It is another year of Think in Cloud. The cause and form of the epidemic have changed. Online + offline will be performed simultaneously, but the excitement remains the same. At this conference, UCloud brought many of the latest product technologies and best practices. I followed it all day and sighed countless times. UCloud can survive in the public cloud market under the influence of giants and become the first public cloud listed. By luck, UCloud has two brushes.
This article takes advantage of the release of UCloud's new product Cube container instance (hereinafter referred to as "Cube") to talk about why UCloud can be successful.
When it comes to containers, everyone should be familiar with it.
With the advent of the cloud computing 2.0 era, cloud native has received unprecedented attention. As the most basic technology of cloud native, container has also achieved rapid popularity in a short time. According to survey data from authoritative organizations, 43.9% of users have adopted container technology in 2019, and 40.8% of users plan to adopt container technology. The development momentum of containers can be said to be very rapid. At the same time, Kubernetes has become the de facto standard, almost dominating the container world.
Then a series of questions are coming. Now that Kubernetes has become a big success, why does UCloud still make a product like Cube? In fact, UCloud also has Kubernetes product UK8S. What is the unique value of Cube? What is the relationship between Cube and UK8S? Let us answer these questions one by one.
01
Cube, in order to benefit more customers
Kubernetes is fragrant, but not all users can share the technical dividends of Kubernetes. Why do you say that? UCloud's customer service journey in the past two years can be said to be the best annotation.
UCloud product manager Zhang Pengbo mentioned a very interesting phenomenon in his speech. When they communicated with some customers two years ago, they said they wanted to embed the business into Kubernetes, but two years later, they were still saying the same thing. .
UCloud is very confused, because UCloud has also done a lot of online and offline technical training. But the result is still very unsatisfactory. What is the essential reason? In fact, in one sentence, Kubernetes is too complicated. Many companies do not have enough capabilities and resources to systematically learn, plan, and fully implement Kubernetes.
In fact, this is also a concentrated reflection of the problem feedback received in the process of serving customers in the two years since UCloud launched the Kubernetes product UK8S in 2018. According to Zhang Pengbo, users' confusion about Kubernetes can be summarized into three aspects:
1. The Kubernetes system is more complex and the learning curve is steep. It requires the customer team to have a certain technical reserve. This is also true for customers who have already used containers but have not yet tried Kubernetes. On the one hand, they need to understand the Kubernetes technical system and on the other hand, they need to modify the application architecture. With Kubernetes.
2. Maintaining a Kubernetes cluster will add additional burdens. In addition to managing applications, users also need to manage back-end resources, and application-centric business management cannot be realized.
Third, Kubernetes will cause a waste of resources, because it builds containers based on a virtual framework instead of directly building containers, which also leads to a long waiting time for application readiness, and does not fully reflect the agile characteristics of containers.
All these factors make UCloud determined to develop a new container product, so that companies without too strong technical capabilities can also use and use it well. This new product is the Cube that this article will talk about.
02
Cube, what's the difference
Let's take a comprehensive look at Cube.
Let's take an intuitive look at the comparison of the use process of Kubernetes and Cube. As shown in the figure below, it is clear at a glance. Compared with Kubernetes, Cube saves three time-consuming and laborious steps. Kubernetes is to learn first, and Cube only needs container image to use.
After having an intuitive understanding of Cube, let's disassemble Cube in detail below.
Simply understand Cube, it is a small Kubernetes system. In Cube, UCloud retains user-friendly functions such as Kubernetes automatic deployment and scheduling, rapid capacity expansion, and fault self-healing, and removes and shields complexities such as Kubernetes architecture, configuration, and networking.
Therefore, Cube can basically be understood as a fool-like platform, and applications can be deployed with just a few clicks of the mouse. As Zhang Pengbo said, through Cube, users only need to provide packaged container images to quickly deploy containerized applications in batches, without the need to purchase cloud hosts or UK8S clusters in advance. In other words, Cube can bring great improvements both in terms of technology and cost.
Specifically, firecracker lightweight virtualization, cri-o + firecracker-containerd container management service, and Kubernetes basic scheduling framework are the basic components of Cube. Of course, every part of UCloud has been optimized accordingly.
