-This paper introduces a renewed gateway to ENEAGRID distributed computing resources, named Fast Access to Remote Objects 2.0 (FARO 2.0). FARO 2.0 is a tool for application and desktop virtualization with a strong focus towards user experience (UX), providing trained as well as untrained users a collection of centralized services that can be seamlessly used on their client through a remote desktop protocol. FARO 2.0 is a JavaFX application whose graphical user interface (GUI) and whose main logics has been implemented through the wellknown Web technologies (HTML5, CSS3, Javascript) for a easier maintainability and customizability, taking full advantage of the WebView component. Its framework has been deployed both as general purpose GUI for remote user access to ENEAGRID resources and as specialized application or workflow oriented GUI. They are applied in a set of applicative domains, ranging from material science to technologies for energy and industry, environmental modeling and nuclear fusion. Some examples and results are also presented.
I. INTRODUCTION
The need for a centralized access to computing resources
has resulted in an extensive research towards software tools,
such as scientific gateways, to allow an intuitive and
easyto-use fruition of remote services (e.g. application and
desktop virtualization, storage, data . . . ). Distributed computing
infrastructures (such as grids, clouds), as well as
generalpurpose yet highly specialized cloud facilities, are now
empowering scientists in a growing number of tasks from their
everyday work life. Furthermore, they are even becoming a
key foundation technology even for untrained users in many
areas beyond scientific research, such as in education. For this
reason, modern scientific gateways should be developed with
a strong focus towards user experience (UX), through a high
degree of interactivity in a pleasing graphical user interface
(GUI), just like any mainstream application for smartphone
and tablets (an “app”) that users already understand and
appreciate. Moreover, they should foresee a high degree of
maintenability and customizability, through the usage of
wellknown technologies. ENEA, the Italian National Agency for
New Technologies, Energy and Sustainable Economic
Development is offering a completely renewed gateway for its
High Performance Computing (HPC) services, called FARO
2.0 (Fast Access to Remote Objects) to let internal as well as
external scientists take advantage of its remote services based
on ENEAGRID distributed computing resources and CRESCO
linux clusters [
1
].
II. MAIN PICTURE
Developed in the contest of the Italian PON “Smart
Cities and Communities” R&C 2007-2013 with the project
“EDOC@Work 3.0 - Education and work in the cloud” [
2
]
in collaboration with Sysman Progetti & Servizi S.r.l. [
3
],
FARO 2.0 is a tool for scientists and students to perform
research as well as train in a real-world HPC environment.
FARO 2.0 has been implemented as a drop-in replacement for
its predecessor, FARO [
4
], it thus relies on its same network
infrastructure, depicted in Fig.1, and remote desktop protocol.
The main goal of FARO 2.0 is to provide users a software
solution with a strong focus towards UX, maintainability
and customizability, being a pleasant interface for real-time
application and desktop virtualization (with remote graphics
acceleration) on conventional as well as on “special” hardware
platforms (featuring Intel MICs or high performance GPUs).
The remote desktop protocol used to deliver applications
and desktops to end-users is based on NX over an SSH.
The compression and transport protocol of NX is used to
enhance the native X display protocol performances in such a
way that services are usable even with slower links. The NX
procotol has been released by Nomachine [
5
], but many open
source server and client implementations exists [
6
], [
7
]. FARO
2.0 remote desktop server is based on a customized version
of FreeNX, where we implemented some advanced features
(e.g. load balancing and session distribution over a cluster).
Accreditated users can connect to FARO 2.0 by downloading
a pre-configured NX client (as well as a session file) from its
web portal (see Fig.2).
It seems important to remark that application and desktop
virtualization is one among the many tools and services that
ENEAGRID offers to its users and administrators, such as web
portals for batch job submission and monitoring (JobRama),
cloud storage (AFSBox), distributed file system monitoring
(AMACA) and help-desk (GridTicket). The Authentication
and Authorization Infrastructure (AAI) of ENEAGRID is
based on Kerberos 5 [
8
]; it is integrated with AFS distributed
filesystem and its Access Control List (ACL) system [
9
] and
is used for all the services provided by the infrastructure (see
Fig.3).
III. FARO 2.0 FARO 2.0 is a JavaFX application whose graphical user interface (GUI) and whose main logic has been imple2
mented through well-known Web technologies (HTML5,
CSS3, Javascript) for an easier maintainability and
customizability, taking full advantage of the WebView component.
