks to the use of key-value storage based on
Consul.
Fed net delete: Performs the opposite action of
the previous use case. It deletes the federated net-
work from all the sites. It also ensures that no
more resources are still deployed in the tenant net-
work before and proceed to remove both the entire
network configuration and the definition files in
our centralized service.
Fed net list: Give us a global view of all the re-
sources in the federation using the Libvirt API to
retrieve the networks deployed at each site of the
federated cloud.
Fed net sdn controller: This function call up-
dates the flow entries, previously defined in a
configuration file. It enables to program and re-
program the network on demand.
4.2 Resource Manegement
Libvirt as wrappers to handle Kernel-based Vir-
tual Machine (KVM, 2019), Linux containers (LXC,
2019) and Consul as distributed service, give us com-
plete resource administration across the federated
cloud, from the resources definition to the resources
removal in an automatic way.
In conjunction with the live migration use case, we
can define the deployment, edit, delete and perform
WAN live migration across the different sites in the
federation. Guaranteeing high availability, elasticity,
scalability of the resources and avoiding duplication
because it gives us a global view of the resources that
are deployed across the federation.
Fed vm create: We can deploy resources on de-
mand and verify if the resource has not been cre-
ated before avoiding duplication. We used pre-
defined templates that define the features of each
resource, the packages required according to VM
role (Rass, DHCPaaS, DNSaaS, LBaaS), user,
password, hostname, IP tables, ssh keys, IPs seg-
ments and other information which is updated and
contextualized on deployment time.
Fed
vm delete: This function is the opposite of
the previous use case and considers the same cri-
teria as before.
Fed vm list: From a single point of our feder-
ated cloud, we can have a complete view of the re-
sources deployed in the federation, as an extended
LAN network.
4.3 WAN-based Virtual Machine live
migration
Having talked about the issues as network and data
migration performance, the continued connectivity
and IP address reassignment in section 2. We decided
to use a file sharing protocol NFS and OFS (Offline
FileSytem tool) to avoid the RRPCs calls made by
NFS.
Fed vm migration: It performs the live migra-
tion, from any point of the federation, of a VM
IWFCC 2019 - Special Session on Federation in Cloud and Container Infrastructures
604
across the cloud sites using both tools mentioned
before, NFS and OFS.
5 TEST AND RESULTS
As we describe in section 3, the federation has three
different sites simulating a WAN network environ-
ment, where our WAN will be the Campus network.
Accessing the different resources deployed across the
federation via a private overlay network.
For the purpose of the test we carried it on two
of the three hosts with different characteristics, as we
can see in table 1, gauging the time performance in
seconds and executing it five times for each lower-
level use cases using four different flavors, as shown
in table 2, mainly for the resource management use
case.
Table 1: Host features.
Processor cores Frequency - (GHz) RAM
Host A intel i5-3470 4 3.20 7.7
Host B intel core 2 Quad - Q9300 4 2.50 7.8
Table 2: Virtual Machine Flavors Description.
Flavor VCPUs RAM(MB) Disk(GB)
Tiny (t) 1 512 2
Small (s) 1 2048 20
Medium (m) 2 4096 40
Large (l) 4 8192 80
5.1 Network Management
Table 3: Average Performance Time - Network Manage-
ment.
Host A Host B Percentage difference
Network Deployment 4.001 4.186 5 %
Network Removal 1.711 1.953 14 %
Network List 0.658 0.689 5 %
Network Reprogramming 0,366 0.348 5 %
Table 3, shows four lower-level processes for
the complete network administration: network
deployment (fed net create), network removal
(fed net delete), network list(fed net list) and
network reprogramming (fed net sdn controller).
Revealing that different features of each host did not
affect the performance of the framework, because the
percentage difference for each use case is between
5% to 14%, which means, a minimum increase of
only milliseconds between the two host.
Table 4: Average Performance Time - Resources Manage-
ment.
Flavor Host A Host B Percentage difference
t 7.280 13.130 80 %
Resource Deployment s 7.828 20.390 160 %
m 7.986 20.927 162 %
l 8.249 21.118 156 %
t 0.572 1.920 236 %
Resource Releasing s 0.577 2.342 306 %
m 0.561 2.483 343 %
l 0.592 2.663 350 %
t 108.186 365.306 216 %
Full boot s 119.122 347.542 218 %
m 111.116 351.260 216 %
l 142.536 350.180 164 %
Resource List - 0.658 0.681 1 %
5.2 Resource Management
This use case has three lower-level processes
for complete management: resource deployment
(fed vm create), resource removal (fed vm removal)
and resource list (fed vm list).
