Resource and Application Description Language (RADL)

The main purpose of the Resource and Application description Language (RADL) is to specify the requirements of the scientific applications needed to be deployed in a virtualized computational infrastructure (cloud). Using a declarative scheme RADL considers distinct features related to

  • hardware, like CPU number, CPU architecture, and RAM size;

  • software, like applications, libraries and data base systems;

  • network, like network interface and DNS configuration; and

  • contextualization, extra steps to set up an adequate environment for the application.

RADL is intended to be more abstract that other standards to specify virtual appliances, like OVF, and easily extensible with other tools, like contextualization languages such as Ansible.

Basic structure

An RADL document has the next general structure:

description <desc_id> (<features>)

ansible <ansible_host_id> (<features>)

network <network_id> (<features>)

system <system_id> (<features>)

configure <configure_id> (<Ansible recipes>)

contextualize [max_time] (
  option <option_name> = <value>
  system <system_id> configure <configure_id> [step <num>]

deploy <system_id> <num> [<cloud_id>]

The description optional keyword can only appear once. It is an special keyword to add some metadata to the RADL. All features on the list are free, except for the name field. The name field is a predefined but optional field used to assign a name to the infrastructure. For instance:

description desc (
   name = 'Infrastructure Name' and
   some_other = 'Other infra metadata'

The keywords ansible, network, system and configure assign some features or recipes to an identity <id>. The features are a list of constrains separated by and, and a constrain is formed by <feature name> <operator> <value>. For instance:

system tomcat_node (
   memory.size >= 1024M and
   disk.0.applications contains (name='tomcat')

this RADL defines a system with the feature memory.size greater or equal than 1024M and with the feature disk.0.applications containing an element with name tomcat.

The network keyword enables representing different networks so that the VMs can be attached to them.

The ansible keyword enables the specification of external nodes that will act as the ansible master node to configure the VMs. These nodes must be connected in a network with all the VMs of the infrastructure.

The sentences under the keyword contextualize indicate the recipes that will be executed during the deployment of the virtual machine.

The deploy keyword is a request to deploy a number of virtual machines. If some identity of a cloud provider is specified the VM will be deployed in the Cloud provider with the “id” specified.

Use Cases

RADL is not limited to deploy different configurations of virtual machines easily. In many applications, infrastructures need management during their life cycle, like deploying virtual machines with new features, changing the features of already deployed virtual machine and undeploying some of them. Next we detail valid RADL examples for every use.

Create a New Infrastructure

A common RADL defines a network and at least one kind of virtual machine and deploys some virtual machines. However the minimum RADL document to create an infrastructure is an empty one.

Add New Definitions

After the creation of the infrastructure, new networks, systems and recipes can be defined. The new definitions can refer to already defined elements, but they must be mentioned. For instance, an infrastructure is created as:

network net (outbound = 'no')
system small_node (
   cpu.arch = 'x86_64' and
   cpu.count = 1 and
   memory.size >= 512M and
   net_interface.0.connection = 'net' and = 'linux'

A new system with more memory and CPUs, and in the same network can be defined as:

network net
system big_node (
   cpu.arch = 'x86_64' and
   cpu.count = 4 and
   memory.size >= 3G and
   net_interface.0.connection = 'net' and = 'linux'

Deploy New Virtual Machines

In the same way, new virtual machines from already defined systems can deployed. For instance, this example deploys one small_node and other big_node:

system small_node
system big_node

deploy small_node 1
deploy big_node 1

Ansible Features

Under the keyword ansible there are the features needed to access the ansible master node with SSH. The supported features are:

host = '<ip or hostname>'

Indicate the hostname or IP of the ansible node.

credentials.username = '<username>'

Indicate the SSH username.

credentials.password = '<password>'

Indicate the SSH password.

credentials.private_key = '<private_key>'

Indicate the SSH private key.

