Datacenter migration object storage evolution.

Datacenter migration，The amount of data created every year is increasing at an exponential rate, and almost every enterprise now relies on technology, and the information generated by these businesses can be said to have become their largest asset. Unstructured data is the best storage type with the largest growth rate. So how do we use object storage technology and what can we expect from it?

What is object storage?
The development of object storage is derived from the highly flexible demand for a large number of structured data stored for a long time. Review the past 20 granny, we have block storage (traditional storage) and NAS devices (usually as a file server). NAS didn't really extend to PB level, nor did it provide flexibility for long-term data retention. Generally speaking, enterprises use tape (Tape) to meet demand, but tapes are slow and inefficient.
The developed object store has filled the gap by providing online access content, and has developed into a mature technology over the years. Using the new protection methods such as erasure coding (EC) and other new protection methods, we usually solve the problem of data protection in large archives.
Object storage uses Web based protocols to store and retrieve data, and in essence, most of them provide four primitive CRUD: create (Create), read (Read), update (Update), delete (Delete). In many cases, Update is just to match Delete and Create. This means that using HTTP embedded data and associated metadata to send REST based API calls to interact with object store is relatively simple.
This simple operation highlights one problem in object storage: the application needs to be rewritten to use the object storage API. vendor to help complete this process by providing SDK, but needs to change the application. This problem shows us the first object of object storage evolution: multi protocol access.
Multi protocol object storage has been implementing multi protocol access for a long time. In the form of gateway or additional software, object is used to store the back end as a large capacity pool. The problem brought by this move is whether they really provide concurrent access to the same data from different protocol stacks. Using NFS to store and retrieve objects can be done, but how about using NFS storage and using Web based protocol access?
Why do companies want to have the ability to store one protocol and access it through another protocol? Providing NFS means that applications can use object storage without modification. By providing concurrent Web based models, analysis tools can access data without introducing performance issues related to NFS protocols, such as lock or multithreading impact on the same object. The typical value read configuration file of the analysis software means that the data can be analyzed without affecting the main application.
Many IOT devices can only be stored through NFS, so it is very important to store such contents into object storage.
Scalability is one of the factors that influence the storage application of objects, which is the ability of contraction, not just expansion. The initial capacity of many object storage solutions is hundreds of TB, which is not practical for smaller IT organizations, and vendors are starting to solve this problem by producing products that can shrink to tens of megabytes.
Obviously, large capacity hard disk and flash memory may become a problem, but it can achieve functional advantages brought by object storage. Therefore, the solution provided by the vendor is pure software, can be deployed on dedicated hardware, and can also be deployed as an internal deployment or virtual instance in a public cloud.
With the continuous development of the Internet of things, there will be an explosive growth of data. A large number of small object storage will prove to be able to meet the advantages of the continuous demand for the Internet of things.
Software definition goes back to pure software solutions, providing pure software solutions means that enterprises can choose the right type of hardware for their environment. In the presence of the hardware supply contract, the enterprise can simply pay the object storage software and deploy it on the existing device, which includes testing the old hardware that may be processed in other ways.
Another development direction of open source object storage software is open source. Ceph is one of the storage platforms developed as an open source mode.  OpenIO provides advanced functions like no server, which can provide the same experience and charge for quality service. Another open source solution, Minio, has recently received 20 million yuan of funds to push its platform to a wider audience, including Docker container.
The attention of the trial version of the software means that the organization is easy to try object storage, in addition to IBM Cloud Storage and DDN, almost all vendors provide some kind of trial process by downloading software or using the company's laboratory environment. The trial version can provide openness for software, so that it can be evaluated and adopted in long term.
In the future, how to consider the future of object storage can be said that the recent development direction is to make solutions easier to use, more emphasis on the usability and installation of software and vendors. Multiple protocols connect more applications, making data more accessible to object storage. In the next few years, we will see that object storage is still an important platform for persistent data storage.