Ctrl+K

ElasticMachineGroup Class

ElasticMachineGroup Class#

An Elastic machine group is similar to the Machine group with the extra property that scales up and down the number of active machines based on the number of simulations in queue. It is composed of a pool of homogeneous machines that work individually and do not communicate with each other in any way.

Hence, an elastic machine group creates a private queue for which workers scale based on the number of tasks in it. This allows running multiple simulations at the same time, with the slight overhead of machines starting, with a more cost-effective strategy since machines won’t stay idle for long.

Instantiating an ElasticMachineGroup object#

To create an elastic machine group the following properties can be configured:

  • the machine_type defines the type of CPU used for each machine. This parameter follows the naming convention set by Google Cloud, e.g., c2-standard-16. This convention is composed of a prefix that defines the CPU series, a suffix that sets the number of virtual CPUs (vCPU) per machine and the middle word refers to the level of RAM per vCPU. In the example, c2 refers to an Intel Xeon Scalable processor of 2nd generation, standard means 4 GB of RAM per vCPU and will contain 16 vCPUs. Currently, this is the list of available machine types available via the API.

  • the min_machines, max_machines sets the number of minimum and maximum machines available in the computational resource. That is, the number of active machines will never go lower than the minimum and never above the maximum. During runtime, there might be a different number of active machines in between. Moreover, the `min_machines`` is the number of machines that the group is started.

  • the data_disk_gb allows the selection of the size of the disk attached to each machine that is reserved for the simulation data in GB.

  • the spot argument determines if the machines will be preemptible or standard. Preemptible machines can be stopped at any time and for that reason are only advised for fault-tolerant workloads. If simulations are running when they are stopped, the simulation is resubmitted to the queue of the machine group again.

For example, the following code creates a MachineGroup with 2 machines of type c2-standard-16 with 100 GB of disk space each:

import inductiva

elastic_machine_group = inductiva.resources.ElasticMachineGroup(
    machine_type="c2-standard-16",
    min_machines=2,
    max_machines=10,
    data_disk_gb=100,
    spot=False)

Creating an instance of ElasticMachineGroup does not start the machines. This only registers the configuration on the API which can now be used to manage it further.

Managing the ElasticMachineGroup#

With your elastic_machine_group object ready, you can launch the elastic machine group with the minimum number of machines active with elastic_machine_group.start().

Within a few minutes, the machines will be set up and ready to pick up several simulations simultaneously. As simulations get into the queue, the number of active machines increases.

At any moment, you can check an estimate of the price per hour of the group as follows:

elastic_machine_group.estimate_cloud_cost()

When you have finished you can terminate it with elastic_machine_group.terminate() or via the CLI with $ inductiva resources terminate api-agn23rtnv0qnfn03nv93nc.

Running simulations will be killed and from this point, the elastic_machine_group object cannot be re-used.

Elastic Machine Group on demand without any hassle.

Contents