Network Features

Under the keyword network there are the features describing a Local Area Network (LAN) that some virtual machines can share in order to communicate to themselves and to other external networks. The supported features are:

outbound = 'yes|no'

Indicate whether the IP that will have the virtual machines in this network will be public (accessible from any external network) or private. If yes, IPs will be public, and if no, they will be private. The default value is no.

outports = <outports_string>

Indicate the ports to be open in the VM at the Cloud provider system. Valid formats:

    • 8899/tcp-8899/tcp,22/tcp-22/tcp

    • 8899/tcp-8899,22/tcp-22

    • 8899-8899,22-22

    • 8899/tcp,22/udp

    • 8899,22

    • 9000:9100/tcp

    • 9000:9100

The usage of - means port mapping the first port (remote) will be opened and redirected the the second port (local). The usage of : means port range. It can also be specified the remote CIDR allowed to access the specified port. The default value is ''.

provider_id = <string>

Indicate the name of the network in a specific Cloud provider. If not set the IM will try to select the “default” option in the Cloud provider. In case of setting this field in a public network in an OpenStack deployment it specifies the name of floating ip pool to get the external floating IP. In case of Amazon EC2 this field follows the format vpc-id.subnet-id. The default value is ''.

create = 'yes|no'

Indicate if the IM must create the network or will try to find the most appropriate from the existing networks. In some connectors (e.g. Azure) the networks are always created independently the value of this parameter. The default value is 'no'.

cidr = <string>

Indicate the CIDR of the network (e.g. in case of network creation. Wildcards can be used (i.e. 10.*.*.0/24) and the IM will select the first option that is not used in the current Cloud provider. The default value is ''.

sg_name = <string>

The name of the Security Group associated with the network that will be created to manage the security in this network. The default value is ''.

router = <string>

Add static routes in the network settings. Currently only supported in OpenStack, GCE and AWS. The format is ‘net_cidr, system_name’ e.g. ‘,front’ to route all the traffic to the net through the front node, or ‘,front’ to route all the traffic through the front node. The default value is ''.

proxy_host = <string>

The URI of a node to use it as an SSH proxy to connect with nodes in private networks. Format of the URI: ‘username:password@servername:port’. The default value is ''.

proxy_key = <string>

Set the private key to connect with the proxy host. If it is specified it overrides the password set in the proxy_host url. The default value is ''.

System Features

Under the keyword system there are the features describing a virtual machine. The supported features are:

ansible_host = '<ansible_host id>'

Set the ansible master node that will contextualize the virtual machine. The ansible host need to have ansible (2.0 or later) installed and the ansible.cfg file configured with similar values than the ansible in the IM server.

image_type = vmdk|qcow|qcow2|raw

Constrain the virtual machine image disk format.

virtual_system_type = '<hypervisor>-<version>'

Constrain the hypervisor and the version used to deploy the virtual machine.

price <=|=|=> <positive float value>

Constrain the price per hour that will be paid, if the virtual machine is deployed in a public cloud.

cpu.count <=|=|=> <positive integer value>

Constrain the number of virtual CPUs in the virtual machine.

cpu.arch = i686|x86_64

Constrain the CPU architecture.

cpu.performance <=|=|=> <positive float value>ECU|GCEU

Constrain the total computational performance of the virtual machine.

memory.size <=|=|=> <positive integer value>B|K|M|G

Constrain the amount of RAM memory (principal memory) in the virtual machine.


Features under this prefix refer to virtual network interface attached to the virtual machine.

net_interface.<netId>.connection = <network id>

Set the virtual network interface is connected to the LAN with ID <network id>.

net_interface.<netId>.ip = <IP>

Set a static IP to the interface, if it is supported by the cloud provider.

net_interface.<netId>.dns_name = <string>

Set the string as the DNS name for the IP assigned to this interface. If the string contains #N# they are replaced by a number that is distinct for every virtual machine deployed with this system description.


Set the availability zone or region where this VM will be launched. It only applies to Google Cloud, Microsoft Azure, Amazon AWS, and Fogbow connectors. In the Fogbow case it specifies the site and cloud where the VM will be launched (in format cloud@site).


