Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

Technical talk-re-talk on soft and hard SDN (1)

2025-03-31 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > Internet Technology >

Share

Shulou(Shulou.com)06/03 Report--

The following articles come from the official account of Wechat: young people performing on the Internet, author KkBLuE

I once shared at a live event jointly organized by SDNLAB and TF Chinese community to discuss the topics of software SDN and hardware SDN. Coincidentally, I saw that domestic Daniel manufacturers were also discussing the path of software and hardware, so I had the first article of this article, which is relatively long and takes 5-10 minutes to read. [live Review] TF Live (KK/ Jianxun: cloudy, SDN, and the Theory of Evolution of Internet Workers

DCN academic school, data center network automatic deployment, how to choose soft and hard SDN?

It is strongly recommended that students who do not understand the software SDN must read it, there are a lot of pertinent words, but there are many sentences to be discussed, so excerpt and talk about their own views. Before I write, I would like to make a few statements:

(1) the selection of this article is not aimed at specific manufacturers, any manufacturer has its advantages, and there are too many things worth learning. (2) the discussion of soft and hard SDN has been a long-term topic. My purpose is not to negate the hardware SDN, but to explain what I have seen and felt for your reference. (3) there is no absolutely perfect solution, only suitable and unsuitable. All the progress of IT and network is the result of cognitive accumulation and continuous reflection and innovation of the original technological scheme.

Please read this article after reading the original text first and emphasize again: this article will not negate any point of view of the original text, but will provide you with another perspective to look at SDN.

First of all, from the perspective of the flexibility of automation capability, the degree of automation depends more on the understanding ability and logical abstraction level of network services of different manufacturers, and the correlation between software and hardware is not so high. For example, in the business distribution of SDN, North American manufacturers use the usual simple multi-level form to distribute business logic. On the other hand, domestic manufacturers have basically used the way of graphical interaction, with a higher level of abstraction and more direct logical understanding, which is worthy of praise for domestic manufacturers. With the gradual rise of intention network, hard SDN manufacturers can use IT business language to automate the configuration of the network, and even achieve a higher level of automation such as pre-verification and post-verification. Soft SDN manufacturers are relatively backward. DCN academic school, data center network automatic deployment, how to choose soft and hard SDN?

In particular, I agree with this sentence that the degree of automation depends more on the ability of different manufacturers to understand network business and the level of logical abstraction. If you look at each manufacturer's plan, you can understand the skills and level of R & D personnel behind it. From the point of view of discussing the realization of technology itself, I prefer that both domestic and foreign manufacturers have their own merits. As the saying goes, there is no first word, no second force, and turnip cabbage. However, in this paper, the example of SDN service distribution is more appropriate to discuss the degree of network automation, and there is a slight disconnection between hardware SDN and software SDN. From another perspective, from the perspective of the data center's demand for SDN, the goal we pursue is the implementation of the IT business language mentioned in this article to the configuration of network devices. The one level of the demand here is the implementation of the management configuration interface, whether it is logical and smooth, whether the interface is clean and beautiful, and whether it meets the basic needs, and the other level is that the implementation of the configuration is completed by software or hardware. If we use SDN software to manage the degree of preference of the interface, to evaluate which soft or hard SDN is stronger, it makes sense, but not the whole picture. The level and ability of the management configuration interface precisely reflects different manufacturers' understanding of the business, and the realization of the degree of automation depends on the implementation of the underlying layer, whether it is software or hardware? A comparison is made in the second paragraph of this paper.

Soft SDN lies in that its business development depends on software implementation, which is more flexible and iterative faster, and its ability does not depend on rapid hardware updates. However, in practice, there is a strict version policy, and it is necessary to consider unloading the features that have a serious impact on performance, such as encryption and encapsulation, on the network card. Overall, the software approach is a little more flexible, but it also faces more constraints because it is completely dependent on CPU. Therefore, at this point, we should focus on different manufacturers regardless of soft and hard ways, their automation capabilities show different horizons and practical levels. DCN academic school, data center network automatic deployment, how to choose soft and hard SDN?

The point of view of this article is very pertinent, the software is more flexible, in fact, whether all rely on CPU is open to question, the mentioned CPU will have more constraints, only see the conclusion, how to prove this conclusion, I need to further study. In fact, too flexible means higher requirements for the people who control it. I think any attempt of technology will eventually achieve results, which has something to do with the radius of the user's ability, the logic of the technology, and the method. When choosing hardware products, we will refer to the color page, look at the port, look at the functional support, by the same token, when choosing a software solution, we also need to consider version adaptation, performance considerations, so as to know ourselves and know each other. The reference aspects of the two are different, but the logic behind it is the same, and more importantly, look at what the business needs for the network.

Secondly, from the point of view of the scalability of the network, the two capabilities are equal. Both of them can build large-scale SDN networks, including cascading networks that support cross-domain and multi-DC. Soft SDN can flexibly sacrifice server resources to replace network resources, so the specifications such as tenant and VPC may exceed the hardware resource limit of hard SDN on the switch. however, there is basically no bottleneck in the current mainstream switch specifications, so this advantage can not be reflected in the actual project value. For the time being, it is considered that the two are tied here. DCN academic school, data center network automatic deployment, how to choose soft and hard SDN?

The scalability of the network, shallowly speaking, depends on the following elements, performance, resilience, and functional support.