The application is delivered to the end-user (as soon as
his NX session is authenticated) by a load balanced graphic
backend in a CRESCO cluster: this means that FARO 2.0 is
executed on ENEAGRID and is served through an entirely
managed ecosystem, where a team of administrators takes care
of the runtime environment in order to ensure its consistent
behavior. Consider that all software packages and components,
including the Java runtime environment required by JavaFX,
are provided and deployed on a distributed filesystem that
connects all computing resources in ENEAGRID. Thus, the
only component that users have to install on their machines is
the NX client, whereas any other tool and service is remotely
executed and rendered on the local screen with near native
performances through the NX protocol. This tool and service
distribution model ensures better security for data and simpler
deployment of complex scientific softwares and their updates,
since everything runs on machines that are monitored and
serviced by our administrators.
The core of FARO 2.0 is its desktop application (a container,
in this context) developed in JavaFX, showing a WebView.
Instead, its GUI and its main logics have been developed in
HTML5, CSS3 and Javascript as a single page application
(SPA), replicating the trending paradigm of cross-platform
hybrid apps. The container is a lightweight piece of software
that implements an ordinary browser, rendering a local web
page. However, the container also injects a custom Java class
(a bridge) in the Javascript global namespace. Thus, through
Javascript calls, the above web page is able to execute any
member function that has been made available from the
container. In this implementation, the bridge exports methods
to launch new processes based on CLI commands and
implements the application logic needed to callback registered
Javascript member functions in case of a new message in the
standard-out or in the standard-error streams; this allows to
virtually execute any command on the host backend server that
is executing FARO 2.0. Moreover, every command executed
is tracked in a user home log that allows administrators to
troubleshoot every problem may occur in the execution of
remote code. A convenient call from the Javascript runtime
environment is, for example, able to let the user remotely
open and interact with the command prompt of a standard
or an advanced node in a cluster, or launch any kind of
scientific software in a dedicated environment (e.g. MATLAB,
COMSOL, IDL). More precisely, each launch initiated from
FARO 2.0 is redirected through the SSH protocol to the ENEA
instance of IBM Load Sharing Facility (LSF), the scheduler
that enqueues requests both for interactive and batch jobs [
10
].
As a matter of fact, any interaction with the ENEAGRID
environment can be routed through FARO 2.0. The main
interface is shown in Fig.4-5-6.
FARO 2.0 is easy to maintain and customize, since most
of its code reside outside the container package and merely
implements a SPA (with bindings to some Java member
functions): this allows to reconfigure quite the entire application
without the need for a compile process. Moreover, the SPA
can run (with a limited functionality) within a web browser; it
can thus be previewed in a very agile way. We are even able
to provide deeply customized interfaces by creating branches
of the SPA, without the need to modify the container. There is
definitely no need to know how FARO 2.0 actually manages
the launch of remote tools and services in order to create
new virtual research environments (e.g. gateways for highly
specialized use cases, collecting scientific tools targeting a
well-defined research context), such as what we call the
“ENEA Virtual labs” (as an example see [
14
]). This science
gateway plays an important role also on computational
chemistry applications based on software like Quantum Espresso
and cp2k, remotely managed by ENEAGRID CMAST Virtual
Lab [
15
]. From a security standpoint, it seems important to
remark that any CLI command executed by FARO 2.0 is
authenticated as if the user wrote the command on a shell, and
there is no way to execute a malicious command with more
privileges (e.g. privilege escalation) from FARO 2.0 code.
Moreover, the WebView scope is constrained to local pages
in order to avoid cross-site scripting.
Through the usage of web standards, FARO 2.0 implements
a full-featured gateway for application and desktop
virtualization in the ENEAGRID distributed computing environment
that can also be used as a boilerplate for new applications,
such as the “ENEA Virtual Labs”.
IV. COMPARISON WITH OTHER TOOLS
FARO 2.0 offers many similarities with a commercial tool
named RemoteApp from Microsoft Corp [
11
]. RemoteApp is a
software tool based on the Microsoft Remote Desktop Protocol
implementing a launcher for applications hosted on Microsoft
Azure. It must be installed on the user device and is available
for Windows, iOS, Mac OS X and Android. This solution is
not as flexible as FARO 2.0 because it requires a set of specific
OS and resources in order to deliver a service remotely.
Another platform that we considered is Citrix Workspace
Cloud, that is built around Citrix proprietary protocol [
12
]. It
offers a complete solution to every remote workspace request.
This protocol was used in the past by ENEA [
13
] but was
abandoned due to commercial costs and more stringent
technical requirements. Instead, FARO 2.0 is distributed toward
users as a remote application itself, and for this reason it works
on any platform where an NX client is available. Besides this,
consider that the JavaFX desktop application at the core of
FARO 2.0 is in some way independent from remote desktop
protocol, as it can be deployed through every protocol one
can choose to render a remote desktop. The NX technology
has been GPL-licensed until the 3.5 version and until now
it offers a set of open source and free tools that can ensure
a lot of flexibility for our solution. Its protocol is optimized
for low bandwidth connections and its performances attracted
interests even from Google [
16
]. We successfully employed
this technology in ENEAGRID even for heavy tasks like
remote 3D rendering [
17
].
