Shulou(Shulou.com)06/01 Report--

Historical experience has repeatedly shown that changes driven by technological progress often take place at a speed and intensity beyond people's imagination.

SDS Software-defined Storage as an emerging technology, even if you do all you can imagine, it is difficult to predict how fast it develops and how far-reaching its impact is. In the face of new changes and trends, while chasing changes, we also need to calmly think about the problems that may be brought. Is the application of SDS in the data center sure to improve performance, reduce costs, and control risks? What kind of problems does SDS hide?

Problem 1. Unable to make full use of the existing infrastructure

The data center is very busy, and the IT department has to accept requirements and data tasks from many different applications at the same time and respond and process them in a timely manner. In addition, the day-to-day work of IT professionals requires maintaining a highly available infrastructure and embedding many workflows.

Most SDS solutions want the data center to be "fully committed", replacing existing storage hardware, infrastructure, and processes in the data center with the infrastructure and processes that come with the SDS solution. They often want data centers to abandon shared centralized storage hardware and support commercial super-converged architectures.

Problem 2. Failure to provide comprehensive data services

Many SDS solutions do not provide complete data service functionality. To take a simple example, many SDS solutions cannot replicate data to a remote location, so they cannot create disaster recovery copies of the data, or even migrate the data to the SDS schema itself.

Since simple data replication cannot be achieved, some data migration feature should be provided to migrate data from the old storage platform to the new platform, otherwise the data must be restored by backing up the copy.

Many two-copy and three-copy solutions are also in the same data center or in the same POD, and there is almost no disaster recovery across data centers. (there is also storage double work and arbitration, which is mentioned here, not advertising. VSAN supports storage duplex and introduces WITNESS components. The goal is to act as a tie-breaking component when usability decisions must be made (in the Virtual SAN cluster) to avoid brain-splitting behavior and meet arbitration requirements)

For SDS, the lack of complete and reliable data service capability means that customers adopting this SDS solution still need to purchase additional storage systems to fill the lack of functionality of the new SDS solution.

Problem 3. Data lifecycle management can not be done well.

With the development of enterprise business and the establishment of various platforms and systems, it also means that the amount of data to be managed is getting larger and larger, the cost of resource maintenance is getting higher and higher, and the efficiency is getting lower and lower. In the face of the requirements of different time limit, different access frequency and different importance of data, it is very necessary to reduce the cost of data storage and maintenance and to manage the life cycle of different data.

Combined with the relevant standards and specifications of the industry, the data of enterprise IT system are classified. At present, according to the business characteristics of the system, the more popular data classification includes three parts: production transaction data, service support data and system data.

According to data classification, different types of data should match different data storage strategies. The data storage strategy is to store different data on a specified storage device. At present, the main storage devices are mainly divided into online storage, near-line storage, offline archive storage, three commonly used storage.

Online storage, also known as work-level storage, storage devices and stored data are always in a state of response at any time, can be read and written in real time, and can meet the speed requirements of the computing platform for data access. General online storage devices are usually disk devices such as built-in disks and high-end disk arrays, which are relatively expensive and have the best performance.

The application of near-line storage, which is located between online storage and offline storage, has relatively low performance requirements, but provides relatively good read performance. The near-line storage strategy generally uses mid-and low-end disk array devices, supplemented by high-compression software, to meet access actions such as fast read and write.

Offline archive storage is the process of moving the data that is no longer frequently used to a separate storage device for long-term storage, and storing the data involved offline for unconventional queries. Archiving equipment in enterprises generally uses tape library, optical disc library, and large-capacity low-end disk array, the price is relatively low.

For software-defined storage SDS, the management of data life cycle has not been fully taken into account at present. Although there is also a distinction between storage areas for hot and cold data, flash drives are basically used for cache acceleration, and hierarchical storage can not be effectively achieved to meet the fine management and life cycle management of a large number of different data.

Problem 4. Lack of a reasonable pricing model

Most of the SDS is licensed under the capacity package. Many software also charge extra for advanced features. This creates a problem because most data centers have never been able to determine their appropriate capacity limit points to take full advantage of capacity packaging prices. For example, a primary capacity package can provide 1-10TB storage services, while a higher-level capacity package supports 10-25TB. What if an enterprise needs 11TB storage capacity? This means that in order to support the excess capacity of 1TB, enterprises must upgrade to the capacity license of 25TB.

Another cost that most SDS solutions ignore is the associated cost of not supporting existing infrastructure and storage systems. Many of these SDS solutions allow users to use commercial hard drives on the same server to provide computing (hyperconvergence), which reduces future costs, but requires users to purchase flash solid state drives and hard drives to install into existing server infrastructure.

The Future of Software-defined Storage

For the large data centers of traditional enterprises, the previous commercial middle and high-end storage have experienced a large number of production environments and practices, and are relatively mature and stable. However, software-defined storage has not been tested by a large number of production environments and enterprise practice verification of complex business scenarios, so it is not mature enough.

However, Software-defined Storage can maximize the advantages of faster CPU, memory, flash and disk with the latest technology through a simplified all-in-one architecture with deep integration of hardware and software. Although SDS still has many problems, it doesn't hinder its excellence. SAN living in the physical world will eventually be replaced by the super-integrated SDS architecture that plays in the virtual world. SDS super-convergence architecture is bound to become a new cornerstone of building a modern data center.

Questions:

Can Q1:SDS meet the standards of enterprise applications? Like a few 9s or something.

A: the high availability of SDS is currently not as high as traditional storage. If you do a two-node plug test, you will find that high reliability is not as good as you think.

Q2: in software-defined storage, is it better to separate computing nodes from storage nodes or to work together?

A: look at the usage scenarios, merge provincial nodes and provincial servers together, and are suitable for merging with requirements for computing power and storage capacity. Separately, it is suitable for mass storage, including some offline storage scenarios.

Q3:SDS are there any other software vendors besides VMware at present?

A: there are SMARTX,EMC 's SCALEIO, Huawei's FUSIONSTORAGE, and open source CEPH! It should be said that these products have their own advantages, and each has its own characteristics!

Q4: are software-defined storage and software-defined networking essentially the same?

A: the essence of thought is the same, it is decoupling, separating the data plane from the control plane.

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.