Host

A host is a machine that can execute tasks. A host consist of the following components:

variable	type	required?	default	description
name	string	no	Host	The name of the host. This is only important for debugging and post-processing
count	integer	no	1	The amount of hosts of this type are in the cluster
cpuModel	CPU	yes	N/A	The CPUs in the host
memory	Memory	yes	N/A	The memory used by the host
cpuPowerModel	Power Model	no	Default	The power model used to determine the power draw of the cpu
gpuPowerModel	Power Model	no	Default	The power model used to determine the power draw of the gpu
cpuDistributionPolicy	Distribution Policy	no	Default	The distribution policy used for the CPU
gpuDistributionPolicy	Distribution Policy	no	Default	The distribution policy used for the GPU

CPU

variable	type	Unit	required?	default	description
modelName	string	N/A	no	unknown	The name of the CPU.
vendor	string	N/A	no	unknown	The vendor of the CPU
arch	string	N/A	no	unknown	the micro-architecture of the CPU
count	integer	N/A	no	1	The number of CPUs of this type used by the host
coreCount	integer	count	yes	N/A	The number of cores in the CPU
coreSpeed	Double	Mhz	yes	N/A	The speed of each core in Mhz

Memory

variable	type	Unit	required?	default	description
modelName	string	N/A	no	unknown	The name of the CPU.
vendor	string	N/A	no	unknown	The vendor of the CPU
arch	string	N/A	no	unknown	the micro-architecture of the CPU
memorySize	integer	Byte	yes	N/A	The number of cores in the CPU
memorySpeed	Double	Mhz	no	-1	The speed of each core in Mhz. PLACEHOLDER: this currently does nothing.

GPU

GPUS are an optional component of a host. The required fields are only required if the host has GPUs.

variable	type	Unit	required?	default	description
modelName	string	N/A	no	unknown	The name of the GPU.
vendor	string	N/A	no	unknown	The vendor of the GPU
arch	string	N/A	no	unknown	the micro-architecture of the GPU
count	integer	N/A	no	1	The number of GPUs of this type used by the host
coreCount	integer	count	yes	N/A	The number of cores in the GPU
coreSpeed	Double	Mhz	yes	N/A	The speed of each core in Mhz
memorySize	integer	Byte	no	N/A	The speed of each core in Mhz
memoryBandwidth	Double	Bytes per Second	no	N/A	The speed of each core in Mhz
virtualizationOverHeadModel	VirtualizationOverHeadModel	N/A	no	N/A	The virtualization model of the GPU, used to determine the performance

Example - No GPU

{
    "name": "H01",
    "cpu": {
        "coreCount": 16,
        "coreSpeed": 2100
    },
    "memory": {
        "memorySize": 100000
    },
    "CpuPowerModel": {
        "modelType": "sqrt",
        "idlePower": 32.0,
        "maxPower": 180.0
    },
    "count": 100
}

This example creates 100 hosts with 16 cores and 2.1 Ghz CPU speed, and 100 GB of memory. The power model used is a square root model with a power of 400 W, idle power of 32 W, and max power of 180 W. For more information on the power model, see Power Model.

Example - With GPU

{
    "clusters": [
        {
            "name": "C01",
            "hosts": [
                {
                    "name": "DualGpuHost",
                    "cpu": {
                        "coreCount": 4,
                        "coreSpeed": 2000
                    },
                    "memory": {
                        "memorySize": 140457600000
                    },
                    "cpuPowerModel": {
                        "modelType": "linear",
                        "power": 400.0,
                        "idlePower": 100.0,
                        "maxPower": 200.0
                    },
                    "cpuDistributionPolicy": {
                        "type": "MAX_MIN_FAIRNESS"
                    },
                    "gpu": {
                        "coreCount": 2,
                        "coreSpeed": 2000,
                        "virtualizationOverHeadModel": {
                            "type": "CONSTANT",
                            "percentageOverhead": 0.25
                        }
                    },
                    "gpuPowerModel": {
                        "modelType": "linear",
                        "power": 400.0,
                        "idlePower": 100.0,
                        "maxPower": 200.0
                    },
                    "gpuDistributionPolicy": {
                        "type": "FIXED_SHARE",
                        "shareRatio": 0.5
                    }
                }
            ]
        }
    ]
}

This example creates a host with 4 CPU cores and 2 GPU cores, both with a speed of 2.0 Ghz. The host has 140 GB of memory and uses a linear power model for both CPU and GPU with a power of 400 W, idle power of 100 W, and max power of 200 W. For more information on the power model, see Power Model.