The resource deployed for the test was a DHCP
server, using Ubuntu Cloud Images (Ubuntu-Cloud-
images, 2019). Therefore, it needs to update its repos-
itory at the booting time and install some packages for
its correct performance, giving us the possibility to
gauge the deployment time made by the framework
and the time it takes to a full boot, as shown in table
4.
Also, we have to consider the fact that we are de-
ploying a resource that has to reach the gateway in
Host B to get access to the internet. Therefore, we
have to reduce the MTU slightly in the gateway inter-
face because the traffic has to to go through the tunnel
Vxlan and will add an extra header to every packet;
as we presented in our previous work (Andrade et al.,
2017).
Resource Deployment: Shows an increase in the
performance from 80% to 156%, this means be-
tween 6 to 13 seconds more in host B than host
A.
Resource Removal: The performance increase
was even greater than in the first use case, with
an increase from 256% to 350%. However, if we
look at table 4, the increase was just 0.5 to 2.6
seconds respectively.
Resource List: This use case presented an in-
crease of only 1%, which means nothing relevant,
just milliseconds.
Full boot: This use case shows a big increase in
the performance around 230 seconds more than
host A.
Easing the Deployment and Management of Cloud Federated Networks Across Virtualised Clusters
605
The test reveals that an increase in three of the four
use case, where we can determine that one of the most
relevant reasons for this increase is the different fea-
ture of each host. Because if we compare each flavor
in each host, there is not much increase, just around
0.5 - 1 second, even though each flavor has different
characteristics.
Where we can see a consistent increase is the full
boot test, due to the slight reduction of the MTU that
we had to do at the interface of our gateway. We can
conclude that the performance time will concern the
WAN network congestion (UPV network) than to the
characteristics of each host.
5.3 VMs Distribution
Migration time depends on the resource activity, so
if the resource is highly active, the memory transfer
process will take longer to complete. For the present
work, we have focused only on VM migration, leav-
ing container migration for further work.
We followed the same procedure as (ZHAW,
2014) to analyze the performance of our framework in
live migration on WAN. Using a stress tool we could
generate a stable memory consumption of around
75%, during migration.
Giving us the opportunity to measure the migra-
tion process time, Virtual machine downtime duration
and the amount of data transferred across the network
during the live migration process between a loaded
against unloaded VM.
Table 5: WAN live migration Performance.
Flavor Unloades VM Loades VM Percentage difference
t 40.298 52.637 31 %
Migration Time s 43.509 183 321 %
m 52.228 196.5 276 %
l 60.12 207 244 %
t 1 31 3000 %
VM Down time s 1 159.5 15850 %
m 9.6 190.5 1884 %
l 12 588.5 4804 %
t 408 514.5 26 %
Total amount of data transferred s 440 1547.5 252 %
m 544 2028.5 273 %
l 623 2040 227 %
On table 5, we can see the three tests to gauge the
live migration performance:
Migration time: Total migration time, in seconds,
of a resource from one physical host to another
while continuously without disrupting its normal
operation.
VM Downtime: the period during which VM does
not reply to ICMP echo request from the remote
host. Total downtime is counted by summing of
all lost packets multiplied by requests interval.
This interval was set up between 2 consecutive
ICMP packets to 200 ms.
Total amount of data transferred: Amount of data
transferred from source to destination host during
migration process via the network measured in
MB. We used the iftop tool (IFTOP-Tool, 2019)
to measure transferred data during the migration
process.
The results show us a considerable percentage dif-
ference for each lower-level use case between a VM
with a high CPU and memory usage (loaded virtual
machine) and a unload virtual machine. This is be-
cause the amount of data to be transferred and the
number of messages to synchronize at the moment of
the migration are higher in a loaded resource than an
unloaded one.
Also, as the previous results for the resource and
management use cases, the difference between the
smaller and larger images increased; but in a reduced
rate concerning the percentage reported in loaded
against unloaded resources.
6 CONCLUSION AND FURTHER
WORK
According to the rapid growth of the cloud federa-
tion paradigm, network federation has become more
important and necessary in a federated networks en-
vironment. The capability of having NFV services
on federated infrastructures as cloud services facil-
itate the implementation of new use cases, such as
geographically-wide workload migration, the defini-
tion of priority network paths, confinement of re-
sources to cloud regions due to legal constraints or
increased multitenancy in network management.
