Trends in the development of cloud data centers
The technological development of cloud data centers (Cloud DC) can be divided into three stages:
Fig. 1. Stages of development of data centers: an analysis of the past, current status and forecasts
Issues related to security vulnerabilities in public cloud security systems are worrying even large services like Apple iCloud and Amazon WebServices. Therefore, almost all more or less large customers, primarily government agencies, prefer to deploy their private clouds, rather than resorting to public services. The target audience of public clouds is individuals, small start-up companies, small and medium businesses. Private clouds are designed to serve customers with much higher requirements for security and reliability, as well as those who are forced to comply with legal restrictions.
Hybrid clouds are intended primarily for users of a private cloud, but allow the use of public cloud services with shared access (for example, when rapid expansion is needed).
')
The most popular cloud service today is IaaS (Infrastructure as a Service). Only a small part of customers consider PaaS (Platform as a Service) as a guideline for development.
The development of the market in recent years shows that the open platform OpenStack is quite capable of becoming a leader in the IaaS business. Corporate users will be able to choose IaaS components from different vendors that support OpenStack to avoid "vendor dependency." A key factor in convincing customers that IaaS can be deployed without any problems in their IT systems can be supported by the vendor of the universal standard OpenStack.
The future of PaaS is unclear, both in terms of customer needs and technological development. Large-scale development and application of this technological concept over the next few years is unlikely.
Obviously, VxLAN (Virtual Extensible LAN) technology has the greatest prospects, while the future of NvGRE (Network Virtualization using Generic Routing Encapsulation) is not so bright. More and more customers indicate in the technical specifications support for VxLAN, and only occasionally support for NvGRE is requested, and then only in case.
However, although VxLAN is becoming a market leader as the basis for integrating cloud platforms and SDN networks, mature and stable demand for VxLAN services is still ahead. The fact is that this technology has not yet reached the desired level of maturity. In addition, VxLAN solution developers use different hardware and software components, network models and implementation methods, and customer requirements also vary. Consequently, the interaction between cloud platforms of different vendors and SDN networks continues to be problematic, and it will take at least a year to solve this problem.
However, this does not mean that “mission is impossible”. There are already a number of successful projects of heterogeneous interaction with the following features:
Some time ago it was believed that the separation of planes in the SDN network (control planes and data planes) would cause dramatic changes in the network architecture. For example, experts predicted the emergence of manufacturers specializing only in the development of SDN controllers and switches with SDN support, and network switches would turn into standard controlled devices with such a development. It is already clear that these forecasts have not come true.
The development of SDN technology is reminiscent of the situation with the so-called Next Generation Network (NGN) next-generation networks on telecoms operator networks. Despite the tendency of abstraction of control planes (Softswitch) and switching (IP network), telecom operators preferred to purchase both Softswitch and IP devices they manage from the same vendor (although they can be bought from different vendors) to reduce technological risks.
Similarly, the use of SDN-controllers and switches from different vendors is not yet very popular. The chances of survival of vendors specializing solely in SDN controllers will be small. In addition, with the further development and maturity of VxLAN, the popularity of Layer 2 protocol TRILL (Transparent Interconnection of Lots of Links) protocol will gradually disappear.
Public cloud platforms support a huge number of users and process massive amounts of data. However, the performance advantage is accompanied by less functionality and the number of supported services.
Private clouds are deployed in different conditions and for different use cases. Therefore, they should provide a wider range of functions. The performance of private clouds at the same time lower than that of public. Is it possible to provide rich functionality and high performance on a single platform? In the long run, this is hardly feasible.
In addition, as in all large systems, the advantages of a large volume of public clouds will be noticeable only after a certain number of services have been deployed. If a certain threshold value is not exceeded, the public cloud will not have distinct advantages over the private in terms of price / performance. Therefore, vendors should choose one thing - either private or public clouds - and, accordingly, develop different versions of equipment for one or another type of cloud solutions. Of course, you can try to develop a universal platform for both public and private clouds, but the lifespan of such a model is unlikely to be long.