V. APPLICATIONS
In the following we will show some interesting applications
and customization of FARO 2.0.
The first application is related to the ADP Virtual Lab [
18
].
The aim is to provide a launcher for a code for the analysis
of small-powered wood biomass energy systems (so called
COGEGNO). The code has been implemented in Matlab and
Scilab environments, and transferred on a web platform, in a
“virtual laboratory” powered by the computational systems of
ENEAGRID. Among the main components of the model there
are the gas recirculation and air staging mobile-grid burner
and the thermic gain obtained from the pre-heating of the air
in input to the boiler. For the model of the boiler the code
uses an helicoidal flow one, with particular attention to the
molten salts, for the advantages arising from the possibility
offered by these fluids to work in a broader temperature
range and at higher temperatures, increasing the gain of
the plant. If the power and the electrical efficiency of the
cogeneration system are known, the code is able to calculate
the performance and the sizes of the main components of
the heat generator not only in nominal conditions but also
if the operational conditions vary. The code is installed on
the AFS filesystem. In particular Scilab code was compiled
and only the executable files are stored in AFS. Using the
FARO 2.0 interface, users can pass the values of the parameters
to the code and choose the operational conditions between
dynamically loaded options. These parameters are passed to
CLI commands through an easy-to-understand JSON format,
that the backend computational core is able to interpret and
then execute. A sample illustration of the customized interface
is in Fig.7-8-9. Consider that the entire architecture design
of FARO 2.0 has strict security policies based on AFS ACLs
and allow software developers to deploy their products hiding
the implementation; in this way any user can execute scientific
codes without accessing directly to them.
A second application of FARO 2.0 is developed for the Web
Crawling Virtual Lab [
19
]. The aim of this tool integrated
in the related ENEA project is to create a simple interface
beetwen the users and the Web crawling environment installed
on ENEAGRID. The Web crawling is an activity that
automatically and systematically explores the Web, in order to
search for contents/documents to download. Starting from the
main html page of the interface, an user can create a web
crawling session, launch the session and monitor the generated
internet traffic. More specifically, this interface is composed of
five tabs, in each of them, one can perform some operations.
In the first tab, there is a form where one can insert some
general informations (such as the title and comments) about
the session that wants to create. In the second tab, a user
can fill in a module with the configuration options to pass
to resource scheduler in order to submit a session job. These
options are split between the run configurations and software
parameters. As run configuration an user can set the running
time of job, the number of nodes to use and the number of
agents per node. An agent is a Java process that plays the
role of a crawler (i.e. a software that explores and downloads
the web pages). As software parameters it is possible to can
set many options for each crawler (e.g., the URLs seed to
initially consider, the number of threads and the size of cache
memories). If the user does not decide to set these options,
automatically the interface creates a session configuration file
with default options. In the third tab, an user can submit a
batch job for the current session. In the fourth tab, once a web
crawling session is submitted, the user can monitoring in real
time the downloaded data amount by every agent. Finally, the
fifth tab allows to view some statistics at the end of the latest
executed session. The tool has been recently equipped with
a further feature called “snapshot”, which allows to schedule
periodic web crawling sessions. This is particularly useful to
create a set of snapshots for some portions of the web, in order
to analyze the changes and the evolution over a time period.
Periodic scheduled web crawling session parameters can be
set in the proper section in the tool interface. Customized
interfaces are shown in Fig.10-11
For both applications, in the submission step, ad hoc shell
scripts are used. These scripts read the parameters defined on
the mask and usually encoded in a JSON format and launch the
codes on execution nodes following the rules defined by the
LSF scheduler. The applications provide also an Help page
with an useful set of links for users (to submit tickets to
ENEAGRID administrators and/or to monitor job execution).
Every customized FARO 2.0 application is a
workfloworiented GUI enabling interaction with ENEAGRID
computational resources and easily interoperates with every set of
scientific tools installed on our grid/cloud environment.
VI. FUTURE DEVELOPMENTS The most challenging part of the above architecture is ensuring our users a consistent behavior of the remote desk
top protocol over as many client configurations as possible.
