Lindorm:Benefits

Feature

Lindorm

Apache HBase

Core features

Data model

Supports multiple data models, such as wide table, time series, search, and file. The wide table model supports multiple endpoints and APIs.

Only wide table

API access

Supports the HBase API and Phoenix SQL. Data is interoperable across different endpoints.

HBase API or Phoenix SQL

SQL

JDBC-compliant and compatible with Phoenix, offering superior stability and performance.

Requires external Phoenix support.

Data type

Supports a rich set of data types. For more information, see Data types.

Only supports byte[].

TTL

Provides enterprise-grade TTL at the table, row, and cell levels.

Supports table-level and cell-level TTL.

Strong consistency

Supports multiple consistency levels, including strong and eventual consistency.

Supported

Global secondary index

Provides a built-in global secondary index. This feature enables transparent queries, delivers high performance, and allows for on-demand redundancy of non-indexed columns.

Requires complex configuration of external components.

Multi-dimensional retrieval

Seamlessly integrates with LindormSearch to provide unified access for massive data storage, multi-dimensional queries, and full-text search. For more information, see Search index overview.

Not supported

Performance

Throughput

Delivers up to 7 times the throughput of a single Apache HBase node. For more information, see Test result analysis.

Not applicable

Request latency spikes

Reduces P99 latency to one-tenth that of Apache HBase. For more information, see Test result analysis.

Frequent latency spikes

Cost

Storage cost

Offers multiple storage types, such as Performance, Standard, and Capacity, reducing costs by up to 80% compared with self-managed instances on cloud disks.

Based on self-managed cloud or local disks, which are costly and inelastic.

Compute and storage separation

Yes. Storage and compute resources scale independently.

No

Data compression

Uses a deeply optimized built-in compression algorithm to achieve a compression ratio of over 10:1, which is more than 50% higher than Snappy.

Supports Snappy, LZ4, and LZO, but compression ratios are low.

Encoding

Employs adaptive, data type-aware encoding for a high compression ratio. This allows for fast searches without decoding.

Supports DIFF with moderate compression. Encoded data cannot be searched.

Hot and cold data separation

Automatically tiers data. Cold data is moved to cost-effective, high-compression storage to reduce costs by 80%, while improving access performance for hot data by 15%. For more information, see Hot and cold data separation.

Not supported

Scalability and elasticity

Minimum scale

Not applicable.

At least 3 nodes

Scalability

Highly scalable. Supports horizontal scaling to thousands of nodes.

Highly scalable. Supports horizontal scaling to thousands of nodes.

Reliability

Active-active redundancy

Provides advanced capabilities such as automated disaster recovery failover and concurrent requests across dual clusters. Supports building a hybrid primary/standby architecture with a self-managed Apache HBase cluster.

Not a productized feature. Failover is not supported.

Cross-AZ strong consistency

Supports cross-availability-zone (AZ) deployment, ensuring automatic recovery and strong data consistency in case of an AZ-level failure.

Not supported

Backup and restoration

Supports backup of datasets that are larger than 100 TB to OSS. Provides advanced features such as on-demand backup, point-in-time recovery (PITR), and a data size-independent Recovery Time Objective (RTO) of less than 30 minutes. For more information, see Enable backup and restoration.

Supported, but with limited capabilities.

Active geo-redundancy

Supported. Allows deployment across multiple geographic regions and units with on-demand data synchronization.

Not supported

Multi-tenancy and security

Authentication and ACL

Supports username and password authentication and ACLs. For more information, see Manage users.

Not supported

Resource isolation

Provides resource groups to enable physical resource isolation between tenants.

Not supported

Quota

Supports global quotas for tenants, including requests and storage.

Does not support multi-tenancy.

Encryption at rest

Supported. Keys are managed by KMS, and all data and logs are encrypted.

Supported, but with limited capabilities.

RPC blacklist

Supports an RPC blacklist to restrict specific calls.

Not supported

Auditing

Currently not supported.

Not supported

Advanced features

Table recycle bin

Lindorm moves deleted tables to a recycle bin, where they can be restored to prevent accidental data loss.

Not supported

Cascading split

Regions can be split successively without waiting for compaction to complete, significantly improving scalability and load balancing.

Not supported

Discrete TTL

Allows you to retain data from multiple non-contiguous time periods.

