On September 1, 2023, the official version of Alibaba Cloud Elasticsearch Serverless was launched. Based on cloud-native Serverless technology, it is committed to creating a lower-cost, elastic, flexible, open and compatible, and out-of-the-box Elasticsearch experience for users on the cloud.
Compared with the public beta version, the official version of Elasticsearch Serverless has higher compatibility with the native Elasticsearch read-write API and ecological components. It provides default Kibana visualization capabilities to help users quickly build data retrieval and analysis applications, and can be flexibly used in website search and enterprise information. In various application scenarios such as retrieval, log analysis, operation and maintenance monitoring, etc.
Background introduction
In the past ten years, under the deterministic trend of enterprises moving to the cloud, the computing power provided by many cloud computing vendors is still in the form of resources such as servers. The threshold for use is high and the problem of resource redundancy is widespread. Cloud computing needs a new form. , making computing power more inclusive and allowing more people to enjoy the technological dividends, thus gradually moving towards a new form - Serverless, making computing power as popular as electricity.
As an open source distributed retrieval and analysis engine, Elasticsearch can be used to search massive amounts of structured and unstructured data. It is widely used in scenarios such as information retrieval and analysis, log analysis, and operation and maintenance monitoring. Since 2017, Alibaba Cloud and Elastic have joined hands to launch the Elasticsearch version of Alibaba Cloud's retrieval and analysis service. The product has always adhered to the upgrade of cloud-native technical capabilities, providing users with an out-of-the-box, faster and better Elasticsearch cloud service experience.
When using the open source self-built and current fully managed PAAS version of Elasticsearch, users need to configure the CPU, memory, disk and other resources of the cluster according to the estimated read and write situation of the business, and reserve a certain water level to ensure the smooth operation of the business. , but still encounter the following problems:
- In peak business scenarios such as major promotions and releases, sudden read and write traffic occurs that cannot be planned in advance, causing cluster downtime and making it difficult to ensure business stability.
- Capacity planning is based on peak business traffic, but a large number of machines are redundant during off-peak periods, resulting in redundancy and waste of resource costs.
- In the context of enterprise IT cost control and overall cost reduction and efficiency improvement, we hope to further optimize cluster configuration and reduce budgets.
With the launch of Alibaba Cloud Elasticsearch Serverless, the above problems can be easily solved. The following first shows a practical case of using Serverless in a business scenario:
Business scene
A user uses the Elasticsearch Serverless service to provide search capabilities in a small website building scenario. The Serverless application experienced a business peak of 8 hours in one day. The writing and query computing resources each used 0.6 CU per hour. During the rest of the time, the writing and query computing resources each used 0.1 CU per hour. The storage space usage was always is 1 GB.
Cost calculation
Total cost per day = writing computing resource cost + query computing resource cost + storage space cost = 2.544 + 2.544 + 0.0504 = 5.1384 yuan
- Writing computing resource cost: 0.3975 yuan/CU/hour x0.6 CU/hour x8 hours + 0.3975 yuan/CU/hour x0.1 CU/hour x16 hours = 2.544 yuan/day
- Query computing resource cost: 0.3975 yuan/CU/hour x 0.6 CU/hour x 8 hours + 0.3975 yuan/CU/hour x 0.1 CU/hour x 16 hours = 2.544 yuan/day
- Storage space cost: 0.0021 yuan/GB/hour x 1 GB/hour x 24 hours = 0.0504 yuan/day
cost comparison
In the above scenario, the Serverless application has completed the tasks that the ECS self-built Elasticsearch 2-core 4GiB single-node instance can complete. Compared with the price of the corresponding specification of 0.407698/hour - the required daily cost of 9.785 yuan, Serverless has achieved nearly 50% of the cost. reduce.
How to achieve ultra-high cost performance?
Alibaba Cloud Elasticsearch Serverless can bring such obvious cost optimization, mainly because it realizes on-demand access to resources, so that each application resource can dynamically expand and contract according to the business load and data volume, and realizes the actual resource usage according to the actual resource usage. paid experience. Compared with the payment method of self-built Elasticsearch with fixed machine specifications, the usage cost can be significantly reduced. In scenarios with large business fluctuations, the resource usage and specification changes of open source self-built Elasticsearch and Alibaba Cloud Elasticsearch Serverless are shown in the figure below.
As can be seen from the figure, self-built Elasticsearch has a lot of resource redundancy and waste during low business peak periods. During peak business periods, insufficient resources may affect business stability. While costs are significantly optimized, Alibaba Cloud's core design concepts of Elasticsearch Serverless are reflected in the following aspects:
- Lower costs : Break the fixed resource payment model and only pay according to usage for the actually occupied resources. It truly realizes pay-as-you-go dynamic matching of load and resources and reduces the redundant cost of idle resources.
- Elastic scaling : Supports automatic second-level elastic scaling within a certain range based on the resources required for business reading and writing. There is no need to manually expand or shrink capacity. In the face of business peaks and valleys, it can effectively reduce operation and maintenance costs and improve business stability.
- Open source compatibility : Supports the connection between Elasticsearch native API and Kibana and other ecological components, retains users' original usage habits, supports rapid migration to the Serverless version, and helps businesses quickly migrate to the cloud.
- Operation and maintenance-free : Users do not need to pay attention to operation and maintenance work such as underlying resource deployment and configuration, cluster expansion and contraction. Concepts such as clusters and nodes are shielded from users, and physical deployment logic is hidden, so that users are unaware, business is not affected, and services are continuously available.
Flexible application in multiple scenarios and multiple groups
- Low-frequency usage scenarios such as development and testing environments.
- Educational scenarios such as individual developer users, school teaching, and student experiments.
- SaaS application scenarios such as website building for small and medium-sized enterprises.
- Businesses have access peaks and troughs, unpredictable fluctuations, and uncertain load scenarios.
- Fully managed or enterprise users who want to be completely free of operation and maintenance.
Welcome to experience
You can enter https://elasticsearch.console.aliyun.com/serverless/cn-hangzhou/introduce to activate the service. You are also welcome to search the DingTalk group number 11205017670 to enter the Alibaba Cloud Elasticsearch Serverless user exchange group to speak freely, and there are also occasional surprise gifts. !