It follows logically that hybrid cloud platforms designed for various types of corporate customers that require diversified functions should be based on the private cloud architecture. In addition, hybrid clouds must have the ability to interact with public cloud platforms and use their services dynamically. The interaction between public and private clouds is reminiscent of the joint work of a public network operator and PBX in an enterprise. So far, there is no unified expert opinion on how services of private and public clouds will be provided - on the same or on different platforms. Time will tell.
The main trend of the ICT industry 15–20 years ago was the transition from private (proprietary) to open systems. Now the main trend is the transition to open source software (OpenSource), and the emergence of cloud technologies only accelerates this trend. Similar transition in different countries is realized differently. Technologically advanced countries see the potential of cloud technologies and OpenSource as one of the incentives for the growth of the ICT industry. For less developed countries, OpenSource means increased controllability and manageability, and therefore, security. However, some experts see some risks in OpenSource.
Fundamental changes in the ICT industry
However, OTT service providers and equipment vendors are not direct competitors. OTT providers specialize in providing public cloud services, while hardware vendors provide private cloud solutions. Therefore, it is more correct to consider the competition between OTT suppliers and vendors as competition between public and private clouds. But, as we said above, their fields of application are different. While public clouds are primarily intended for private users, Internet startups and SMBs, then private clouds are mainly aimed at large enterprises and government organizations. Some of them have modified their IT platforms to fit the private cloud.
Nevertheless, the situation on the market for private and public clouds is unlikely to change drastically in the coming years due to technology restrictions, information security requirements, legislation, etc. Thus, the relationship between OTT and vendors is not direct competition, but rather market sharing.
Traditional equipment vendors are under pressure from OTT in four areas:
These factors affect the revenues and margins of traditional equipment vendors and lead to the erosion of established value chains in the ICT industry.
OTT is an emerging industry, and market competition is not so fierce here. OTT players now have sufficient profit and margins to support their own service development. However, as the OTT segment matures, competition will increase, and OTT providers will inevitably face lower revenues and margins.
In addition, the development of data center technologies (in particular, DC 3.0) will require more investment in research and development (R & D). The declining revenues and the growth of R & D will lead to the fact that the number of OTT, leading their own development, will decrease, and their technological niches will be reduced. This will lead to the start of OTT cooperation with traditional vendors of equipment (approximately the same way equipment suppliers cooperate with telecom operators).
Companies will strive to acquire complete integrated cloud platform complexes, rather than buying and assembling them in parts (servers, network devices, cloud software), as they did before. Customers will take care of functions and specifications related to cloud platforms, rather than specific approaches to implementing solutions on the equipment of various vendors. Today, there are many types of all-in-one equipment that can be considered as early prototypes of modular IaaS (for example, this is Huawei's FusionCube solution).
Competences of traditional vendors in terms of possible ways of technology development will weaken for two reasons. First, as mentioned above, companies are starting to change their purchasing behavior and are moving from purchasing individual components to using modular IaaS. That is, they redirect attention from equipment to cloud services. In addition, cloud systems are moving to OpenSource platforms, where OpenStack is becoming a key technology. As a result, traditional vendors (for example, network equipment) will inevitably lose some of their influence. The growing influence of OTT will also lead to a decrease in their weight in the market. A typical example is the Open Compute Project (OCP) open computing project initiated by Facebook, in which many companies participate.
As follows from the above opinions, the size of the market for vendors providing a limited number of technologies and products will be reduced. It becomes clear why so many vendors are concerned about mergers and acquisitions, the restructuring of their business and why they are seeking to gain influence in the cloud platform market.
The business of integrators will also shift to cloud platform sales — from equipment sales to sales of cloud services and OTT services. Thus, they will be transformed from suppliers of a limited set of technological equipment (for example, for corporate networks) to suppliers of modular IaaS.
DC 3.0 Technology Forecast
Today, the final definition of the functions and technologies of DC 3.0 has not yet happened. So far, the industry has only strengthened the view that the DC 2.0 architecture is unable to meet the future explosive growth of data traffic, therefore new solutions and new approaches are required.
The need for services from a private cloud is estimated by business requirements. In the case of public clouds, it is also necessary to take into account the size of the user base, and it can be hundreds or thousands of times larger than the number of corporate users. Therefore, it is not surprising that the data processing capacity in them should also increase hundreds and thousands of times. The gap between the requirements of private and public clouds is increasing. Therefore, the use of fundamentally different architectures for private and public clouds and for different market segments would be the right approach. This trend will be more apparent with the development of next-generation data center technologies.