Not supported

Operations and diagnostics

O&M tools

Provides a GUI-based cluster management tool for managing tables, namespaces, groups, and ACLs. For more information, see Log on to the cluster management system.

HBase Shell

Data query

Supports interactive SQL queries in the GUI-based cluster management system. For more information, see Data Query. It also supports open source tools like HBase Shell and CQLsh.

HBase Shell

Ecosystem

Data migration

Supports online, cross-version, automated, and efficient migration from various versions of Apache HBase. The migration process has zero impact on your applications and requires no code changes. For more information, see Lindorm Tunnel Service.

Only offline migration is supported.

MySQL data synchronization

Supports full import and incremental synchronization of MySQL data to Lindorm by using Lindorm Tunnel Service.

Requires third-party tools. Does not support online incremental synchronization.

Spark analysis

Offers deep, productized integration. You can incrementally sync Lindorm data to Spark, analyze it with Spark SQL, and write the results back to Lindorm.

Not optimized. Data integration requires significant development effort.

MaxCompute

Provides productized integration to incrementally archive Lindorm data to MaxCompute.

Data integration requires significant development effort.

Log Service

Supports real-time data subscription from Log Service to Lindorm by using Lindorm Tunnel Service.

Data integration requires significant development effort.

Service and support

Availability SLA

Backed by an SLA. Provides 99.95% availability for a single-AZ instance and 99.975% for a multi-AZ instance.

Not provided

Operational cost

A fully managed service that eliminates the need for complex database operations.

High operational cost

Technical team

An expert team of Apache Project Management Committee (PMC) members and Committers provides technical support.

Not provided

Proven experience

Proven at scale with tens of thousands of deployed instances, supporting Alibaba's 11.11 Global Shopping Festival for nine years.

Not applicable

Feature

LindormTSDB

OpenTSDB

Operations and management

Service availability

99.9%

Requires you to build and manage clusters and dependencies to ensure availability.

Data reliability

99.9999%

Requires you to build and manage clusters and dependencies to ensure reliability.

Hardware and software investment

No hardware or software investment. Pay as you go.

Relatively high cost for database servers.

Maintenance cost

Managed service

Requires dedicated database administrators (DBAs), leading to high labor costs.

Deployment and scaling

Provides instant activation, rapid deployment, and elastic scaling.

Requires time-consuming hardware procurement, data center hosting, and machine deployment.

Dependencies

O&M-free

Depends on AsyncHBase and HBase, which leads to high operational costs.

Parameter tuning

Uses default parameters based on best practices.

Requires manual tuning of parameters such as SALT, connection count, synchronous flushing, and compaction.

Table creation statements

Table creation is managed by the service and transparent to users.

Requires O&M personnel to write static table creation statements.

Monitoring and alerting

Provides a complete, self-monitoring pipeline.

Requires external tools for setup.

Features

Data model

Supports both multi-value and single-value data models.

Supports only the single-value data model.

SDK

Java SDK

The open source SDK does not support queries.

Data type variety

Supports multiple data types, such as numeric, boolean, and string.

Supports only numeric types.

SQL query capability

Supports SQL for analytical queries.

Not supported

Chinese character support

Supports English and Chinese characters.

Supports only English characters.

Tag requirement

Tags are optional.

Tags are required.

Number of tag keys

Up to 16

Up to 8

Integration

Offers a rich ecosystem with seamless integration with Flink and IoT Platform.

As an open source product, it has limited integration capabilities with cloud services.

Storage cost

Data compression

Uses a specialized compression algorithm for time series data, achieving a high compression ratio.

Uses a general-purpose compression algorithm, resulting in a low compression ratio.

Stability

Data reads

Separates read and write thread pools for easy connection management and stable read/write performance.

Couples read and write operations, which can lead to connection exhaustion and a high rate of read/write failures.

Aggregator

Uses stream aggregation with fine-grained memory management for greater control.

Uses in-memory materialized aggregation, which can easily lead to out-of-memory (OOM) errors.

Feature

LindormSearch

Elasticsearch

Apache Solr

Core features

Data model

Supports multiple data models, such as wide table, time series, search, and file. The search engine can seamlessly serve as an index store for other engines.

Only search

Only search

API access

Supports Phoenix SQL and the Solr API.

ES API

Solr API

TTL

Provides enterprise-grade TTL at multiple granularities, such as table and row.