Get the instance ID assigned by the Cloud provider for this VM.


Set the instance name for this VM.


Set the instance type name of this VM.


A set of keypair values to be set to the VMs. With the following format: key=value,key2=value2 …


Features under this prefix refer to virtual storage devices attached to the virtual machine. disk.0 refers to system boot device.

disk.<diskId>.image.url = <url> or [comma separated list of urls]

Set the source of the disk image. The URI designates the cloud provider:

  • one://<server>:<port>/<image-id>, for OpenNebula;

  • one://<server>:<port>/<image-name>, for OpenNebula;

  • ost://<server>:<port>/<image-id>, for OpenStack or EGI;

  • aws://<region>/<ami-id>, for Amazon Web Service;

  • aws://<region>/<snapshot-id>, for Amazon Web Service;

  • aws://<region>/<snapshot-name>, for Amazon Web Service;

  • gce://<region>/<image-id>, for Google Cloud;

  • azr://<image-id>, for Microsoft Azure Clasic;

  • azr://<publisher>/<offer>/<sku>/<version>, for Microsoft Azure;

  • azr://[snapshots|disk]/<rgname>/<diskname>, for Microsoft Azure;

  • <fedcloud_endpoint_url>/<image_id>, for FedCloud OCCI connector.

  • appdb://<site_name>/<apc_name>?<vo_name>, for FedCloud OCCI, OpenStack or EGI connectors using AppDB info (from vers. 1.6.0, 1.8.6 and 1.10.2 respectively). In case of EGI connector the vo_name is not required as it will be get from auth data.

  • appdb://<apc_name>?<vo_name>, for FedCloud OCCI, OpenStack or EGI connectors without setting site_name.

  • docker://<docker_image>, for Docker images.

  • fbw://<fns_server>/<image-id>, for FogBow images.

  • lin://linode/<image-id>, for Linode images.

  • ora://<region>/<image-id>, for Orange Flexible Engine images.

In case of using a list of URLs, the IM will select the final image based on the credentials provided by the user.

disk.<diskId> = <string>

Set the source of the disk image by its name in the VMRC server.

disk.<diskId>.device = <string>

Set the device name, if it is disk with no source set. It specifies the device where the disk will be located in the system (hdb, hdc, etc.). Depending on the Cloud provider the meaning of this field may change. In Docker connector the device refers to a path to create a bind in the container, if it starts with character / or the name of a volume otherwise.

disk.<diskId>.mount_path = <string>

Set the mount point, if it is disk with no source set. It specifies a path to mount the device. In Docker and Kubernetes connectors this path refers to the directory in the container to mount a PVC or the bind host directory specified in device.

disk.<diskId>.fstype = <string>

Set the filesystem, if it is disk with no source set. It specifies the type of the filesystem of this disk. If specified the contextualization agent will try to format and mount this disk in the path specified in mount_path field. In case of Docker the fstype refers to the driver to use in case of using a volume.

disk.<diskId>.size = <positive integer value>B|K|M|G

Set the size of the disk, if it is a disk with no source set.

disk.<diskId>.type = <string>

Set the type of the disk, if it is a disk with no source set. The types depends on the provider: e.g. in GCE posible types are: pd-standard | pd-ssd, in EC2 possible values are: standard | io1 | gp2. In OpenStack possible values are ephemeral, or any volume type supported by the provider.

disk.0.free_size = <positive integer value>B|K|M|G

Set the free space available in boot disk.

disk.<diskId> = linux|windows|mac os x

Set the operating system associated to the content of the disk.

disk.<diskId>.os.flavour = <string>

Set the operating system distribution, like ubuntu, centos, windows xp and windows 7.

disk.<diskId>.os.version = <string>

Set the version of the operating system distribution, like 12.04 or 7.1.2.

disk.0.os.credentials.username = <string> and disk.0.os.credentials.password = <string>

Set a valid username and password to access the operating system with sudo privileges.

disk.0.os.credentials.public_key = <string> and disk.0.os.credentials.private_key = <string>

Set a valid public-private keypair to access the operating system with sudo privileges. = <string> and = <string>

Changes the credentials of the user with admin privileges.

disk.<diskId>.applications contains (name=<string>, version=<string>, preinstalled='yes|no')

Set that the disk must have installed the application with name name. Optionally a version can be specified. Also if preinstalled is yes the application must have already installed; and if no, the application can be installed during the contextualization of the virtual machine if it is not installed.