Conclusion
Public and private clouds differ significantly from each other in technology and target markets, and these differences are increasing. The point of view of the public cloud adherents who claim that the ICT industry will one day become as public as today's power grids (powergrids) are gradually losing relevance. Obviously, the ICT industry has much more personalized requirements than power grids and traditional telephone networks.
Business models of both vendors and integrators will also change. For traditional vendors, success or failure in the corporate market depends on their persuasion (credibility) and market share. The OTT / ISP market is more optimistic about future development. As OTT and public service markets mature, the corresponding ICT technologies will become increasingly complex and diversified. It is likely that equipment vendors and OTT will establish cooperation similar to today's partnership of telecom operators and vendors.
The development of cloud computing and OTT is a headache for leading vendors in the traditional corporate market. Their persuasiveness regarding technological development will weaken. They must radically change their business models. These changes are inevitable, and we foresee their future rebirth.
- DC 1.0 is a consolidation stage, that is, a transition from autonomous, scattered data centers to cloud-based distributed ones.
- DC 2.0 is a modern stage characterized by a fundamental shift in the system design and deployment of data centers, when, due to resource virtualization and dynamic orchestration of DSO (Dynamic Service Orchestration) services, the load factor of resources and the flexibility of their distribution significantly increased.
- DC 3.0 is the next stage of development, caused by the rapidly growing needs for ICT services, as well as the growth of data volumes in the networks of large Internet companies. While most of the technology DC 3.0 is in its infancy, and the future path of their development is not clear enough.
Fig. 1. Stages of development of data centers: an analysis of the past, current status and forecasts
- The main problem of public clouds is security.
Issues related to security vulnerabilities in public cloud security systems are worrying even large services like Apple iCloud and Amazon WebServices. Therefore, almost all more or less large customers, primarily government agencies, prefer to deploy their private clouds, rather than resorting to public services. The target audience of public clouds is individuals, small start-up companies, small and medium businesses. Private clouds are designed to serve customers with much higher requirements for security and reliability, as well as those who are forced to comply with legal restrictions.
Hybrid clouds are intended primarily for users of a private cloud, but allow the use of public cloud services with shared access (for example, when rapid expansion is needed).
')
- IaaS is the focus of cloud computing, and OpenStack is the clear winner in the race for survival.
The most popular cloud service today is IaaS (Infrastructure as a Service). Only a small part of customers consider PaaS (Platform as a Service) as a guideline for development.
The development of the market in recent years shows that the open platform OpenStack is quite capable of becoming a leader in the IaaS business. Corporate users will be able to choose IaaS components from different vendors that support OpenStack to avoid "vendor dependency." A key factor in convincing customers that IaaS can be deployed without any problems in their IT systems can be supported by the vendor of the universal standard OpenStack.
The future of PaaS is unclear, both in terms of customer needs and technological development. Large-scale development and application of this technological concept over the next few years is unlikely.
- VxLAN is the most promising tool for integrating data centers and SDN networks.
Obviously, VxLAN (Virtual Extensible LAN) technology has the greatest prospects, while the future of NvGRE (Network Virtualization using Generic Routing Encapsulation) is not so bright. More and more customers indicate in the technical specifications support for VxLAN, and only occasionally support for NvGRE is requested, and then only in case.
However, although VxLAN is becoming a market leader as the basis for integrating cloud platforms and SDN networks, mature and stable demand for VxLAN services is still ahead. The fact is that this technology has not yet reached the desired level of maturity. In addition, VxLAN solution developers use different hardware and software components, network models and implementation methods, and customer requirements also vary. Consequently, the interaction between cloud platforms of different vendors and SDN networks continues to be problematic, and it will take at least a year to solve this problem.
However, this does not mean that “mission is impossible”. There are already a number of successful projects of heterogeneous interaction with the following features:
- Most customers are Internet companies. Almost all of them use open source software to integrate their cloud platforms with data center infrastructure.
- Solutions for interoperability of cloud platforms and SDN networks can be tailored to the specific needs of customers (customization).
- Currently, the interface between the SDN controller and the service platform is much more important than the interface between the control plane (SDN controller) and the data plane.