For example, at the virtualization layer, UCloud has fully streamlined and optimized the kernel/rootfs/init process of firecracker, and only retained the most basic functions to speed up the startup speed, reduce the security attack surface, and reduce resource consumption. In addition, UCloud also has built-in infra container in firecracker, so that when Cube runs as a pod, there is no need to mount additional infra containers.
In the container management layer, UCloud has modified the cri-o management container group architecture and adopted a single pod corresponding to a single shim model, which can significantly reduce shim resource consumption and simplify container management.
Scheduling layer, for Kubernetes, on the control plane, UCloud uses a custom scheduler, which can better meet the needs of task priority, scheduling speed, and resource management in multi-tenant scenarios. On the host node, in view of the characteristics of Cube operation, UCloud has streamlined some functions that do not need to be implemented by kubelet, such as mounting the configmap/volume directory on the host, running the cni plug-in, collecting specific directory logs, etc., enhancing the relationship between the container and the host Isolation security.
From the above content, it is not difficult to see that Cube is not a product that is simply modified based on open source software, but a product that UCloud really starts from the user's demand side and polishes it a little bit. In fact, this is also a decisive factor for UCloud's success. Focusing on customers, focusing on product innovation, reviewing the development history of UCloud, there are countless examples such as the safe house is another typical.
Comprehensive innovation makes Cube have distinctive features, such as:
Free operation and maintenance. Without the need for a cloud host, the maintenance process is naturally omitted, and users are truly application-centric.
No operating system occupation. There is no need for a basic operating environment, so that as many resources can be applied for, as many resources can be used by the container, and can be used up, paying on demand, minimizing waste.
Smaller specifications. The container can be created small enough, and the smallest Cube specification can achieve 0.1 core 128Mi, giving users more fine-grained business choices.
Strong isolation at the virtual machine level. Firecracker's lightweight virtualization technology can isolate each container group. Note that it is not process-level isolation, which improves the security of container operation.
Pay by the second. Cube realizes that users only need to pay for the life cycle of container operation.
There are many other features of Cube worth mentioning. For example, each Cube instance has independent internal and external IPs. After the Cube instance is restarted, the internal and external IPs remain unchanged and can be used as the back-end service node of ULB. Exposure services to the outside world and provide stable and reliable services. For another example, Cube currently supports directly mounting UFS as a persistent storage when it is created, which is even better than cloud hosting in terms of convenience.
In summary, Cube has two keywords: simplify and save. The purpose is to lower the threshold, so that more users can use advanced technology such as containers and enjoy the dividends of the times.
03
Connecting the past, the great use of Cube
Finally, talk about the relationship between Cube and Kubernetes.
As mentioned earlier, the scheduling layer of Cube also uses Kubernetes, but some modifications have been made. Therefore, the relationship between Cube and Kubernetes is actually very simple. Cube can be connected to Kubernetes upward. Within UCloud, Cube is defined as a product that connects the past and the next.
On the bottom, Cube can be used by customers who are concerned about technology and expenditure. On the top, if the scale of customer container applications is gradually expanded, it can seamlessly connect with the more powerful Kubernetes system. In other words, UCloud provides customers with a layered construction path.
Because of this, in the first few months since UCloud launched the public beta of Cube in early July, it has attracted many container customers. Typical scenarios such as business elastic expansion and contraction, such as Dada, belong to this, and data collection and forwarding services. , Because Cube can be implemented in smaller specifications, which can help customers save costs.
Zhang Pengbo said that the scope of application of Cube is very wide. For head customers who have already used Kubernetes, Cube can be used as a resource pool for flexible expansion and contraction. For waist customers and small customers, Cube can lead them to get started and enjoy the bonus of container technology in advance.
Summarizing the full text, we not only see the ingenuity of UCloud products through Cube. Cube is more like a mirror through which you can understand a lot of UCloud. Being customer-centric, not only focusing on the needs of large customers, but also the pain points of a large number of small and medium customers, this is the key to UCloud's success. In the future, Cube will continue to move towards the goal of simplification and saving. UCloud will also adhere to its original intention to help customers go to the cloud and complete digital upgrades.