There are some special type of application that starts with ansible.roles. (ansible.modules. in < IM 1.14 ) or ansible.collections.. These applications installs ansible roles or ansible collections that can be used in the configure sections of the RADL. These roles/collections will be installed with the ansible-galaxy tool so the format of the string after ansible.xxxx. must follow one of the supported formats of this tool (see Ansible Galaxy docs for more info):

There are three type of ansible modules/roles:

  • Ansible Galaxy roles: ansible.roles.micafer.hadoop: The user specifies the name of the galaxy role afther the string ansible.roles.

  • HTTP URL: ansible.roles.|hadoop: The user specifies an HTTP URL afther the string ansible.roles.. The file must be compressed. It must contain the ansible role content. Furthermore the user can specify the rolename using a | afther the url, as shown in the example.

  • Git Repo: ansible.roles.git+|hadoop: The user specifies a Git repo (using the git scheme in the URL) afther the string ansible.roles.. Furthermore the user can specify the rolename using a | afther the url, as shown in the example.

nat_instance = yes|no

Set that this instance will be used as a NAT router for a set of nodes. It will configure the node to enable nat with the appropriate iptables rules (experimental).

gpu.count <=|=|=> <positive integer value>

Constrain the number of virtual GPUs in the virtual machine.

gpu.vendor = <string>

Constrain the vendor name of the GPU in the virtual machine like NVIDIA or AMD.

gpu.model = <string>

Constrain the model name of the GPU in the virtual machine like Tesla-v100 or Radeon RX 5000

rg_name = <string>

The name of the Resource Group associated with the system. It only applies in a subset of connectors.

Disk Management

In the RADL documents there are two different types of disks: disk.0 as the boot disk with the O.S. and the rest of disks assumed as data disks. In the first case if you are using an VMRC server you can specify the features of the requested O.S. and let VMRC to get the most suitable image:'linux' and
disk.0.os.flavour='ubuntu' and

Otherwise you can directly specify the image and, if required, the credentials to access the O.S.:'linux' and
disk.0.image.url = 'one://' and
disk.0.os.credentials.username = 'ubuntu' and
disk.0.os.credentials.password = 'somepass'

In case of the rest of disks you can specify the requirements of the data disk to be attached:

disk.1.size=1GB and
disk.1.device='hdc' and
disk.1.fstype='ext4' and

The fields fstype and mount_path are optional and they enable the IM (through Ansible) to format and mount the disk in the specified path. The device field is optional in most of the connectors but some of them require it to correctly attach the disk to the VM.

You can also specify an image to be attached to the VM:

disk.1.image.url = 'one://' and

Parametric Values

RADL documents can use parametric values to be requested to the user in launch time. It make easy to launch different infrastructures without modifying the RADL document, only changing a set of values in launch time. This parametric values are requested to the user in the launch time by the client application (CLI or Web).

This values are specified with the following syntax:


In the following example the user will be asked for specifing the CPUs and the NumNodes variables (in the CLI and in the Web Interface):

system node (
   cpu.count = @input.CPUs@ and
   memory.size >= 512M
deploy node @input.NumNodes@


RADL documents also enable to specify contextualization, extra steps to set up an

adequate environment for the application.

Configure Recipes

Contextualization recipes are specified under the keyword configure. Only Ansible and Cloud-Init recipes are supported currently. They are enclosed between the tags @begin and @end, like that:

configure add_user1 (
  - tasks:
    - user: name=user1   password=1234

In the Ansible case, to easy some contextualization tasks, IM publishes a set of variables that can be accessed by the recipes and have information about the virtual machine.


