Data center mobility policy: what needs to be considered in the transition to SDDC?

When the data centre migration, it takes a lot of time and careful study to move your traditional equipment onto the large platform. In mobile data center, what strategies should be kept in mind?

SDDC: a virtualization and software technology for all physical and hardware resources in the data center. SDDC relies on virtualization and cloud computing technology, SDDC is the goal of all the physical resources data center virtualization, virtualization technology, construction consists of a virtual resource pool of resources, not only is the server virtualization also includes storage virtualization and network virtualization etc.. Not only can simplify the server changes, storage changes, network configuration difficult, but also makes the server, storage, network management and configuration operations with repeatability and sustainability.
SDDC enables hardware resources to be configured and scheduled by software, improving flexibility and agility, and a significant advantage is that it greatly reduces the cost of data centers. With the centralized software management layer provided by SDDC, management becomes simpler. At the same time make SDDC software to manage the network, make the network become a part of the data center, a proprietary network can greatly improve the efficiency of the hardware, thus the importance of software can not be ignored. Software defined data centers will become a new direction and trend of data center evolution.
So, when you've decided to start building or to SDDC (software defined data centers) over, you need to think about some issues. For example, when you're on the hot channel of the data center at the moment, you might be frustrated. How do you deal with the existing devices in the data center and the applications that run on them? The only answer is: like most things in information technology - "it depends on the situation."".
For mobile data centers, two basic models should be considered: first, run some old and new ones in parallel, or integrate your existing devices with the new SDDC in some form of transformation. Second, the integration of existing network devices into an independent data center, wrong, not concentrated to one, is two, or integrated in the pod layer, or integrated in existing devices running SDDC top layer.
The first model is to run the old and new data centers in parallel, which may seem simpler, even desirable. But even in the ideal case, there are some problems to be solved. Will workloads be transformed between data centers? If so, how does this happen? Of course, the answer to this question largely depends on the application itself. There are several important questions to be asked in this field.
How does the new data center architecture meet the requirements of each specific application? It is important to consider the issues that are usually considered, such as bandwidth utilization, delay and jitter (Jitter is composed of deterministic content and Gauss (random) content. The so-called jitter is a kind of jitter. How do you explain that? Its definition delays from the source address will be sent to the target address, there will be a different delay, such a delay change is jitter.) and other needs. But considering the domain name system, the dynamic management of elephant traffic, the creation of security domain, overlay network and other factors are also very important.
Elephant flows: in computer networks, elephant traffic is a very large (total byte) continuous traffic that is established by measuring TCP (or other protocol) traffic on the network link. Although the flow of elephants is not very large, it can occupy a disproportionate share of the total bandwidth over a period of time. It is not clear who created the "elephant", but on the Internet the word published in 2001 in the beginning to happen, when observed, a small amount of traffic carrying most of the Internet traffic, the remaining traffic flow contains a large number of these network traffic is very small (mouse flow). For example, researcher Mori et al studied the traffic volumes of several Nihon University and research networks. In the WIDE network, they found that the elephant traffic accounted for only 4.7% of all traffic, but during this time occupied the 41.3%. of all data transmission
The actual impact of elephant traffic on Internet traffic is still a field of research and debate. Some studies have shown that elephant flow is likely to be highly correlated with traffic peaks and other elephant flows. There are different definitions of elephant flow among researchers, including over 1% of the total flow in a period of time, measuring the duration of the flow, and observing the magnitude of the flow greater than the mean, plus the standard deviation of traffic during this period. One of the main goals of elephant flow research is to develop more efficient bandwidth management tools and Internet prediction models. For example, researchers focus on providing better service quality for small size (small flow) traffic by prioritizing traffic flows.
Even in the initial design stage, it is best to avoid some technical problems related to the late service provided by the data center. In fact, in the process, there are always some omissions.
Because there are very few application developers who know what services the application is relying on, or make invalid assumptions during the execution of such lists, at least they don't have complete inventory.