We have entered the 5g era, which will be an era of connectivity. In the previous 4G era, there were only a few devices, such as our personal computers, mobile phones and tablets, connected together. In the 5g era, there will be more and more other devices connected together, which will become an era of interoperability of everything, with explosive data growth. At the same time, the storage challenges are getting higher and higher. Therefore, in order to meet the arrival of 5g era, the importance of data and storage will be growing and improved.

From our point of view, the CPU centric era in the past has entered the data centric era from now on. The impact is not only on the current mobile phones and computers, but also on the rapid growth of data. At the same time, machine learning and driverless driving, which rely on big data, also show explosive data growth. Under such requirements, it is bound to bring about the vigorous development of our storage industry, This trend is reflected in the “super cycle” of the semiconductor industry that we have repeatedly mentioned in recent years.

In addition to the vigorous development of big data, the most popular hot spots at present are machine learning and artificial intelligence. We have learned from various channels that machine learning has surpassed human capabilities in many aspects. In this regard, Samsung’s storage has various types of devices from SSD to other types. Under this great social change, we expect to use our various storage products to meet epoch-making changes.

First of all, based on our changes and trends in data, we are now in the era of big data. Data is generated anytime, anywhere, and needs to be processed in the data center in real time. Real time analysis, real-time processing, and continuous learning will inevitably bring challenges and growth to the storage business. Therefore, our understanding of the cloud era is that more and more devices are connected to the cloud through the cloud, and then our devices connected to the cloud through the cloud are interconnected and integrated.

In this case, various devices connected by the cloud generate a large amount of data, which will inevitably lead to great changes to our existing cloud infrastructure. Therefore, in view of this development, we put forward three core elements for Samsung’s next-generation storage products. The first is low latency, high bandwidth and large capacity. Next, I will introduce the whole line of products one by one based on these three core elements.

In order to meet the growing demand in different scenarios, we have now built a comprehensive product line from z-ssd to QLC SSD from a single SSD product. In addition to SSD products, the most popular is storage level memory (SCM). We are also building our SCM products through standardization building and ecosystem building. At present, SCM products have passed standardization certification.

Z-ssd is a product developed by us for real-time data analysis and high-performance applications. Z-ssd has the characteristics of low latency and high bandwidth. We can see that compared with traditional reading, z-ssd does not use a new storage medium, but uses our mature flash technology. In addition to improving the performance of flash technology, it further improves the overall performance of z-ssd by optimizing the performance of flash controller.

Z-ssd feedback collected from the industry, such as feedback received from server providers and cloud service providers, is generally positive. This low latency feature meets the needs of many scenarios. In addition, we can see from foreign technical media that the performance of our z-ssd with low latency is close to that of competitors. They generally welcome the emergence of our innovative products.

Products with high bandwidth and high capacity. Taking PCIe gen4 as an example, the pm1733 supports 32tb at the maximum capacity, and uses 32tb to replace the servers with two cabinets of 10000 RPM hard disks. It can be seen that the storage devices that used to need two cabinets now only need two servers to replace them. The advantage is that when used in the data center, the occupation of space and the consumption of resources are greatly saved. At the same time, our design improves the performance by at least two times, At the same time, the performance of 4U servers has doubled compared with 2U servers.

QLC is mainly aimed at some applications with reading as the main application scenario, such as archived videos, emails, notes, including photos and videos of social video websites. Such scenarios are very suitable for QLC, mainly for reading and writing applications. QLC is mainly used to replace 10000 RPM enterprise disks. Compared with 10000 RPM disks, the main advantage of our QLC SSD is that in the whole life cycle, the TCO product is equivalent to 1/3 of 10000 RPM disks, which provides a very competitive choice from the perspective of cost. We expect QLC SSD to be rapidly expanded in the 10000 RPM disk market.

We use NF1 SSD to build a system. For example, if we use 16TB NF1 SSD, 1U servers can reach 576tb, and 2U servers can be expanded to 1.15pb. Of course, because it supports upward expansion, the performance of 2U is twice as fast as 1U, reaching 20million IOPs. It is easy to build PB level systems.

With the rise of the Internet, our requirements for traditional databases need to be highly concurrent and easy to expand. In this case, we know that the efficiency of traditional databases is getting worse and worse. In this case, more and more manufacturers turn their attention to kV (key value) databases. At the same time, we also see some problems faced by deploying kV databases on traditional storage.

We found that we need to go through the file system, block devices, and then transfer them to the SSD. We also need to go through the mapping of logical addresses and physical addresses. This layer by layer data mapping brings about the consumption of system resources. With the increasing speed of SSD, this consumption has become the bottleneck of the system more and more. In view of this situation, Samsung has launched a new SSD – kV SSD. Our approach is to build part of the kV engine into the kV SSD, At the same time, some kV interfaces are provided, so that you can directly enjoy the kV request of the application. This greatly reduces the data call process, improves the speed of data access, and saves resources. Of course, although this has greatly improved the system performance, you can see that there are great changes to the traditional storage architecture. We need to build a new software ecosystem with new products. This part may be expanded in detail later.

The next step is to show the revolutionary changes of our kV SSD with examples. First of all, we chose the domestic mainstream kV storage engine. The test and comparison results show that the delay is reduced by more than 7.6 times, and the IOPs is increased by 7.7 times. The change is very huge. At the same time, the domestic mainstream distributed storage is selected. It can also be seen that the bandwidth is increased by 4.8 times, and the IOPs is increased by at least 2 times. Therefore, these works are being carried out closely with customers. I hope that those who have more information will have more detailed introductions.

In particular, openmpdk, a software component developed by Xi’an team, can be found on our home page. If you want to know about it and join us, you can visit the home page. You are welcome to participate and jointly promote the revolutionary change of this storage structure. Finally, with regard to the design of the current database, we emphasize diversity and flexible configuration. Samsung currently holds an open attitude and welcomes various user manufacturers to cooperate to develop targeted SSD optimization and solutions for their specific scenarios. This is our expectation to build a storage ecosystem in China.

Leave a Reply

Your email address will not be published. Required fields are marked *