Hostname of the virtual machine (without the domain).


Domain name of the virtual machine.


Complete FQDN of the virtual machine.


Private IP of the virtual machine. In case that the VM has more that one the first one will be returned.


Public IP of the virtual machine. In case that the VM has more that one the first one will be returned.


The value of the substitution #N# in the virtual machine.


Cloud type where the VM has been deployed.


Cloud server where the VM has been deployed (if available, if not this variable is not defined).


Hostname (without the domain) of the virtual machine doing the master role.


Domain name of the virtual machine doing the master role.


Complete FQDN of the virtual machine doing the master role.

IM_<application name>_VERSION

The version installed of an application required by the virtual machine.

IM_<application name>_PATH

The path to an installed application required by the virtual machine.


The identifier asigned by the Cloud provider to the virtual machine.

IM_NODE_NET_<iface num>_IP

The IP assigned to the network interface num iface num.


The identifier asigned by the IM to the infrastrucure this VM belongs to.


The RADL in JSON format: networks, systems and deploys. (from ver. 1.6.2). It enables to use RADL values in Ansible recipes. The . in the properties are replaced by _ (e.g. net.interface.0.dns_name is replaced by net_interface_0_dns_name). It can be used in combination with the Ansible json_query filter to extract values as shown in this example:

NODENAME: '{{IM_INFRASTRUCTURE_RADL|json_query("[?id == ''front''].net_interface_0_dns_name|[0]")}}'

Ansible json_query filter is built upon jmespath so this library must be installed on the managed node that uses this function. IM installs it on the master VM but no in the rest of VMs. If you want to use it on other VMs you have to prepare them installing jmespath in a previous step.

Including roles or collections of Ansible Galaxy

To include a role available in Ansible Galaxy a special application requirement must be added: it must start with: “ansible.roles” as shown in the following example. In this case the Ansible Galaxy role called “micafer.hadoop” will be installed:

network net (outbound = 'yes')

system node_ubuntu (
   cpu.arch = 'i686' and
   memory.size >= 512M and
   net_interface.0.connection = "net" and = "linux" and
   disk.0.os.flavour = "ubuntu" and
   disk.0.applications contains (name="ansible.roles.micafer.hadoop")

Then the configuration section of the RADL can use the role as described in the role’s documentation. In the particular case of the “micafer.hadoop” role is the following:

configure wn (
 - roles:
    - { role: 'micafer.hadoop', hadoop_master: 'hadoopmaster' }


You can request an specific version/tag/branch of a galaxy role using the following format:

disk.0.applications contains (name="ansible.roles.micafer.hadoop,v1.0.0")

Similarly, to include a collection available in Ansible Galaxy it must start with: “ansible.collections” as shown in the following example. In this case the Ansible Galaxy collection called “community.crypto” will be installed and can be used in any of the configuration recipes:

system node_ubuntu (
   disk.0.applications contains (name="")

Disable Contextualization

By default the contextualize is performed in all the infrastructures. If the user wants to disable this step he must add an empty contextualize section:

contextualize ()

Advanced Contextualization

By default the IM will apply the configure section to the nodes with the same name of the system defined. Furthermore all configure sections will be executed at the same time, in parallel.

But RADL also enables to specify the order in which the configure sections will be performed and which configure sections will be executed to a specific type of node. It can also be specified the contextualization tool to use en each case.

The contextualize section has the next structure:

contextualize <max_context_time> (
   option <option_name> = <value>
   system <system_id> configure <configure_id> [step <num>] [with (Ansible|cloud_init)]

The max_context_time value enables to set a timeout for the contextualization step to enable to kill the process if some of the steps takes more time than expected.