The emerging WebRTC APIs implemented in modern web
browsers [
20
], open up the new opportunity for a
crossplatform web-based remote desktop experience, that could
even avoid the need to perform the client software download
to start using FARO 2.0. In the meanwhile our research is
targeted towards remote desktop protocols that can enhance
the features of FARO 2.0. An experimental setup has been
realized with Cendio ThinLinc [
21
], that is based on SSH
and VNC (see Figs.12-13-14), and we are planning also tests
with other clientless remote desktop gateway, like for example
Guacamole [
22
] or with remote viewer open source software
like Spice [
23
].
As it was stressed before, the main interface is independent
from remote desktop protocol, so it can be reproduced over
different connection types between desktop clients and the
ENEAGRID computing environment.
FARO 2.0 is a gateway for application and desktop
virtualization, as well as a boilerplate to easily develop new virtual
research environments. Being distributed toward users as a
remote application itself, FARO 2.0 is executed on ENEAGRID
and is served through an entirely managed ecosystem, thus
ensuring better security for data and simpler deployment of
complex scientific softwares and their updates, since
everything runs on machines that are monitored and serviced by
our administrators. Most part of FARO 2.0 has been developed
through web technologies (HTML5, CSS3, Javascript), and
thus is easy to maintain and customize, even without the need
for a compile process. FARO 2.0 uses an open source
implementation of NX as its remote desktop procotol, and an NX
client is the only software component that users have to install
to start using it. The future usage of WebRTC based clients
may lead to an even leaner remote desktop experience. We
believe that application and desktop virtualization constitute an
excellent tool and service distribution model, that lets scientists
focus more on research since relying on a high performance
environment that is completely managed by our administrators.
ACKNOWLEDGMENT
The authors would like to thank Alessio Rocchi who
was the former developer of FARO application and all
the people involved in the management and operation of
ENEAGRID/CRESCO infrastructure [
24
]. We also thank
Matteo Caldera who is in charge of the computational core of
the ADP Virtual Lab. FARO 2.0 has been developed in the
contest of the Italian PON “Smart Cities and Communities”
R&C 2007-2013 with the project “EDOC@Work 3.0 -
Education and work in the cloud”. Part of this activity has been
supported by the ENEA-Forschungszentrum JULICH GmbH
contract “Fornitura da parte ENEA di attivita` di modellistica
molecolare”.
[1]
G.Ponti
et al,
The role of medium size facilities in the HPC ecosystem: the case of the new CRESCO4 cluster integrated in the ENEAGRID infrastructure inProceedings of the International Conference on High Performance Computing and Simulation
,
HPCS2014
, Bologna, Italy,
21
-
25July
,
2014[2] https://www.edocwork.it/
[3] http://www.sys
-man
.it/
[4]
A.Rocchi
,
S.Pierattini
, G. Bracco,
S.Migliori
,
F.Beone
,
A.Santoro
, C. Scio`, S. Podda, FARO - The Web portal to access
ENEAGRID Computational Infrastructure in Proceedings of the International Workshop on Science Gateways (IWSG2010)
, Catania, Italy,
R.Barbera
,
G.Andronico and G. La Rocca
(Eds.),
ConsorzioCOMETA
(
2010
).
ISBN 978-88- 95892-03-0[5] https://www.nomachine.com/[6] https://sourceforge.net/projects/freenx.berlios/[7] http://opennx.net/[8] http://web.mit.edu/kerberos/[9] https://www.openafs.org/[10] http://www-03.ibm.com/systems/platformcomputing/products/lsf/[11] https://azure.microsoft.com/en-us/services/remoteapp/[12] https://www.citrix.com/products/workspace-cloud/overview.html
[13]
G.Bracco
et al,
CRESCO HPC System integrated into ENEA-GRID environment inProceedings of the Final Workshop of the Grid Projects of the Italian National Operational Programme2000
-2006 Call 1575,
Catania
, Italy,
10
-
12
February, 2009
ConsorzioCOMETA
(
2009
).
ISBN978
-88-95892-02-3
[14]
F.Ambrosino
,
A.Colavincenzo
,
G.Guarnieri
,
A.Funel
,
G.Ponti
,
A.Mariano
,
F.Palombi
, G. Bracco, S. Migliori,
NEPTUNIUS: the ENEA HPC portal for multiphysics simulations
,
ISC2015
, Frankfurt 12-
16July
2015
[15] http://www.afs.enea.it/project/cmast/[16] https://code.google.com/archive/p/neatx/[17] https://www.ark3d.enea.it/home.php[18] http://www.afs.enea.it/project/adp/[19] http://www.afs.enea.it/project/webcrawl/[20] https://webrtc.org/[21] https://www.cendio.com/thinlinc/what-is-thinlinc[22] http://guac-dev.org/[23] http://www.spice-space.org/[24] http://www.eneagrid.enea.it/people/2015EneaGridPeople.html