Some time ago it was believed that the separation of planes in the SDN network (control planes and data planes) would cause dramatic changes in the network architecture. For example, experts predicted the emergence of manufacturers specializing only in the development of SDN controllers and switches with SDN support, and network switches would turn into standard controlled devices with such a development. It is already clear that these forecasts have not come true.
The development of SDN technology is reminiscent of the situation with the so-called Next Generation Network (NGN) next-generation networks on telecoms operator networks. Despite the tendency of abstraction of control planes (Softswitch) and switching (IP network), telecom operators preferred to purchase both Softswitch and IP devices they manage from the same vendor (although they can be bought from different vendors) to reduce technological risks.
Similarly, the use of SDN-controllers and switches from different vendors is not yet very popular. The chances of survival of vendors specializing solely in SDN controllers will be small. In addition, with the further development and maturity of VxLAN, the popularity of Layer 2 protocol TRILL (Transparent Interconnection of Lots of Links) protocol will gradually disappear.
- Public and private cloud platforms will evolve in different ways.
Public cloud platforms support a huge number of users and process massive amounts of data. However, the performance advantage is accompanied by less functionality and the number of supported services.
Private clouds are deployed in different conditions and for different use cases. Therefore, they should provide a wider range of functions. The performance of private clouds at the same time lower than that of public. Is it possible to provide rich functionality and high performance on a single platform? In the long run, this is hardly feasible.
In addition, as in all large systems, the advantages of a large volume of public clouds will be noticeable only after a certain number of services have been deployed. If a certain threshold value is not exceeded, the public cloud will not have distinct advantages over the private in terms of price / performance. Therefore, vendors should choose one thing - either private or public clouds - and, accordingly, develop different versions of equipment for one or another type of cloud solutions. Of course, you can try to develop a universal platform for both public and private clouds, but the lifespan of such a model is unlikely to be long.
It follows logically that hybrid cloud platforms designed for various types of corporate customers that require diversified functions should be based on the private cloud architecture. In addition, hybrid clouds must have the ability to interact with public cloud platforms and use their services dynamically. The interaction between public and private clouds is reminiscent of the joint work of a public network operator and PBX in an enterprise. So far, there is no unified expert opinion on how services of private and public clouds will be provided - on the same or on different platforms. Time will tell.
- The transition from open systems to open source software.
The main trend of the ICT industry 15–20 years ago was the transition from private (proprietary) to open systems. Now the main trend is the transition to open source software (OpenSource), and the emergence of cloud technologies only accelerates this trend. Similar transition in different countries is realized differently. Technologically advanced countries see the potential of cloud technologies and OpenSource as one of the incentives for the growth of the ICT industry. For less developed countries, OpenSource means increased controllability and manageability, and therefore, security. However, some experts see some risks in OpenSource.
Fundamental changes in the ICT industry
- The development of the OTT (Over the Top) ISP business has the following implications:
- Telecom operators have a choice: to purchase either the equipment from the vendor to implement the required functionality or to purchase the OTT service.
- System integrators are faced with the need to reorient their business from equipment integration services to integrating OTT services.
- Corporate IT professionals also have a choice: to purchase either equipment or OTT service.
- For OTT OpenStack and OpenFlow themselves will be the main technologies.
However, OTT service providers and equipment vendors are not direct competitors. OTT providers specialize in providing public cloud services, while hardware vendors provide private cloud solutions. Therefore, it is more correct to consider the competition between OTT suppliers and vendors as competition between public and private clouds. But, as we said above, their fields of application are different. While public clouds are primarily intended for private users, Internet startups and SMBs, then private clouds are mainly aimed at large enterprises and government organizations. Some of them have modified their IT platforms to fit the private cloud.
Nevertheless, the situation on the market for private and public clouds is unlikely to change drastically in the coming years due to technology restrictions, information security requirements, legislation, etc. Thus, the relationship between OTT and vendors is not direct competition, but rather market sharing.
Traditional equipment vendors are under pressure from OTT in four areas:
- OTT suppliers reduce the profits of traditional vendors, delaying market share.
- OTT is changing the way companies look at the development of traditional IT systems.
- OTT or OEM's own developments reduce the amount of equipment purchased from traditional vendors.
