With the launch of 5G networks, IoT devices are expected to become truly mainstream to some It is estimated that their number will exceed 50 billion by 2022.
Experts note that the existing infrastructure will not cope with the increased load. By estimated Cisco, in two years the traffic passing through the data center will reach 20,6 zettabytes.
For this reason, IT companies spend billions on the development of network infrastructure. For example, Google deal laying new submarine cables in Asia and Europe to reduce delays in data transmission at points remote from data centers. Also, IT giants are building hyperscale data centers - for their creation AWS, Microsoft and Google in total already spent over 100 billion dollars.
Obviously, in such (and simpler) systems, it is impossible to manually monitor the correct operation of all switches, servers, and cables. This is where software-defined networks (SDN) and special protocols (for example, OpenFlow).
At Statista they saythat by 2021 the volume of traffic passing through the SDN systems of data centers will more than double: from 3,1 zettabytes to 7,4 zettabytes. For example Fujitsu introduced SDN technology in hundreds of its data centers located in different parts of the world. Uses software-defined networks and one of the local cloud providers in the United States.
Experts from IDC expect the SDN market to continue to evolve. By 2021 it volume will reach $13 billion, given that in 2017 it was estimated at $6 billion.
Switch to virtual machines
The popularity of virtualization in recent years is associated with the development of a large number of tools that automate the management of VMs and increase their availability.
IaaS providers also provide automation tools to their customers. For example, we are in 1cloud offer a software interface that allows you to raise a new virtual machine in a couple of minutes. It is also possible to manage infrastructure via API. For example, you can configure the shutdown of virtual machines according to a specified schedule so as not to pay for their "idle" work. The API can also be used to change the number of cores and the amount of RAM.
Virtualization solution management systems are developing towards the use of machine learning technologies that automatically distribute the load between VMs. For example, this functionality has a solution for VMware NSX virtual environments. It already helps IaaS providers load balance in multi-cloud and hybrid environments.
Implement DCIM systems
DCIM solutions (Data Center Infrastructure Management) are software that monitors the operation of data center engineering systems: energy consumption of servers, storages, routers, power distributors, humidity levels, etc. Such systems are available in Dataspace and Xelent data centers, where 1cloud hosts its hardware.
In the first case, the DCIM system manages electricity and water supply, air conditioning server rooms and video surveillance throughout the building. In the second - automatically stabilizes output voltage in the mains, protecting servers and eliminating micro-breaks.
Artificial intelligence systems have also reached this area. Smart algorithms predict server breakdowns by analyzing their βbehaviorβ. For example, Litbit works over Dac technology. The system monitors the state of iron using sensors installed in the machine room. They analyze ultrasonic frequencies and floor vibrations.
Based on this data, Dac detects anomalies and determines if all equipment is working properly. If there are problems, the system notifies data center operators or independently disables faulty servers.
So far, these technologies are not very widespread, but recently they have significantly strengthened their positions. By analysts' forecasts, in 2022 the volume of the DCIM market will be $8 billion, which is twice as much as in 2017. Soon, these solutions will begin to appear in all major data centers.