How to Test RAM
The performance of RAM is determined by the number of read and write operations performed per second, as well as the amount of data transferred per second.
To test RAM performance, the sysbench utility will be used.
-
If you have not done so already, contact technical support or your VK Cloud account manager with a request to allocate the required resources for testing.
Wait for confirmation that the resources you need have been allocated.
-
Make sure that SDN Sprut is connected for your project. Otherwise, contact technical support with a connection request.
-
Create a test virtual machine with the following parameters:
-
Virtual machine category: select a category depending on the type of resource requested, or
All current virtual machine types. -
Virtual machine type: select
STD-3 (Ice Lake) 4 vcpu, 16 Gb RAMfrom the list, or the type specified when requesting the resource. -
Disk type:
SSDis selected by default. To get maximum performance values, select the High-IOPS or Low Latency NVMe type. -
Disk size:
256 GBis selected by default. To get maximum performance values, it is recommended to select1500 GB, provided thatHigh-IOPSorLow Latency NVMEis selected in the Disk type field. For more information, see the section on disk performance. -
Operating system: select
Ubuntu 22.04 LTS. -
At the Configuration step, in the Network field, select the existing SDN Sprut or create a new network with the Sprut type.
Leave the remaining parameters at your discretion.
-
-
Connect to the created VM and install all available updates. Installation methods depend on the selected OS.
-
Connect to the VM.
-
Install the sysbench utility on the VM:
sudo apt install sysbench -
Start executing sequential read and write operations:
sysbench --test=memory run
Save the command output to a file. The most indicative parameters are:
Operations per second— the number of read and write operations performed per second;MiB transferred— the amount of data transferred per second.
You can now compare this data with the performance of other cloud platforms.
The created resources are billed. If you no longer need them, delete them: