5. Introduction to virtualization features
1. Cluster characteristics
1.1HA
HA (Hith Available) overcomes the limitations of a single host. When a server is damaged, the virtual machines running on the damaged server will automatically migrate to other servers that are running normally, and the user will not be aware of the entire migration process.
1.2 Load balancing
When there are many virtual machines running on a certain server, the virtualization software will automatically find that a certain server has a heavy load, and will migrate some virtual machines running on this server to other servers with lower load, so that all Server load reaches a balanced level; improve user experience and hardware life.
1.3 Easy to expand
The virtualization cluster will form a virtualized hardware resource, and the resource pool will shield the underlying differences. The resources in the virtual resource pool will be allocated to the virtual machine, a small part of the resources will be released after the virtual machine is used, and most of the resources will be occupied by the virtual machine. As a result, the resources in the virtual resource pool become less and less. You only need to expand the hardware resources, and then convert the hardware resources into resources in the new virtual resource pool. No cutover or capacity expansion is required. You only need to add hardware resources to the cluster without affecting the running of the program.
1.4 Memory multiplexing
Make the sum of virtual memory of all virtual machines in the server exceed the physical memory capacity of the server, improve hardware utilization and reduce costs.
a. Memory sharing
Different virtual machines share the same physical memory, and the Hypervisor will map a common memory space to multiple virtual machines at the same time.
b. Memory bubbles
A part of the memory of the virtual machine with low memory utilization is allocated to the virtual machine with high memory utilization.
c. Memory replacement
The part of the virtual machine is infrequently used, and the system allocates the infrequently used data to the storage, and redistributes the data from the storage to the virtual machine if it needs to be used.
2. Virtual machine features
2.1 Rapid deployment
a. Template deployment
Copy a virtual machine (template) that is always shut down and configure it to add new personalized data. Using the template deployment can generate completely consistent virtual machines except for personalized data (IP address, MAC address, etc.).
b. Virtual machine replication
After the virtual machine is shut down, multiple identical virtual machines (including personalized data) can be replicated, but these virtual machines cannot be run simultaneously.
2.2 Resource Hot Add
When resources are insufficient, administrators can directly add resources.
2.3Console control
Virtual machines can be operated through Console (console).
2.4 Snapshots
Use the snapshot to save the state of the virtual machine at a certain moment, and you can restore the state of the virtual machine at any time when the snapshot was taken.
a. Create a snapshot
After the snapshot is created, a new mapping will be generated, and the user can read and write to the new mapping. The original data will be sealed, and the user can only read the saved data.
b. Delete the snapshot
After deletion, the newly generated mapping will not change. The deleted snapshot is merged with the original data, and the merged data can only be read.
c. Restore the snapshot
The data does not change, and the object that the user performs read and write operations changes.
2.5 NUMA
NUMA divides a computer (physical machine or virtual machine) into multiple nodes, each node is called NUMA-Node, each node has one or several CPUs, and a common node controller is used inside the node. Interconnection module (Interconnet) for connection. The CPU access between the same Node is called Local Access, and the CPU access between Nodes is called Remote Access. Local Access is faster than Remote Access. The function of NUMA is to allow the CPU and memory used by a virtual machine to come from the same Node.
3. Features of Huawei virtualization products
3.1 Prerequisites for Advanced Features - Tools
Tools includes two parts: the hardware driver in the kernel state, and the vm-agent process in the user state.
The virtualization platform judges whether the Tools are running normally by detecting the status of the user mode process. Only the platform with Tools kernel mode installed can obtain the hardware information corresponding to the virtual machine, and can complete functions such as snapshot, live migration, online adjustment of virtual machine specifications, and network card QOS; the vm-agent in user mode is a virtual machine running in Process, through vm-agent, you can obtain some information inside the virtual machine or issue some instructions, such as obtaining the IP address of the virtual machine, obtaining the status of the virtual machine, soft shutdown of the virtual machine, and restart of the virtual machine.
3.2 Cluster Features
a. HA
HA has detailed how to deal with the problems of the host, storage, and virtual machines respectively.
b. Load balancing - power management
During working hours, the server, the virtual machine on the server, and the applications in the virtual machine are guaranteed to run normally, and the power of the equipment is sufficient to ensure the running speed of the application; during idle time, the system automatically migrates the virtual machines on some servers to several servers according to the load condition power off the servers without running virtual machines to save power.
c. Load balancing - DRS rules
① Gather virtual machines
When a virtual machine on a physical host is migrated to another physical host, another virtual machine on the same physical machine is also migrated to the same physical machine. Ensure that some virtual machines must run on the same physical machine at the same time; for example, some related applications run faster on the same server.
②Mutually exclusive virtual machine
When a virtual machine is migrated to another physical host, another virtual machine on the migrated physical host will be automatically migrated to other physical machines. Ensure that some virtual machines must not run on the same physical machine at the same time; for example, in order to ensure high availability of applications, another virtual machine will not fail after one virtual machine fails.
d.IMC
IMC can ensure that the hosts in the cluster provide the same CPU function set to the virtual machines. Even if the actual CPUs of these hosts are different, the migration of virtual machines will not fail due to incompatible CPUs. In FusionCompute, set the IMC policy of the cluster so that virtual machines can be migrated between hosts with different CPU types. Currently, the IMC policy only supports hot migration of CPUs of different models from Intel, and this function cannot be configured on CPUs from other manufacturers.
When setting the cluster IMC policy, if there are hosts or virtual machines in the cluster, the following conditions must be met:
- The CPU feature set of hosts in the cluster must be equal to or higher than the set target baseline feature set.
- The CPU feature set of the running or hibernating virtual machines in the cluster must be equal to or lower than the target baseline feature set. If there is a virtual machine that does not meet the conditions, you need to shut down the virtual machine or migrate it out of the cluster.
3.3 Virtual Machine Features
a.CPU&Memory Qos
Shares: Allocate CPUs to each virtual machine in proportion.
Reservation: Determine the lower limit of the CPU and memory occupied by the virtual machine. No matter what happens, the system must reserve enough resources.
Limit: Determine the upper limit of the CPU and memory occupied by the virtual machine, and limit the maximum amount of resources that the virtual machine can occupy.