1. Why is flexibility so important
Be flexible so that IT capabilities can easily keep up with the user's business development, so that it can be more, faster, better and less expensive.
- Many: There are many choices, you can build a website globally, there are many models to choose, and there are many supporting cloud services.
- Fast: deployment is fast, the construction time of self-built IDC is in years, and what you see is what you get on the cloud.
- Good: Good maintenance, no need to consider IT hardware failure maintenance, work energy can be fully considered business development.
- Province: Create IT resources on demand based on computing tasks, solve the problem of vacant IT resources, and optimize IT costs unprecedentedly.
2. How to do flexibility
The construction of the supply side is mainly the responsibility of cloud vendors. The purpose is to provide cloud users with more flexible resources while reducing operating costs. How to achieve the flexibility of supply-side construction to be "inclusive of all rivers" can be said to be the unique internal strength of each cloud vendor, and it is also an important criterion for evaluating the merits of a cloud vendor.
The construction of the consumer side needs to be jointly built by cloud users and cloud vendors. Cloud vendors build the stage and cloud users perform. Cloud vendors should provide users with the most convenient and flexible operation interface and API for resource creation and resource scheduling. What do cloud users need to do?
Figure 1: Flexible consumption side and supply side
3. How to deal with consumer elasticity
1. First choose one or several cloud vendors as cloud vendors. How to choose cloud vendors for flexibility is mainly based on the market share and stability of this cloud vendor. Because cloud vendors with a larger market share generally have larger inventory pools in various regions, more inventory allocation methods, and more secure supply chains, they can provide us with more flexible resources. Considering the stability of word of mouth is naturally to consider the stability of your business.
2. Identify which resources are flexible and which cannot.
Figure 2: Flexible and inflexible
3. Capacity assessment
Inflexible resources should be evaluated based on mid-to-long-term business planning and peak estimation when evaluating capacity, because their construction period is generally very long or they cannot respond elastically when there is a sudden peak.
The elastic resource capacity evaluation generally uses the resources required during the daily average business flow as the baseline level, and slightly adjusts the baseline level in consideration of the delivery speed of elastic expansion resources and the resource level of cloud vendors.
Figure 3: Resource water level
4. Flexible costing
Cloud resources are generally purchased in two common forms: annual subscription and monthly subscription and pay-as-you-go. A small number of cloud products support competitive bidding instances
- Pay-as-you-go: Similar to the post-payment method of mobile phones, paying according to the actual length of use. This is the most flexible way to purchase. It is always on the go, but the price is usually the highest, suitable for temporary expansion of explosive business. For example, temporary testing, business peak hours, etc.;
- Annual and monthly subscription: Similar to the prepaid package for mobile phones, it is bound to a specific cloud server instance, and promises to be used for a long time. The flexibility in time is less, but the price is relatively favorable. It is suitable for running fixed 7x24 services, such as web services, Database, etc.;
- Preemptive bidding: This is a very unique way of selling. We use temporarily surplus resources at a very low price to allow users to buy in a manner similar to bidding, but when the supply situation changes, these cloud servers may be clouded Manufacturers "preempt" to go back, so it is more suitable for computing tasks that can be interrupted at any time and then resumed afterwards, such as stress testing, scientific computing, batch computing, video transcoding, etc.
The resources within the flexible baseline are highly utilized, so it is recommended to purchase them on a yearly and monthly basis, which is more cost-effective.
If the daily peak lasts for more than 1 hour and the data needs to be stored for a long time, it is recommended to purchase by volume.
If the hotspot peak is relatively short within 1 hour and there is no requirement for data persistence, you can consider buying an auction preemptive instance.
5. Flexible resource scheduling
- Timing elastic scheduling If the business traffic is very regular every day, you can consider using the timing elastic scheduling method to expand the capacity before the business peak, and shrink the capacity after the peak.
- Dynamic Elastic Scheduling Dynamic resource scheduling can automatically expand and shrink capacity according to business traffic trends. To achieve dynamic resources, you need to do the following steps.
Figure 4: Elastic water level assessment
Define the water level for expansion and contraction
Figure 5: Elastic water level
6. Design a flexible strategy for high availability
- Try to make the business district more than can fit more of the available models do elastic deploy multiple availability zones to deploy disaster recovery can do multi-room
multi-zone deployment can be used in parallel to create a resilient resources in a multi-room, increase the flexibility of speed
and more resilient model can do Reduce the difficulty of flexible inventory protection and expand flexible scale - Inflexible resources should also be considered for high availability. If there are dedicated lines, try to consider dual-line dual-access points to connect to cloud vendors,
such as caches. After expansion, it needs to be warmed up for a long time. Doing normal deployment and considering multi-zone deployment can allow elastic resources to call the cache nearby and optimize the call time.
This article is the original content of Alibaba Cloud and may not be reproduced without permission.