Shulou(Shulou.com)06/02 Report--

What is Ceph? It is a software-defined open source distributed object storage solution for PB-level massive data storage platform. Originally developed by Inktank in 2012, the company was acquired by Red Hat in 2014. With the development of big data in recent years, Ceph has received a lot of attention in the field of distributed storage because of its excellent performance in performance, reliability and scalability.

As an enterprise-class open source platform, Ceph can provide unified software-defined storage on standard economical servers and disks, combining the storage of data blocks, objects and files into one platform, and is a reliable, automatic rebalancing and automatic recovery distributed storage system.

In Ceph, data is stored or processed as objects, unlike traditional data storage solutions that typically write and read data to storage volumes through sectors and sector offsets (often referred to as blocks). When dealing with large-scale data, it is a good way to treat the data as objects, and it is easier to manage. This kind of object storage enables Ceph to achieve simplified scalability, thus easily meeting the needs of users. These objects are copied to the entire node cluster, making Ceph fault-tolerant and further reducing single point of failure.

Last year, the Ceph community released support for erasure Code Pool (Erasure Coded pool), which enables storage with less space, but has some negative impact on performance. In addition, the design of Ceph also includes self-correction and self-management functions.

In terms of accessibility, Ceph exposes three interfaces to user space. The first is object storage. This object store is accessible through the RESTful interface and supports OpenStack Swift and Amazon Simple Storage Service (S3). In this way, Web applications can send direct PUT, GET, and DELETE methods to the object store without rewriting the application code or worrying about where the object is stored.

The second interface is a sparsely allocated block device. The goal is to allow Ceph to access existing computing environments directly, without rearchitecting applications and virtual environments for file / block volumes, but still be able to take advantage of most of the features, functionality, and resilience provided by Ceph. The advantage of Ceph's object-based model is that block device and file system interfaces are well used for snapshots, clones, and better load balancing support.

The third interface is the file system. Although the file system provides a lot of accessibility and functional block devices anyway, in the Ceph implementation, the built-in file system does remove the block device layer (reducing the number of total stacking layers) and can connect directly to the object storage back end. This does simplify maintenance and debugging.

Ceph architecture diagram

Ceph is completely managed by the command line. Red Hat has rereleased Ceph,Calamari through a Web-based user management interface called Calamari to simplify general Ceph management. It comes with a server and client component that provides a Web-based dashboard. It communicates directly with the server through RESTful API.

Although Ceph itself solves many problems in the industry, especially how to manage and scale data, data management challenges are more than these. Ceph is designed to handle two things: 1, it supports fault tolerance by distributing data (replication or erasure coding) across node clusters; and 2, it provides users with access to the same data. What happens above and below is entirely up to the storage administrator. For example, how is the hardware monitored under the Ceph framework? How to detect and correct drive failures? How do I export blocks and file system volumes on top of the framework? How do I enable high availability for these same volumes?

This is the value of software redistributors. Vendors such as Red Hat, SUSE, Canonical (Ubuntu), etc., combine all of these parts and unify them into one management space. In order to increase the credibility of products, many large companies in the data storage industry have joined Ceph, including SanDisk, SolidFire (now part of NetApp), and so on, all of which use Ceph in some form.

It is not difficult to see that Ceph is likely to have a very bright future. Of course, like any other data storage solution, it cannot meet all data storage needs, but since it has emerged and has received a lot of recognition, it shows that it is very competitive.

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.