It may seem that by setting weights for all vCPUs we will succee in limiting the CPU consumption of the virtual machine. However this trick will not work as expecte. In fact it will only work if you specify the correct cpu shares value for the entire virtual machine. In the case of OpenStack work nees to be done to scale the value of cpu shares. I agree the code eiting is small but this solution will nee to be thoroughly teste. Conclusion A very cool solution that is suitable for managing a large number of processes or virtual machines. Well scalable and automate.
A Private Cloud Is More Reliable Due
We limit on the knee There is an even simpler solution without the use of additional software. Principle of operation In the simplest version you can try to run more virtual cores than iron ones on the host. So if there are six CPUs on the host then by running six virtual machines with vCPUs we can assume that each vCPU will have only . CPUs available. This Bermuda Email List approach lacks the flexibility of setting different limits for different virtual machines. But this is better than a hard limit on consumption from above as is the case with cpu limit. This method scales well by adding hosts. Virtual machines will not be limite in principle and will only experience steal time when everyone is in a sudden nee of CPU time. pros Does not require additional development.
To Efficient Load Balancing Between Physical
Easy to understand. In the same way as with cgroups the virtual machine will have the opportunity to take more CPU time than state if there is a free resource. Minuses You nee to keep track of how many virtual machines are running on the host and not BM Leads excee the specifie CPU vCPU ratio. If the ratio is violate we risk getting virtual machines with a lower guarantee percentage than state. Scales only by adding new hosts. Since virtual machines are essentially limite only by the parameters of the hardware on which they are running. If a virtual machine limite via cgroups can be run anywhere then in the current solution only on special hosts. Conclusion This way of organizing interest bearing instances is convenient because it is very easy to implement and understand.