Current solutions focus mainly on user-based and
centralized solutions (such as the set up of VPN
servers), the experiments performed in this article
show the possibility to have a flexible middleware tool
that allows us a complete network federation and at
the same time supports VM live migration in a WAN
network. Further updates to the framework will al-
low, first, support to container networks and combin-
ing them with OpenVswitch to create overlay WAN
networks in a containers environment and second test
the framework in an intercontinental scenario inte-
grating the model with the fogbow federation system
(Brasileiro et al., 2016).
IWFCC 2019 - Special Session on Federation in Cloud and Container Infrastructures
606
ACKNOWLEDGEMENTS
This work is supported by SENESCYT (Secretar
´
ıa
de Educaci
´
on Superior, Ciencia, Tecnolog
´
ıa e Inno-
vaci
´
on - Ecuador) and partially funded by the AT-
MOSPHERE Project (funded by the European Com-
mission under the Cooperation Programme, Horizon
2020 grant agreement No 777154).
REFERENCES
Andrade, I., de Alfonso, C., and Blanquer, I. (2017). Defin-
ing and testing virtual federated networks across mul-
tiple cloud sites. In 2017 Second International Con-
ference on Fog and Mobile Edge Computing (FMEC),
pages 107–112.
Bradford, R., Kotsovinos, E., Feldmann, A., and Schi
¨
oberg,
H. (2007). Live wide-area migration of virtual ma-
chines including local persistent state. In Proceedings
of the 3rd International Conference on Virtual Exe-
cution Environments, VEE ’07, pages 169–179, New
York, NY, USA. ACM.
Brasileiro, F., Vivas, J. L., d. Silva, G. F., Lezzi, D., Diaz,
C., Badia, R. M., Caballer, M., and Blanquer, I.
(2016). Flexible federation of cloud providers: The
eubrazil cloud connect approach. In 2016 30th In-
ternational Conference on Advanced Information Net-
working and Applications Workshops (WAINA), pages
165–170.
Calico (2019). Calico official website.
https://www.projectcalico.org/. Accessed: 2019-
01-12.
Carlini, E., Coppola, M., Dazzi, P., Ricci, L., and Righetti,
G. (2012). Cloud federations in contrail. In Alexan-
der, M., D’Ambra, P., Belloum, A., Bosilca, G., Can-
nataro, M., Danelutto, M., Di Martino, B., Gerndt, M.,
Jeannot, E., Namyst, R., Roman, J., Scott, S. L., Traff,
J. L., Vall
´
ee, G., and Weidendorfer, J., editors, Euro-
Par 2011: Parallel Processing Workshops, pages 159–
168, Berlin, Heidelberg. Springer Berlin Heidelberg.
Consul (2019). Consul official website.
http://www.consul.io. Accessed: 2019-01-10.
Coreos (2019). Coreos official website.
https://www.coreos.com/etcd/. Accessed: 2019-
01-10.
EC-Managed (2019). Ec managed official website.
https://www.ecmanaged.com/. Accessed: 2019-01-
10.
Elmroth, E., Marquez, F. G., Henriksson, D., and Fer-
rera, D. P. (2009). Accounting and billing for fed-
erated cloud infrastructures. In 2009 Eighth Interna-
tional Conference on Grid and Cooperative Comput-
ing, pages 268–275.
Flannel (2019). Flannel official website.
https://www.coreos.com/etcd/. Accessed: 2019-
01-10.
Floodlight-Project (2019). Floodlight project official web-
site. http://www.projectfloodlight.org/floodlight. Ac-
cessed: 2019-01-13.
Hassan, M. M., Song, B., and Huh, E.-N. (2010). A market-
oriented dynamic collaborative cloud services plat-
form. annals of telecommunications - annales des
t
´
el
´
ecommunications, 65(11):669–688.
IFTOP-Tool (2019). Iftop tool. http://www.ex-
parrot.com/pdw/iftop/. Accessed: 2018-01-28.
Kaur, K., Kaur, S., and Gupta, V. (2016). Software de-
fined networking based routing firewall. In 2016 In-
ternational Conference on Computational Techniques
in Information and Communication Technologies (IC-
CTICT), pages 267–269.
Khondoker, R., Zaalouk, A., Marx, R., and Bayarou, K.
(2014). Feature-based comparison and selection of
software defined networking (sdn) controllers. In
2014 World Congress on Computer Applications and
Information Systems (WCCAIS), pages 1–7.
Kubernetes (2019). Kubernetes official website.
https://kubernetes.io/. Accessed: 2019-01-13.
Kuribayashi, S. (2013). Improving quality of service
and reducing power consumption with wan accel-
erator in cloud computing environments. CoRR,
abs/1302.1921.
