Comparison and comments on military storage products of various clusters

This article mainly introduces the related concepts of cluster storage and the characteristics of Scale-out. The focus of military storage is to compare and comment on the main cluster storage products on the market.

Main classification method

The main classification methods of storage products at present:

First, both manufacturers and users are accustomed to dividing storage systems into high-end, mid-range, low-end, or entry-level products.

Secondly, it is divided according to data types, namely NAS, SAN, and unified storage.

[Two directions for unified storage: NetApp essentially uses a NAS server to virtualize a block and present it to the host. Through this "fake SAN" method, the file service and block service are jointly embedded in a storage controller. The other is to use virtual machine technology (such as VMware) to run block-level software and NAS server software in the controller at the same time to provide SAN/NAS and iSCSI services in a pair of storage controllers. 】

According to the expansion type, that is, Scale-Up and Scale-Out.

There is also a classification method, main storage, near-line storage and backup disk library. We will not get involved in the latter two products at present.

Our main focus is on clustered NAS products with Scale-Out features.

Note: YeeStor strictly defines a cluster file system. It not only supports NAS access (NFS/CIFS), YeeStor client software provides standard POSIX interfaces, but also supports multiple access protocols such as ISCSI, HTTP, and FTP.

What is clustered NAS?

A cluster is a loosely coupled collection of computing nodes composed of multiple nodes, which cooperate to provide external services. Clustered NAS refers to a NAS (NFS/CIFS) service that cooperates with multiple nodes to provide high performance, high availability, or high load balancing.

Unstructured data is currently showing a rapid growth trend. IDC research report analysis pointed out that by 2012, unstructured data will account for more than 80% of the total data storage. Clustered NAS is a scale-out storage architecture with the advantages of linear expansion of capacity and performance, and has been recognized by the global market. From EMC's acquisition of Isilon, HP's IBRIX, DELL's acquisition of Exanet, Compellent, etc., as well as IBM's SONAS and NetApp's release of Data ONTAP 8, it can be seen that clustered NAS has become one of the mainstream storage technologies.

In China, we also see clustered NAS solutions such as UIT UFS, LoongStore, Kyushu Chuzhi CZSS, and Midison YFS. Clustered NAS has a huge potential market in the future, and will gradually be widely used in high-performance computing HPC, broadcasting and television IPTV, video surveillance, cloud storage and other industries.

High performance, high capacity, and scalability are the most popular features for the development of cluster storage.

Traditional NAS expansion can easily bring about the following problems:

1. System management. Even though NAS system management is much easier than SAN storage management, it still takes time and resources to manage.

2. Manage client and application access to data. The prerequisite for access is that the NAS system must be mounted on the corresponding server or workstation. Mounting will interrupt the access of the application, so you need to reserve the downtime of the application on the server when mounting. The more NAS systems are mounted, the more downtime is brought.

3. File location. The strategy for defining file storage locations should be based on performance, accessibility, file creation time, access frequency, storage cost, availability, data protection and many other factors. The policy setting itself is not too difficult, but in fact, moving files to a suitable NAS system is a time-consuming manual data migration process. Moreover, as the NAS system becomes larger and larger, its complexity becomes higher and higher.

The Scale Out NAS provides a single interface and interface by distributing data to the storage of each node in the cluster, so that users can conveniently use and manage all data in a unified manner. All disk devices in the cluster are integrated into a single shared storage pool and provided to the front-end application server, which facilitates access and improves disk utilization.