- OTT providers can offer more flexible contract terms due to their large-scale purchases.
These factors affect the revenues and margins of traditional equipment vendors and lead to the erosion of established value chains in the ICT industry.
- OTT providers and traditional vendors are gradually becoming partners.
OTT is an emerging industry, and market competition is not so fierce here. OTT players now have sufficient profit and margins to support their own service development. However, as the OTT segment matures, competition will increase, and OTT providers will inevitably face lower revenues and margins.
In addition, the development of data center technologies (in particular, DC 3.0) will require more investment in research and development (R & D). The declining revenues and the growth of R & D will lead to the fact that the number of OTT, leading their own development, will decrease, and their technological niches will be reduced. This will lead to the start of OTT cooperation with traditional vendors of equipment (approximately the same way equipment suppliers cooperate with telecom operators).
- The change in “purchasing behavior” is a transition from purchases of individual components to the acquisition of IaaS modules.
Companies will strive to acquire complete integrated cloud platform complexes, rather than buying and assembling them in parts (servers, network devices, cloud software), as they did before. Customers will take care of functions and specifications related to cloud platforms, rather than specific approaches to implementing solutions on the equipment of various vendors. Today, there are many types of all-in-one equipment that can be considered as early prototypes of modular IaaS (for example, this is Huawei's FusionCube solution).
- Changes in the industry chain.
Competences of traditional vendors in terms of possible ways of technology development will weaken for two reasons. First, as mentioned above, companies are starting to change their purchasing behavior and are moving from purchasing individual components to using modular IaaS. That is, they redirect attention from equipment to cloud services. In addition, cloud systems are moving to OpenSource platforms, where OpenStack is becoming a key technology. As a result, traditional vendors (for example, network equipment) will inevitably lose some of their influence. The growing influence of OTT will also lead to a decrease in their weight in the market. A typical example is the Open Compute Project (OCP) open computing project initiated by Facebook, in which many companies participate.
- Space for highly specialized vendors will be reduced.
As follows from the above opinions, the size of the market for vendors providing a limited number of technologies and products will be reduced. It becomes clear why so many vendors are concerned about mergers and acquisitions, the restructuring of their business and why they are seeking to gain influence in the cloud platform market.
The business of integrators will also shift to cloud platform sales — from equipment sales to sales of cloud services and OTT services. Thus, they will be transformed from suppliers of a limited set of technological equipment (for example, for corporate networks) to suppliers of modular IaaS.
DC 3.0 Technology Forecast
Today, the final definition of the functions and technologies of DC 3.0 has not yet happened. So far, the industry has only strengthened the view that the DC 2.0 architecture is unable to meet the future explosive growth of data traffic, therefore new solutions and new approaches are required.
- Processing speed in public clouds in the future will grow hundreds, if not thousands of times.
The need for services from a private cloud is estimated by business requirements. In the case of public clouds, it is also necessary to take into account the size of the user base, and it can be hundreds or thousands of times larger than the number of corporate users. Therefore, it is not surprising that the data processing capacity in them should also increase hundreds and thousands of times. The gap between the requirements of private and public clouds is increasing. Therefore, the use of fundamentally different architectures for private and public clouds and for different market segments would be the right approach. This trend will be more apparent with the development of next-generation data center technologies.
Conclusion
Public and private clouds differ significantly from each other in technology and target markets, and these differences are increasing. The point of view of the public cloud adherents who claim that the ICT industry will one day become as public as today's power grids (powergrids) are gradually losing relevance. Obviously, the ICT industry has much more personalized requirements than power grids and traditional telephone networks.
Business models of both vendors and integrators will also change. For traditional vendors, success or failure in the corporate market depends on their persuasion (credibility) and market share. The OTT / ISP market is more optimistic about future development. As OTT and public service markets mature, the corresponding ICT technologies will become increasingly complex and diversified. It is likely that equipment vendors and OTT will establish cooperation similar to today's partnership of telecom operators and vendors.
The development of cloud computing and OTT is a headache for leading vendors in the traditional corporate market. Their persuasiveness regarding technological development will weaken. They must radically change their business models. These changes are inevitable, and we foresee their future rebirth.
Source: https://habr.com/ru/post/280598/
All Articles