Apache StratOS Cloud Partition

Cloud Partitioning

The following sections explain in detail what a cloud partition is:

Partition

Stratos partitions can be mapped to a logical partition of the IaaS. This logical partition can be one of the following levels: Provider level, region level, zone level, or rack level. A logical partition needs to define at least one provider. A division can also be specified when a logical partition is defined. To achieve high availability, devops users can define composite regions and create resource instances into different regions, such as different region, zone, or rack. For example, if you create a resource instance to a different zone that is available in EC2, the application that is deployed when there is a problem with one zone is still available.

650) this.width=650; "Src=" Http://s4.51cto.com/wyfs02/M02/86/81/wKiom1fBQ6iRmWdnAABsUsne_Ks899.png-wh_500x0-wm_3 -wmp_4-s_2320074522.png "title=" network partition. png "alt=" wkiom1fbq6irmwdnaabsusne_ks899.png-wh_50 "/>

Network partition

The network partition is also associated to a partition group. A network partition is mapped to an area of IaaS, subject to an actual network of IaaS (such as region). Each region is geographically isolated. Each region has multiple isolated locations, called multiple zones. However, some partitions can also be within a network partition. To achieve high availability, DevOps users can define multiple partition groups, which belong to multiple partitions. The partitioning algorithm for multiple partitions within the same network partition is defined in the deployment rule, while the partitioning algorithm between different network partitions is defined in the rules that are applied.

Network Partitioning algorithm

When automatic scaling occurs between multiple partitions, it is necessary to work according to a predefined partitioning algorithm. The extension mechanism defined by the partitioning algorithm is required when creating resource instances on multiple network partitions. The partitioning algorithm can be applied to a network partition, depending on the partitioning algorithm to change between multiple network partitions. For example, with two network partitions named NP1 and NP2, with two isolated partitions named P1 and P2, in the same partition group PG1, the partitioning algorithm works within PG1 and P1 of the P2 partition group, and can change to function between the network partition NP1 and NP2.

650) this.width=650; "Src=" Http://s2.51cto.com/wyfs02/M02/86/81/wKioL1fBRE6zAZWpAABdVPZ2VRc877.png-wh_500x0-wm_3 -wmp_4-s_2592384440.png "title=" network algorithm. png "alt=" wkiol1fbre6zazwpaabdvpz2vrc877.png-wh_50 "/>

There are several partitioning algorithms:

Sequential algorithms

With this algorithm, the resource instance is created on the same partition until the resource is all used, and it continues to be created on the next partition. For example, two partitions are named P1 and P2, the resource instances are created under the same partition, and the resources are created under P2 until the partition P1 resources are exhausted. This partitioning algorithm is used between multiple partitions or multiple network partitions.

Stochastic algorithm

With this algorithm, resource instances are randomly created on multiple partitions. For example, there are two partitions P1 and P2, the first time the resource instance is created on the P1 partition, and the next time the resource instance is created in the P2 partition. This partitioning algorithm is used only between different partitions.

Match all

The resource instance is created into all the network partitions used by the application. This partitioning algorithm is applied between multiple partitions or multiple network partitions.

Sample Scenario

In a scenario where a typical cloud traffic suddenly increases, you can use the sequential algorithm. For example, if the enterprise has its own private cloud, all resource instances are placed in the private cloud, and when the capacity is insufficient, the next partition points to the public cloud. Stochastic algorithms are used in scenarios where reliable high availability is required, and resource instances are distributed evenly across different iaas platforms.

Network partition definitionThe Auto Scaling service is based on the definition of the network partition. Manage the number of resource instances in Apache Stratos by defining network partitions. The network partition is defined in the network partition JSON definition file. The definition of a deployment policy can be defined at the group level or at the cartridge level. The maximum number of resources can be defined at the group level or at the cartridge level.

This article is from the "Wesley" blog, please be sure to keep this source http://szk5043.blog.51cto.com/8456440/1843270

Apache StratOS Cloud Partition