KVM (2019). Kvm official website. https://www.linux-
kvm.org/page/Main Page. Accessed: 2019-01-16.
Libvirt-project (2019). Libvirt official website.
https://libvirt.org/. Accessed: 2018-01-24.
LXC (2019). Linux container official website.
https://linuxcontainers.org. Accessed: 2019-01-
15.
Marosi, A., Kecskemeti, G., Kert
´
esz, A., and Kacsuk, P.
(2011). Fcm: an architecture for integrating iaas cloud
systems. pages 7–12.
McKeown, N., Anderson, T., Balakrishnan, H., Parulkar,
G., Peterson, L., Rexford, J., Shenker, S., and Turner,
J. (2008). Openflow: Enabling innovation in cam-
pus networks. SIGCOMM Comput. Commun. Rev.,
38(2):69–74.
Melhem, S. B., Agarwal, A., Daraghmeh, M., Goel, N., and
Zaman, M. (2017). Live vm migration across cloud
data centers. In 2017 IEEE 14th International Confer-
ence on Mobile Ad Hoc and Sensor Systems (MASS),
pages 654–659.
Mesos (2019). Mesos official website.
http://mesos.apache.org/. Accessed: 2019-01-13.
Moreno-Vozmediano, R., Montero, R. S., and Llorente,
I. M. (2012). Iaas cloud architecture: From virtu-
alized datacenters to federated cloud infrastructures.
Computer, 45(12):65–72.
Newhouse, S. J. and Brewer, S. (2012). Egi: An open e-
infrastructure ecosystem for the digital european re-
search area and the humanities. In Ioannides, M.,
Fritsch, D., Leissner, J., Davies, R., Remondino, F.,
and Caffo, R., editors, Progress in Cultural Her-
itage Preservation, pages 849–856, Berlin, Heidel-
berg. Springer Berlin Heidelberg.
NFS (2019). Network file system official website.
http://linux-nfs.org/wiki/index.php/Main
Page//. Ac-
cessed: 2019-01-14.
OFS (2019). Offline file system official website.
http://offlinefs.sourceforge.net/wiki/. Accessed:
2019-01-20.
Easing the Deployment and Management of Cloud Federated Networks Across Virtualised Clusters
607
OpenStack (2019). Openstack official website.
http://www.openstack.org. Accessed: 2019-01-
11.
OpenVswitch (2019). Openvswitch official website.
https://www.openvswitch.org/. Accessed: 2019-01-
22.
Rochwerger, B., Breitgand, D., Levy, E., Galis, A., Nagin,
K., Llorente, I. M., Montero, R., Wolfsthal, Y., Elm-
roth, E., Caceres, J., Ben-Yehuda, M., Emmerich, W.,
and Galan, F. (2009). The reservoir model and ar-
chitecture for open federated cloud computing. IBM
Journal of Research and Development, 53(4):4:1–
4:11.
Rodriguez, V. K. Q. and Guillemin, F. (2016). Performance
analysis of resource pooling for network function vir-
tualization. In 2016 17th International Telecommu-
nications Network Strategy and Planning Symposium
(Networks), pages 158–163.
Romana (2019). Romana official website. ttp://romana.io/.
Accessed: 2019-01-12.
Silvera, E., Sharaby, G., Lorenz, D., and Shapira, I. (2009).
Ip mobility to support live migration of virtual ma-
chines across subnets. In Proceedings of SYSTOR
2009: The Israeli Experimental Systems Conference,
SYSTOR ’09, pages 13:1–13:10, New York, NY,
USA. ACM.
Ubuntu-Cloud-images (2019). Ubuntu official website.
lhttps://cloud-images.ubuntu.com/. Accessed: 2018-
01-24.
Weave (2019). Weave official website.
https://www.weave.works/. Accessed: 2019-01-
11.
Wood, T., Ramakrishnan, K. K., Shenoy, P., der Merwe,
J. V., Hwang, J., Liu, G., and Chaufournier, L. (2015).
Cloudnet: Dynamic pooling of cloud resources by live
wan migration of virtual machines. IEEE/ACM Trans-
actions on Networking, 23(5):1568–1583.
Zhang, F., Fu, X., and Yahyapour, R. (2018). Layermover:
Fast virtual machine migration over wan with three-
layer image structure. Future Generation Computer
Systems, 83:37 – 49.
ZHAW (2014). An analysis of the perfor-
mance of block live migration in openstack.
https://blog.zhaw.ch/icclab/an-analysis-of-the-
performance-of-block-live-migration-in-openstack/.
IWFCC 2019 - Special Session on Federation in Cloud and Container Infrastructures
608