Elasticsearch:Features of the AliES Kernel-enhanced Edition

Enhancement category

Feature name

Feature description

Version 7.16

Version 7.10

Version 6.7

Performance enhancement

Analyticsearch retrieval and analysis

Improves query performance in log scenarios. It accelerates Kibana Discover queries and supports concurrent queries to significantly reduce query time.

For more information, see Use the analytic-search plug-in.

×

Supported by kernel version 1.7.0 and later

×

Physical replication plug-in

Provides an index-level switch for real-time incremental synchronization between primary and replica shards. This greatly reduces cluster CPU overhead and improves write performance by 60%.

For more information, see Use the physical replication feature of the apack plug-in.

×

√

Supported by kernel version 1.2.0 and later

Bulk write aggregation plug-in

Aggregates bulk write requests in batches based on a specified request size and time interval. This increases cluster write throughput by 20%.

For more information, see Use the bulk aggregation plug-in (faster-bulk).

√

√

√

Time-series query pruning

In time-series scenarios, queries crop data by time range. This improves the query performance for ranges on time-series fields by 30%.

For more information, see Use the time-series query pruning feature.

√

√

√

Stability improvement

QoS throttling plug-in

Implements index-level read/write throttling for the cluster. It downgrades indexes based on specified rules to control cluster read/write traffic and improve stability.

For more information, see Use the cluster throttling plug-in (aliyun-qos).

√

√

√

Cost optimization

Index compression plug-in

Supports multiple compression algorithms, such as Brotli and Zstd, to compress index files. This reduces the index storage size by more than 40%.

For more information, see:

Versions 7.16 and 7.10: Use the aliyun-codec plug-in

Version 6.7 (only for instances created before April 2024): Use the index compression plug-in (beta) (codec-compression)

√

√

√

Feature optimization

Timestream time-series plug-in

Supports creating, reading, updating, and deleting time-series indexes. It also supports downsample write queries and Prometheus Query Language (PromQL) queries. This greatly reduces the storage and usage costs for time-series data.

For more information, see Introduction to the TimeStream time-series enhancement engine.

Supported by kernel version 1.8.0 and later

Supported by kernel version 1.7.0 and later

×

Aliws tokenizer plug-in

Integrates an analyzer and a tokenizer from Alibaba DAMO Academy's NLP technology. It provides a more comprehensive tokenization dictionary for data retrieval and analysis.

For more information, see Use the AliNLP tokenizer plug-in (analysis-aliws).

√

√

√

Feature name

Feature description

Billing description

Indexing Service

Suitable for scenarios with high write transactions per second (TPS), large fluctuations in write traffic, and low queries per second (QPS), such as log retrieval and time-series metric analysis.

• Based on a read/write splitting architecture, it forwards data writes to a cloud service for index building. This offloads the write pressure from your cluster.

• Cloud write resources scale automatically to match resources with traffic dynamically. This provides over 10 times the write elasticity, easily handling write traffic fluctuations and peak write bottlenecks.

• You pay for writes based on actual traffic. You do not need to reserve resources for peak writes, and no resources are idle during off-peak hours. In high-write scenarios, cluster computing resource costs can be reduced by more than 50%.

For more information, see Introduction to Indexing Service.

You are charged for the actual write traffic.

Openstore intelligent hybrid storage

Suitable for scenarios that require long-term storage of massive data, have low data query QPS, and have a relatively high tolerance for query latency. Examples include log retrieval, metric analysis, and data archiving.

• Based on a storage-compute separation architecture, you do not need to plan and purchase cluster data storage space in advance. You are charged for the actual amount of data stored. The unit price for storage is 70% lower than for cloud disks, which greatly reduces cluster storage costs.

• It goes beyond traditional hot and cold data separation architectures by automatically tiering data based on query frequency. With hybrid storage, you do not need to configure index lifecycles.

• Multiple replicas share one copy of the data without increasing storage costs. This provides 99.9999999999% data high availability.

For more information, see OpenStore intelligent hybrid storage engine.

You are charged for the actual Openstore storage space.