Amazon OpenSSearch Service provides automated hourly images as a critical backup and recovery mechanism for customer data. These images serve as a point-in-time backup that you can use to restore your OpenSearch domains to previous state, helping to ensure the durability of data and continuity of business. Although this function is necessary, it is equally important that the image process works smoothly without basic domain operations. The workflow of images must be sufficiently efficient in order for the main performance of searching and indexing operations to represent the ability to scalance with increasing workload and support the overall stability of the cluster.
In this blog post, we will tell you how we have improved the efficiency of the Amazon OpenSsearch service and carefully held these critical operational aspects. These Snapshot optimizations are allowed for all OpenSearch optimized instances of instances (or1, or2, OM2) from version 2.17.
Background
In the traditional Snapshot of OpenSSearch mechanism, it includes a process of recording files of the additional segment from each Shard to Amazon Simple Storage Service (Amazon S3). The workflow begins when the initiation of the cluster manager for creating and coordinates with nodes holding primary shards to capture their appropriate images. During this process, the data nodes continually communicate with the cluster manager to report their progress. To ensure resistance to the leader failure, the cluster condition maintains permanent monitoring of all ongoing images. This condition is shared with all data nodes. However, this approach introduces meaningful overhead costs of communication, especially with extensive deployments.
Connect the cluster with nodes of primary shards. Each snapshot operation requires at least the N Cluster status update, while the shipping call flows into the AZ cluster node to the manager to the data nodes (including the update of one cluster status for each primary share and transport calls for each update) as shown in the following diagnostic diagram. In large domains with hundreds of nodes and thousands of shards, this intensive communication Pattenn can potential overwhelming cluster manager, which affects its ability to manage further tasks of critical cluster management.
OpenSSearch optimized instance family has introduced significant progress in data durability and image efficiency. OpenSearch OpenSearch Oppensearch Optimized Instances, built to supply a high Thrighput with 11 nine dubility, keep a copy of all indexed data in Amazon S3. This architectural design eliminated the need to reload the data while creating images. Instead, the system refers to an existing data checkpoint in the frame metadata. The data checkpoints monitor the data status on the shards at the moment to help ensure consistency and durability. We also prevent data cleaning from Amazon S3, which refers to the metadata of the image. Thanks to this approach, images were significantly lighter and faster than the conventional method.
Improved OpenSSearch Optimized Instances, also called shallow V1 image, referring to the checkpoint by creating an explicit lock files for each share checkpoint. This flow is illustrated in the following diagnosis diagram, where in the fourth step, instead of uploading segment data, we upload a set of checkpoint lock.
While this approach has successfully solved the data redundancy problem by replacing the recording of segment data on the creation of a checkpoint lock, it introduced its own challenge set. Communication overhead costs between the nodes remained unchanged during the operations and deletion of films. In addition, the system creates blocking files for each share in each frame, regardless of whether SHARD gets active operation or not. This design option generated an excessive number of remote shop calls to form a set of lock on SHARD during an image operation, which is particularly problematic for a larger opensearch domain.
Revised shallow image (v2)
At its core, the shallow Snapshot V2 reimaginates how we process data backup in OpenSearch. The shallow image of the V2 occupies a more intelligent approach of implementing a system that refers to a timing system that reduces data duplication and at the same time eliminates the direction of communication. In the shallow image of V2, as shown in the following diagram, the interaction of an explicit lock into the SHARD remote warehouse control point set is inserted by an implicit lock based on the frame and checkpoint set. We follow these time stamps in Timetamp files and record them to a remote store. With this implicit lock, there are no checkpoints that match the time stamps in Timetamp files, cleaned from the Amazon S3. With this architectural change, Don Don’s data nodes must send SHARD updates to the cluster manager and avoid the following cluster status updates. The frame recovery process works by reading a Timetamp file that corresponds to your image that helps search the data node and download the correct version of the Amazon S3 data.
Key benefits
Let’s explore the main advantages of using a shallow V2 image.
Performance of improvement
The performance benefits of the shallow V2 image are considerable and versatile. Minimizing the amount of data to be uploaded to a remote store, and the number of cluster status updates that need to be communicated between the nodes of bold images, the system significantly reduces I/O and network operations. This translates shortening to faster images and lower use of system resources during backup operations.
The evaluation listed in the following table, which we have done to assess the effect on images when the domain experiences a significant load.
| Domain configuration | The time to create a picture | |||
| Number of nodes | The number of shards | Traditional | Shallow picture V1 | A shallow picture v2 |
| 10 | 100 | 15-20 minutes | 1-2 minutes | <1 second |
| 10 | 10,000 | 30-40 minutes | 5-10 minutes | <5 seconds |
| 100 | 100,000 | > 1 hour | > 1 hour | <10 seconds |
Scalabibility
With a fixed number of communication calls between the node during images, the creation of the image is a single -digit second and the node, index and Shard Counte grow. When it was tested on 1,000 nodes in the Amazon OpenSsearch Service, a shallow time was observed between 10-20 seconds. For organizations managing large Amazon OpenSearch service domains, the shallow V2 image offers special benefits. Decreased storage costs from shallow image and faster images from shallow V2 frames allow more frequent backups without huge storage sources or impact system performance.
Architectural simplification
Architectural improvements in the shallow image V2 exceed optimization. The new implementation contains a more efficient and maintainable code base, which reduces the need for tuning and suggests improvements. Simplified architecture reduces the complexity of the image and restore process, leading to more reliable operations and less possible point of failure in use that require frequent backups such as compliance -based scenarios or slight development levels. This means that you can set a low restoration target to recover after a disaster. Effective processing of additional changes in Snapshot V2 allows more detailed backup plans without punishment.
Storage
The cornerstone of the shallow film V2 is its innovative approach to storage management. Instead of creating multiple copies of unchanged data intelligent links to existing data blocks. This implicit mechanism of reference counting based on the time limit prevents the creation of explicit locks on the shard. In S. Svinds, where there are drops for bonus, the storage can store a shallow V2 image to save costs. Reference -based approach helps to ensure optimal use of available storage space while detaining an understanding of backup coverage.
We look forward to seeing ahead
The introduction of a shallow image V2 means the beginning of our journey to more efficient data backup solutions. Based on the frames created by the shallow image of the V2, we can implement other features such as time restoration (Pitr), better cluster status integration and various performance optimization.
Conclusion
The shallow V2 is a significant progress in Opensearch’s backup capabilities. By combining efficiency, improved performance and architectural simplification, it provides robust solutions for modern data backup challenges. If you use the instance type from an optimized instance family, a shallow V2 is already allowed for you. Whether you are using a large domain or working in storage restrictions, shallow snapshot v2 tangible slaves for your Amazon OpenSearch Service domains.
About the authors
Sachin Kale He is a senior engineer for the development of AWS software who works on OpenSearch.
Bukhtawar Khan is the main engineer working on Amazon OpenSearch Service. He is interested in building distributed and caromaous systems. He is an administrator and active contributor to OpenSearch.
Gaurav Bafna is a senior software engineer working on OpenSearch at Amazon Web Services. It is fascinated by the solution of problems in distributed systems. He is an administrator and active contributor to OpenSearch.