Real-time decision-making for 5G applications

Background overview

Although many suppliers in recent years have repeatedly reiterated their professional recognition of VoltDB's continuous output, VoltDB has also continued to add some popular technical vocabulary with the development of technology, but it really allows everyone to understand the characteristics of a certain technological product's continuous evolution, relying solely on Adding a few technical terms is still far from enough. Relying on the list of nouns without detailed descriptions will only aggravate the level of information confusion and make the decision-making process of technical decision-makers more complicated.

As far as VoltDB is concerned, the most prominent example is the use and misuse of "real time" in the 5G environment. Before 5G entered reality, VoltDB had been operating in the millisecond world for ten years. At present, global partners are paying attention to 5G and believe that the advent of 5G will open up a new blue ocean. At the same time, various articles on major breakthroughs in "real-time decision-making" and other similar statements are also emerging.

Two recent examples highlight this challenge:

Case 1:

MemSQL's growth mirrors need for real-time data analysis and rapid decision-making (see the original link at the end)

In the article, the author fully explained the importance of "combining real-time information and historical information in the cheapest and easiest to deploy flexible architecture", and then described how the product is currently used for COVID tracking in Thailand. Reporting COVID results and aggregation is a valuable activity, but it is not real-time or decision making.

Case 2:

How 5G and Public Clouds Will Shape the Future of Global Applications (see the original link at the end)

Case 2 is actually a re-release of the previous article, discussing "...Two technical building blocks are inevitably close to each other. The first is the continuous improvement and accessibility of the public cloud, and the second is the upcoming 5G "The author claims "These two features will release the delay level previously only applicable to Fortune 500 companies.

We are not sure about the connection between the two. The second article also discussed the 100 millisecond rule extensively, but this discussion is meaningless. Although 100 milliseconds is a delay that can be felt by human nerves, the future growth of data traffic will come from devices and computers-"devices" in the Internet of Things, not from humans.

VoltDB has realized this for a long time, and we already have a large number of client applications running in millisecond delays.

The core of 5G application lies in "problem decision-making"

The first thing that many vendors lack is the "problem decision" feature in 5G applications-making decisions in business. The transaction does not store the results of the COVID test, what you can do may produce multiple results, and until you do, you will not know what kind of results will happen.

Scalable transaction is a problem that traditional RDBMS strives to solve, but NoSQL has never really solved it. Although you can extend your "read" capabilities by having multiple copies of data, "transaction" and "write" operations are quite different.

The actual magnitude of future transactional workloads can be found in the Industrial Internet of Things, which makes the Industrial Internet of Things rapidly becoming the focus of 5G. From our extensive experience in telecommunications companies, we know that a telephone or mobile data interaction involves dozens or even hundreds of transactions behind the scenes, which results in the system processing more than 500K TPS per second. Assuming that the density of 5G equipment is 1 million units per square kilometer, we can estimate that the capacity required for automatic decision-making in the future will be very large.

The more important issue is that even under ideal circumstances, the low latency requirements of 5G will force you to deploy applications near the edge. To meet these latency requirements, architectural changes are required to reduce the number of visits to the back-end server. The actual time delay limit is not the 100 milliseconds required for human-computer interaction, but the few milliseconds specified by the 5G specification.

to sum up

While other suppliers are still talking about what they are going to do for their customers, we have already done it. In the fast data analysis business, the scene forces the upgrade of the basic software architecture. This is an era of change and an exciting time. The future will be the era of millisecond-level large-scale automated transactions. We have seen the rise of a series of market segments, and many database vendors can indeed find solutions to meet the needs of these areas, but when it comes to large-scale automatic millisecond-level decision-making, there will be fewer and fewer real competitors.