Only table-level TTL is supported.

Only table-level TTL is supported.

Unified storage and retrieval

Seamlessly integrates with LindormTable and LindormTSDB to provide unified multi-modal storage and retrieval.

Not applicable

Not applicable

Performance and cost

Throughput

Delivers 130% to 200% of the throughput of a single Apache Solr node.

Not applicable

Not applicable

Storage cost

Offers multiple storage types, such as Performance, Standard, and Capacity. Reduces storage costs by up to 80% compared to self-managed instances on cloud disks.

Based on self-managed cloud or local disks, which are costly and inelastic.

Based on self-managed cloud or local disks, which are costly and inelastic.

Compute and storage separation

Yes. Storage and compute resources scale independently.

No

No

Data compression

Uses a deeply optimized built-in compression algorithm to achieve a compression ratio of over 10:1, which is more than 50% higher than Snappy.

Not applicable

Not applicable

Hot and cold data separation

Automatically separates data into tables based on a time attribute. Cold data uses high-compression, cost-effective storage to reduce costs, while access performance for hot data is improved.

Not supported

Not supported

Elasticity

Storage elasticity

High. Decouples storage from compute and supports one-click scaling. Storage scaling takes effect in seconds, and compute scaling takes effect in minutes.

Low. Scaling out requires data migration and takes hours.

Low. Scaling out requires data migration and takes hours.

Single-writer, multiple-reader

Data shards support a single-writer, multiple-reader model. Read replicas can be scaled out horizontally online, with changes taking effect in seconds.

Supported, but adding read replicas requires data migration and takes hours.

Supported, but adding read replicas requires data migration and takes hours.

Ecosystem

Data migration

Supports online, automated, and efficient data migration from Apache Solr or Elasticsearch clusters to Lindorm with zero application impact or code changes. For more information, see Lindorm Tunnel Service.

Only offline migration is supported.

Only offline migration is supported.

MySQL data synchronization

Supports full import and incremental synchronization of MySQL data to Lindorm by using Lindorm Tunnel Service.

Requires third-party tools. Does not support online incremental synchronization.

Requires third-party tools. Does not support online incremental synchronization.

Spark analysis

Offers deep, productized integration. You can analyze Lindorm data with Spark SQL, incrementally sync Lindorm data to Spark, and write offline analysis results back to Lindorm.

Not optimized. Data integration requires significant development effort.

Not optimized. Data integration requires significant development effort.

Log Service

Supports real-time data subscription from Log Service to Lindorm by using Lindorm Tunnel Service.

Data integration requires significant development effort.

Data integration requires significant development effort.

Service and support

Availability SLA

Backed by an SLA. Provides 99.95% availability for a single-AZ instance and 99.975% for a multi-AZ instance.

Not provided

Not provided

Operational cost

A fully managed service that eliminates the need for complex database operations.

Not applicable

Not applicable

Technical team

An expert team of Apache PMC members and Committers provides technical support.

Not provided

Not provided

Proven experience

Proven at scale with tens of thousands of deployed instances, supporting Alibaba's 11.11 Global Shopping Festival for nine years.

Not applicable

Not applicable

Feature

LindormDFS

HDFS

Product positioning

Distributed file system

Distributed file system

HDFS compatibility

HDFS communication protocol

Supported

Supported

Basic read/write APIs

Fully supported

Fully supported

Advanced management APIs

Fully supported

Fully supported

Cost

Storage price (The actual price on the purchase page prevails.)

Starts at USD 0.019/GB/month

Starts at USD 0.023/GB/month

Storage elasticity

Supports smooth online scaling.

High entry threshold and large scaling increments.

Compute and storage separation

Supported. Decoupled from compute engines for independent scaling.

Not supported. Co-located with compute engines.

Tiered storage

Multi-tier storage with intelligent data tiering.

Not supported

Scalability

Number of nodes

Not applicable

0 to 1,000

Storage capacity

0 to 1 EB

0 to 10 PB

Number of files

Supports hundreds of billions of files.

Tens of millions

Ecosystem

Open source big data ecosystems like Hadoop and Spark, as well as the Alibaba Cloud data ecosystem.

Open source big data ecosystems like Hadoop and Spark.

Usability

LindormDFS is O&M-free and easy to maintain.

Stateful service that is complex to maintain.