Shulou(Shulou.com)06/03 Report--

General Electric GE

General Electric GE, founded in 1892, is the world's largest multinational company in technology and services. Since Thomas Edison founded GE, it has operations in more than 100 countries and employs 315000 people.

GE has a rich product line and operation experience in aviation, power, transportation, energy and other industries. At the same time, GE also helps customers to carry out product operation and maintenance, data analysis and improvement through digital means. For this reason, GE has established its own digital platform of the Internet of things.

GE uses a micro-service architecture and runs stateful applications through containers. To this end, the need to establish CSI (Container Storage Interface), GE tried some methods, but from the use is not mature, can not provide users and partners of the platform with easy-to-operate container storage and data management methods. There needs to be a basic layer that can run both stateless and stateful applications.

In the end, GE chose to work with Portworx to establish CSI to manage container storage and data effectively.

Portworx establishes a software-defined storage abstraction layer on top of physical storage, and users do not need to think about which physical storage they are using, nor do they need to know which public or private cloud is hosting the application.

Before there is such an abstraction layer, users need to manually allocate physical storage volumes to a container. Traditional storage is driven by virtual machines and operating systems, but it is not suitable for containers. Because containers are usually choreographed by Orchestrator to run in a multi-node environment. Not all applications run in a single container. Cassandra, for example, is usually deployed on a series of containers. A Cassandra cluster may have 3, 10, 15 Cassandra containers, deployed on 15 different virtual machines, or even in different physical data centers. So when we try to add a volume to such a distributed system, there will be a lot of problems. These problems require operation and maintenance engineers to spend a lot of time adjusting to map volumes to such distributed systems. Let's say a 5-node Cassandra cluster, which virtual machines are these nodes running on? Which storage is it on? So we have to correspond the application to the virtual machine because we are using the storage resources corresponding to the virtual machine. If the virtual machine goes down, we have to find the corresponding storage manually and match it to the new virtual machine. This is not in line with the cloud native thinking and the positioning of the container scheduler Orchestrator. At the same time, new questions will arise, how to set a password for storage in such a distributed system? How to take a snapshot? These questions will be left to our users, which is even more problematic.

As GE, we don't want to leave such complex infrastructure crawling to users.

Portworx itself is a container-based super-fusion architecture, which organically combines computing resources with storage. At the same time, Portworx integrates seamlessly with scheduler, the scheduling software of K8S.

Stateful containerized applications such as databases require persistent data on distributed nodes. Portworx uses stateful Stateful Fabric to manage data volumes, or container-SLA-aware, to do this. Replicate volume data to ensure its state while meeting the performance and availability of containerized applications. More importantly, Portworx can manage its snapshots, clone replicas, and replication operations at each container level, allowing DevOps to manage microservices separately, rather than the binding groups used as traditional storage systems like LUNs. Using Portworx to manage stateful container Stateful Containers is convenient, and the availability and management of data at each container level is simple and highly automated.

If you need to deploy a Cassandra cluster, but do not want all the nodes on the same ring network, the same Availability Zone or Failure domain,Portworx can help users better build these distributed applications. In addition, we usually want to have more than 80% utilization of physical resources. We need to make different applications coexist in the same hardware without IO conflicts. Portworx does not directly connect storage or physical LUNs to applications. We provide a virtual storage volume layer to avoid IO conflicts and implement container encryption or snapshots. Especially when a container goes down and then recovers from another location, we can quickly find the original storage and restore it in the new container. At the same time, Portworx's security and compliance management features help GE meet the company's internal security compliance requirements for the platform architecture, as well as the storage encryption and snapshot needs of users and customers.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.