Cloud storage is closely related to the cluster NAS architecture

Today's cloud computing boom is overwhelming, and in terms of storage, the demand for storage from cloud services can often be divided into two areas. On the one hand, performance is a key requirement, especially for cloud computing architectures. On the other hand, cost effective storage is also critical, which is often required by the cloud storage architecture. Both aspects require the ability to extend support for thousands of users and petabytes of storage. Cluster NAS architectures and clouds have almost become synonymous with clusters that can meet this expansion requirement. There are two kinds of clustering methods, tight binding and loosely binding, and both of these methods play different roles in cloud storage.

Loosely-coupled clusters can often meet the needs of the environment that is dominated by cloud storage. In these clusters, each node is a separate entity, many volumes are assigned to nodes, and other nodes in the cluster cannot access the volumes. Here, the smallest unit is the file, which belongs to the node. When a file is stored in a cluster, its data is completely stored in a particular node in the cluster. Although multiple copies of this file are relocated to other nodes to provide redundancy, multiple nodes cannot provide a single instance of the file. This is acceptable to many cloud storage applications.

The smallest unit of a tight-knit cluster is a block of data. As files are saved to a cluster, they are divided into chunks of data that are accessible to any other node in the cluster. When a file request is made, each node accesses different blocks of the file to handle the application or the user's request. The more available nodes in response to a request, the higher the performance.

In both of these clustering methods, the surface of the application or the user is facing is an entity. They do not need to identify multiple nodes below multiple points. However, for storage administrators, loose-knit clusters have some attachment because they need to manage distributed data and ensure that access to different nodes is a relatively balanced strategy. There is only one access point in a tight-knit cluster because all nodes that participate in file access are peers.

The performance of the cloud is often overlooked by people. After all, people access data through a relatively slow internet connection. The processor data-especially in cloud storage-is accomplished by accessing users or applications. Although there may be a performance requirement for thousands of users to access data, the performance of individual nodes is generally acceptable. Storage performance is important in both cases for the cloud infrastructure.

First, cloud computing or the application is the service environment. In the cloud computing environment, thousands of users often use an application Web front-end. All requests are received and processed within the cloud computing provider's data center. This is roughly equivalent to thousands of internal user access and request applications.

The second example is the data center that wants to deploy scalability within the cloud storage architecture. As the example mentioned above, potentially thousands of users simultaneously access storage clusters, all internally, and storage performance becomes important if high-performance NAS services are required.

The challenge is that, when providing a non-traditional environment, a computing provider cannot use traditional methods to manage storage performance and scalability. In the traditional approach, all components are prepaid or front-end purchases, and then the next five years capital expenditure is gradually devalued.

The cloud provides a pay-as-you-go pattern. Most cloud computing environments are not starting with thousands of users who have already prepaid a service. Users usually pay monthly, quarterly, or even yearly fees. As a result, cloud providers cannot afford to lease a large front-end storage infrastructure and expect users to order and pay immediately. They are also aware of another problem, that their applications or services are so successful that they quickly exceed the initial purchase capacity and then need to upgrade to a faster and more expensive system.

Cloud providers need a pattern that matches their business model and expands on services. The cloud storage market has started to quickly adopt loosely-coupled storage clusters-through in-house-developed technology or off-the-shelf software-to gain this flexibility. In contrast, cloud computing and application service providers migrate to tightly bonded cloud storage clusters to provide scalability in performance and capacity.

First, in most tight-knit clusters, all storage-system components are provided by one vendor. In this architecture, communication between nodes requires a high similarity of hardware. Also, tight-knit clusters tend to manage multiple levels of storage in a cluster. Some clusters can only handle one type of storage (either all fibre Channel drives, either SATA drives or all solid-state drives).

Symantec's Filestore products are beginning to combine the flexibility of loosely-coupled clusters with the performance and management simplification of tightly coupled clusters. These solutions create a hybrid storage cluster by using software that can be loaded into off-the-shelf Intel servers, providing front-end storage processing capabilities for the shared storage backend. This storage is provided by several vendors and is divided into multiple storage pools. The data is then automatically migrated between storage pools based on file properties.

When it comes to backing up large repositories, many loosely-clustered clusters can be protected through NDMP. Some clusters also add support for data protection software. Because many clusters run Linux operating systems, it may require a secure Linux backup agent, which is more in the form of authentication rather than integration. Symantec has done more in this area, with the exception of standard NDMP support, Symantec provides tight integration with NetBackup Enterprise backup solutions. This can dramatically improve performance and further simplify data protection operations.

This provides the flexibility to use common Intel hardware for the cloud computing environment, but retains the performance characteristics of a tightly knit cluster. This is an ideal choice for data centers that want to leverage private cloud to provide internal file storage. Now you can take advantage of the performance features of tight-knit clusters and existing storage resources to provide users with a high-performance NAS platform, all of which are based on the economic characteristics of the cloud.

Original source from "Bit net", reprint please keep the original link: http://storage.chinabyte.com/375/12162375.shtml

(Responsible editor: The good of the Legacy)