Datacenter migration：Procurement infrastructure

Datacenter migration，The Internet industry's interest in the super fusion infrastructure (HCI) has replaced the all flash memory array in a sense, which has become a hot topic in recent years. Enterprises and organizations have shifted interest to HCI, mostly because of the simplicity they bring. Users are no longer willing to (or are forced to) spend time planning, building and implementing the three tier architecture that is designed around independent servers, servers, storage and network components. HCI seems to be a better choice to integrate these functions into a single device.

However, the rapid deployment of hyper fusion systems at multiple locations always brings a sense of urgency, and for a variety of reasons, some IT managers spend more money on it than they actually need.
The following are three erroneous choices that should be paid attention to when choosing and deploying HCI. Many people may have fallen into these pits more or less.
Error 1: apply the solution to the data center on the edge.
Most HCI solutions are designed for data center usage, and may be too much for edge deployment. Usually, deploying such a data center class in a small remote environment requires too much cost and management costs and may be very hard and hard for the operation and maintenance managers.
At the stage of RFP, several key problems should be put forward to ensure that they will not oversource resources because they buy more suitable solutions for large data centers. The first thing to be asked is the minimum number of nodes or devices needed to achieve high availability solutions. If a node or device fails, another node or device will automatically achieve overall interoperability. Secondly, what kind of network should be determined to run local device clusters and witness servers. Many solutions require 1Gbit or 10Gbit connections, and the edge environment is usually located in remote locations where high bandwidth networks can not be implemented. Third, you can review and refuse to increase the data center class characteristics that are not needed in the marginal environment. Many solutions have data center characteristics, which inevitably improves the overall cost, and many resources have never been used.
Error 2: the number of deployed hardware exceeds the actual demand.
Because many HCI solutions are pre configured devices, enterprises tend to fall into the "standardized" trap of deploying nodes, which are more CPU, network, or storage space than actual requirements. In order to maintain cost compression within the budget, especially for edge deployment, users should analyze their precise IO, storage, and network requirements to properly configure the hardware and reduce excess waste.
IT staff usually extend resources to HCI devices for simplicity, but there are often many hidden costs. For example, if the business application needs additional storage capacity or performance, the customer finds it too late, so they have to buy a new device or node - plus the license fee - to effectively expand the cluster. This means that it is very necessary to further understand the details of the management requirements of the high availability cluster group and to witness the hardware requirements remotely. Some remote witness solutions need the relationship of 1:1, that is, a witness of each site, and have strict hardware requirements. If users don't think carefully, they will spend a lot of unnecessary budgets.
Error 3: use full flash at unnecessary time.
All flash arrays and super fusion devices are often touted by suppliers as "the best performance". However, not all applications need to do this, and because the cost of flash memory is usually 10X, the overall cost is always high. Users should study automatic adjustment or caching software to better optimize the placement of data between flash and rotating disks, and in some cases, system memory can also deliver the same performance without the need for additional costs.
For the specific applications that require high performance, such as financial services, HPC environment, reservation system, etc., the performance of all flash array is meaningful. But for those daily applications that only need good performance, the performance of the hybrid storage medium is enough, and the cost is only a small part.
Last
So, what will the future of HCI be? According to the research of MarketsandMarkets, a famous international market research and consulting company, by 2022, the HCI market is expected to reach $13 billion, and the annual compound growth rate (CAGR) can reach 44%.. The future of the hyper fusion infrastructure will be a major trend in the data center. The final thing is to consider the cost problem. It is very necessary to reasonably deploy HCI devices according to their own business needs. The above error deployment methods are some common problems in the industry. I hope this guide to prevent pit entry can provide you with some small help.