Performance: generally refers to throughput, forwarding, and related table item support

Elasticity: generally refers to the degree of automation, the degree of expansion, and whether network equipment can change rapidly with the business.

Functional support: in addition to functional support in the network, support such as NAT,FW and corresponding service chains are also added to virtualization or cloud DC.

From my point of view, expansibility is a comprehensive project. if the barrel theory is applied, the above mentioned boards are all boards in the bucket, and the expansibility depends on which piece is the shortest. It is estimated that we can't finish tens of thousands of words, so we are still simple and rough, listing the following two questions:

Software SDN is more flexible than hardware SDN, but its performance is not as good as hardware SDN. There is no doubt about the performance of hardware SDN, so what is the performance of software? Can the demand be met? Do we still use the early impressions of the performance of the early software SDN to look at the performance of the current software SDN?

Hardware SDN is not as flexible as software SDN, so what are the advantages and disadvantages of hardware SDN in terms of flexibility, which are natural deficiencies, and what can be followed up?

There is probably no perfect solution, and the real use of landing is a combination of objective technology and subjective preferences. A long time ago, I had two job choices. I asked a senior what to do. The elder said that you should take a piece of paper. The title is two jobs An and B, and then A writes clearly what is good and what is bad, and B writes clearly what is good and what is bad. It is best to add your current job. I am not sure whether you have chosen right or not, but at least I have thought about it. I may have chosen wrong and do not regret it.

When it comes to soft SDN and hard SDN, the truth is probably the same.

However, technicians always have the desire to pursue perfection. If we disassemble the functions of SDN and hand over some of the functions to the software and some to the hardware, and adopt a combined scheme to enhance our strengths and circumvent our weaknesses, there may be more interesting topics, which we can discuss further in later articles.

Third, in the actual network deployment, reliability and stability is the top priority. No one will tolerate a network that drops links frequently. This hard SDN naturally has a better position, and the commercial reliability it has accumulated from its early days has a huge advantage. Soft SDN also provides the ability to fail over through vSwitch, but as a business software, it is still limited by the unique reliability problems of the software. In addition, soft SDN will bring new challenges to the network operation and maintenance team, the operation and maintenance boundary needs to be extended to the inside of the server, server operation and maintenance overlap in part of the network and business team, there may be conflicts. So we see that those who dare to use soft SDN are basically Internet manufacturers, and it is also an important reason that they have strong technical ability to overcome this problem. DCN academic school, data center network automatic deployment, how to choose soft and hard SDN?

There has always been no doubt about the stability of the network. The more close the business is to the market, the more sensitive it is, and the more profitable the business is, the more it cannot be lost. At this point, we can see that Internet companies, operators, financial enterprises, and other large enterprises all have a desire for stability, and stability is just another barrel.

In this bucket, in addition to the stability of the hardware, it also includes the highly available design of the network, the mechanism of fault recovery and so on. There may be no problem with the operation of the hardware, but a ring of spanning tree will make the network workers die again and again. For the stability and reliability of hardware SDN and software SDN, if there are no sufficient cases and materials, I dare not make an evaluation. I hope my brothers and sisters who have read this article can give me more materials.

As mentioned in the article, Internet manufacturers dare to use soft SDN. They must have their own views on stability and find an opportunity to chat with their brothers on the Internet. as far as I know, domestic AWS,Azure and GCP, domestic BAT, and software SDN are still in the majority. Combined with what I said before, in fact, their technical ability and control ability are related, this is a new topic, I will not say much here.

A topic raised in this paper attracts more attention, not the implementation of technology, but the boundary of management. in short, when using soft SDN, because the execution point is on the server, who cares? Extension point, who will configure it and who will maintain it? To put it crudely, if something goes wrong, who cares?

It must be admitted that the conflict between operation and maintenance is an objective phenomenon, but we do some reflection, this is a management problem? Or is it a technical problem?

Cloud brings changes to the industry, not only in the form of business and technology, but also in organizational structure, management model, division of labor and cooperation.

It is precisely that the market needs the agility of the enterprise's business, which leads to the agility of the enterprise's business requirements of IT, and it is precisely because of the agility demand of IT that there will be the rise of cloud, and it is precisely with the rise of cloud that there will be such a high correlation between network and business. Therefore, the various departments that used to perform their duties under the IT organization now need to sit down together and reformulate the strategy and business division. We can see that many customers within the IT have become cloud business units. The core reason is to break down the original barriers and integrate the whole IT organically.

But organizational change is bound to bring challenges, management authority, ability learning, how to work together between departments, there is too much thinking here, and this conflict, I am more inclined to management thinking, the thinking of the future enterprise IT helmsman and all employees, and the discussion of the stability of soft and hard SDN, we still need to pay more attention to the technical implementation and design architecture.

Accidentally, more than 3000 words, digital transformation, cloud computing, SDN has never been a single problem, more angles, will bring us more thinking, Huawei's article, help us to see the various possibilities to achieve SDN, my text, hope to provide another perspective, more inclined to lead to more thinking, rather than arbitrarily draw conclusions, words also have certain limitations I hope to have more communication and study with my colleagues. I'll see you next time.

No winter will not pass, no spring will not come!

[live Review] TF Live (KK/ Jianxun: cloudy, SDN, and the Theory of Evolution of Internet Workers

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.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

Internet Technology

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report