General Electric GE
General Electric GE, founded in 1892, is the world's largest technology and services to multinational companies. Since Thomas Edison created General Electric Company, operates in more than 100 countries and employs 315,000 people.
GE has a wealth of product lines and operational experience in aviation, electric power, transportation, energy and other industries. GE while also helping customers through digital approach to product operation and maintenance, data analysis and improvement. GE To this end set up their own things digital platform.
GE micro-service architecture, and runs through the container stateful applications. To this end, it is necessary to establish CSI (Container Storage Interface), GE tried a number of ways, but it is not from the use of mature, can not give customers and partners the platform, providing container storage and data management simple and easy to operate. There can be a need to run the application stateless and stateful application of the base layer.
Finally, GE chose to work with Portworx, to establish CSI containers to store and manage data effectively.
Portworx established software-defined storage abstraction above the physical storage layer, the user does not need to use that particular Reflection physical storage, carrying an application not need to know which public cloud or private cloud flowers.
在没有这样一个抽象层之前，用户需要手动的把物理存储卷来分配到某个容器上。传统的存储，都是通过虚拟机和操作系统来驱动存储的，对于容器来说则很不适用。因为容器通常被编排程序Orchestrator排程在多节点的环境下来运行。应用程序也不都是在单一的容器内运行。比如Cassandra, 通常是部署在一系列的容器上。一个Cassandra集群可能会有3个、10个、15个Cassandra容器，被部署在15个不同的虚拟机上，甚至可能在不同的物理数据中心里。所以当我们尝试把某个卷添加到这样一个分布式系统里的时候，就会出现非常多的问题。这些问题需要运维工程师花大量的时间来做调整，让卷与这样的分布式系统产生映射。假如说一个5节点的Cassandra集群，这些节点都运行在哪些虚拟机上呢？又是在哪个存储上呢？于是我们不得不把应用跟虚拟机对应起来，因为我们在使用虚拟机对应的存储资源。如果虚拟机停机了，我们就不得不去手动寻找相对应的存储，然后把它和新的虚拟机对应起来。这跟云原生的思想和容器排程器Orchestrator的定位并不对路。同时新的问题又会产生，如何在这样的分布式系统里为存储设定密码？如何做快照？这些问题都将留给我们的用户，这就更有问题了。
作为GE，我们并不想把这样的复杂的基础架构爬坑工作留给用户。
Portworx本身是一种基于容器的超融合架构，将计算资源与存储有机结合在一起。同时Portworx与K8S的调度软件scheduler无缝集成。
Such as database applications have a state type container need persistent data in the distributed nodes. Portworx use stateful Stateful Fabric to manage data volumes, namely container-SLA-aware, to do this. Copy volume data to ensure its status, while meeting performance and availability of the application container. More importantly, Portworx can manage their snapshots in each container level, the cloned copy and copy operations, DevOps can separate the micro-management services, rather than used as a binding set of LUNs that traditional storage systems. Use Portworx management stateful container Stateful Containers easy, availability and management of each container-level data is also very simple, and highly automated.
If you need to deploy a Cassandra cluster, but do not want all nodes in a ring network, in the same Availability Zone or Failure domain, Portworx can help users better to structure these distributed applications. In addition we usually hope to have the physical resources of more than 80% utilization. We need to make different applications to coexist within the same hardware, without creating a conflict of IO. Portworx not directly to the store or physical LUNs connected with the application, we provide a virtual storage volume of IO layer to avoid conflicts, and encryption containers or snapshots. Especially when a container is down, and then from another location after a recovery, we will be able to quickly find the original storage and recovery in the new container. Meanwhile Portworx security and compliance management capabilities to help GE meet internal compliance requirements for secure platform architecture, but also to meet the user's storage encryption and customers, such demand storage snapshots.