The optional “option” lines enable to specify some contextualizacion option values. Currently only ansible_version is supported. It enables the user to specify the ansible version to be installed in the “master” VM that will be used to configure all the VMs of the infrastructure. For example:

option ansible_version = '2.6.20'

Each line inside the contextualize section enables to specify which configure section configure_id will be applied in the nodes of type system_id. Optionally a step number can be specified to set the execution order. For example:

system nodeA (

system nodeB (

configure conf_server (

configure conf_client (

configure launch_client (

contextualize 1200 (
   system nodeA configure conf_server step 1
   system nodeB configure conf_client step 1
   system nodeB configure launch_client step 2

This RADL specifies that the configure section conf_server will be applied to the nodeA type nodes in the first step. In parallel the the configure section conf_client will be applied to the nodeB type nodes. Finally the configure section launch_client will be applied to the nodeB type nodes. This is a tipical example of a client-server application where the client must be launched afther the server has fully configured.


Hello Cloud!

The next RADL is a simple example that launches two virtual machines in the default cloud provider with at least 512M of RAM:

system node (
   memory.size >= 512M
deploy node 2

Deploy ten Ubuntu

The next RADL deploys ten Ubuntu of 32 bits with version 12.04 at least, that can be accessed from extern networks and with DNS names node-0, node-1, …, node-9:

network net (outbound = 'yes')

system node_ubuntu (
   cpu.arch = 'i686' and
   memory.size >= 512M and
   net_interface.0.connection = 'net' and
   net_interface.0.dns_name = 'node-#N#' and = 'linux' and
   disk.0.os.flavour = 'ubuntu' and
   disk.0.os.version >= '12.04' and
   disk.0.applications contains (name='toncat')

deploy node_ubuntu 10

Including a recipe from another

The next RADL defines two recipes and one of them (add_user1) is called by the other (add_torque):

configure add_user1 (
  - tasks:
    - user: name=user1   password=1234

configure add_torque (
  - tasks:
    - include: add_user1.yml
    - yum: pkg=${item} state=installed
      - torque-client
      - torque-server

Using Cloud-Init contextualization

The next RADL deploys a single node that will be configured using Cloud-Init instead of Ansible:

network privada ()

system node (
   cpu.count>=1 and

configure node (
    - [ wget, "", -O, /tmp/index.html ]

deploy node 1

contextualize (
   system node configure node with cloud_init

It depends on the Cloud provider to process correctly the cloud-init recipes of the configure section. More information about Cloud-Init in Cloud-Init documentation).

JSON Version

There is a JSON version of the RADL language. It has the same semantics that the original RADL but using JSON syntax to describe the objects. This is a complete example of the JSON format:

    "class": "ansible",
    "id": "ansible_jost",
    "credentials.username": "user",
    "credentials.password": "pass",
    "host": "server"
    "class": "network",
    "id": "publica",
    "outbound": "yes"
    "class": "system",
    "cpu.arch": "x86_64",
    "cpu.count_min": 1,
    "": "linux",
    "id": "front",
    "memory.size_min": 536870912,
    "net_interface.0.connection": "publica"
    "class": "configure",
    "id": "front",
    "recipes": "\\n---\\n- roles:\\n- { role: 'micafer.hadoop', hadoop_master: 'hadoopmaster', hadoop_type_of_node: 'master' }"
    "class": "deploy",
    "system": "front",
    "vm_number": 1,
    "cloud": "cloud_id"
    "class": "contextualize",
    "items": [
        "configure": "front",
        "system": "front",
        "ctxt_tool": "Ansible"

The RADL JSON document is described as a list of objects. Each main object has a field named class that described the type of RADL object (ansible, network, system, configure, contextualize or deploy). In case of ansible, network, system and configure, the must also have and id field. Then the other fields correspond to the features described in the RADL object. A particularity of the JSON format is that it does not uses the comparators (<= or >=) so it is expressed using the _min and _max suffixes as show in the example in cpu.count_min and memory.size_min. Also the JSON format does not use units in the amount of memory or disk size, so all these quantities are expresed in bytes.

Currently this format is only supported in the REST API (not in the native XML-RPC one).