No Matter Any Data,

No Matter How Big !

- GoldiLocks Cluster -

 Goldilocks Cluster Overview

현대 사회의 산업 환경이 IoT, Cloud, Big Data, Mobile로 급격히 전환되면서 데이타가 폭증하고 있습니다. 국제적인 시장 조사기관 IDC에 따르면, 2020년까지 데이타는 44ZB(64GB USB 약 1조 개)만큼  폭증할 것이라고 합니다. 또한, 산업이 진화함에 따라 실시간 데이터 처리와 분석에 대한 요구는 급증하고 있습니다. 

이러한 문제는  2000년대 중반부터 대두되어, Big Volume, Horizontal scalability를 지원하는 NoSQL 기술이 등장했습니다. 그러나, NoSQL은 데이터 정합성을 보장하지 못하고, SQL이 지원되지 않아 기업 내부에서 중요한 정보를 담아 분석하기에는 한계가 있었습니다. 그렇다면 ZB시대로 표현되는 IT패러다임 변화에 데이터 처리를 어떻게 해야할까요? 기존의 정합성이 보장되는 관계형 데이터베이스의 장점과 폭증하는 데이터도 수평적 확장으로 데이터 처리가 가능한 기술이 바로 NewSQL입니다. 

| IT패러다임 변화에 따른 데이터 처리 요구사항 |

  • 폭증하는 데이터를 실시간을 처리할 수 있는가?

  • 24X365 서비스를 위한 가용성을 제공하는가?

  • 표준 SQL지원으로 개발의 편의성, 생산성을 보장 하는가?

  • ​안정적인 운영환경을 지원하는가?

Old SQL, NoSQL and New SQL

NewSQL은 데이터와 트랜잭션의 폭증에 대응할 수 있도록 쉽게 확장할 수 있는 아키덱처야 합니다. 또한 전통적인 RDBMS처럼  SQL을 지원함과 동시에 트랜잭션이 안전하게 처리되는 것을 보장하는 ACID(원자성(Atomicity), 일관성(Consistency), 고립성(Isolation), 지속성(Durability)을 보장해야 합니다. 궁극적으로는 비즈니스 연속성을 보장할 수 있도록 24x365 의 가용성을 제공해야 합니다.

Goldilocks Cluster는 폭증하는 데이터 처리를 Scale Out Cluster로 처리하는 NewSQL DBMS입니다. SQL과 트랜잭션을 지원하면서도 유연한 Scale Out이 가능합니다. 기존의 관계형 데이터베이스는 SQL, ACID, 가용성은 제공했지만, HW 확장도 쉽지 않았으며, 확장을 했다고 하더라도 이에 비례한 선형적인 성능확장은 불가능했습니다.

​​

Goldilocks Cluster는 폭증하는 데이터 처리를 Scale Out Cluster로 처리하는 NewSQL DBMS입니다. SQL과 Transaction을 지원하면서도 유연한 Scale Out이 가능합니다. HW의 추가로 선형적인 성능이 확장이 가능하여 실시간 고속 트랜잭션 처리 및 분석처리에 최적화 되었으며, NoSQL과는 달리 SQL과 ACID를 지원합니다.

Goldilocks Cluster는  변화하는 IT서비스 환경에 대응할 수 있는 유일한 New SQL입니다.

​귀사의 애플리케이션에 날개가 필요하십니까? 

 Goldilocks Cluster Features

Architecture

-In-memory based Scale Out Cluster -

 Golidlocks Cluster Features 

Goldilock Cluster is DBMS that allows expansion of capacity and performance of data base server without interrupting your business service with specialized sharding technology. With additional server, improvement of total through put is achieved and is optimized to process increasing data and transaction.

  • Database server capacity / performance expansion without interruption of business

  • Add Group/member, Delete Group/ member

  • Auto/Manual Data rebalancing

  • Distributed database design through horizontal scale out method

  • Data join and processing of global transaction supported through database sharding

  • Redistribution of data allows scale up and scale out without interruption in data processing

  • With the increase of cluster node, linear performance enhancement is achieved

  • All master, All Active Node's Architecture

  • I/O overhead optimization(minimization of I/O overhead) , without async processing

Inherit RDBMS advantages

- ACID, Global Transaction, Legacy SQL  -

Goldilocks Cluster provides NoSQL's horizontal scale out with sharding technology , keeping all advantages of RDBMS, allowing easy implementation on various applications. Hence allowing identical method of data processing as existing RDMS, in the work environment of persistent increase of data and transaction.

  • ACID in Shared nothing Cluster         

        - 2 phase commit protocol

       - Global statement level consistency

  • ACID Compliant with MVCC

  • Networked DML

  • With SQL Shipping, Parallel processing supported

  • No I/O overhead of multi-node storage system

  • ​Data Joining process among multi-nodesScale up & out 가능 

  • When a problem occurs in the node during the data processing, another node within the group is used (should an error occur on the primary node; provides a node within the group)

  • Node Failure

        - No Single point Failure

        - Connection Fail-over

        - DML Fail-over

Do you need wings for your application?

 Goldilocks Cluster Performance

Goldilocks Cluster allows performance enhancement in a linear format with additional nodes. All Master, All Active Node's Architecture increases available nodes improving the performance. Optimized I/O overhead minimizes performance loss in input and output process, allowing linear performance improvement.

HW SPEC:  Intel Xeon E5 2697v3 2.6GHz 14C * 2EA / 128G mem

DATA :       3,600,000 records, 200Byte/record,   Clients : 32, 64, 96 ea

Partition:   Range Partition/Sharding

 

Goldilocks Cluster References

National Police Security Dream

(Support System for the weak in the society)

OVERVIEW

For the weak in the society (kids, seniors, disabled), simultaneous finger print verification and search among large amount of data(big data searching) within Police support system became possible through Goldilocks Cluster.

National Police Team has merged Lost-kid-center with Emergency Operation Center for Woman, into a combined service called Dream Safety for prompt rescue and reporting violent assault to the ever increasing crimes to the weak of the society. Trying to increase speed in finding lost child and lost person through connection with national finger print search system.

CUSTOMER

PAINPOINT

​ AS-IS 

  • Demand for approximately 300 million cases/ day data processing : In connection with national finger print identification system , finger print image data of approximately 300 million cases needs matching process.

  • Demand for fast responsive speed: various accidents and criminal cases needs to be checked as soon as possible to reduce the occurrence rate, therefore demanding high performance data processing

  • Distributed data processing environment : Demand for optimized data processing environment with distributed data system to efficiently process high volume of data

 Adopt Goldilocks Cluster 

  • Optimized data separation environment : Goldilocks™ Cluster  has achieved improved data separated environment guaranteeing fast responsive speed and stable performance

  • Data mismatch prevention(avoid) function: Goldilocks™ Cluster's separated data environment runs on each active node preventing mismatch of data

  • Large Data processing : provides performance adequate to process approximately 300 million cases matching with transaction

SOLUTION 

​RESULT

  • Approximately 300 million cases of data processing a day : large amount of data transaction is processed through stable performance, peak time security is guaranteed

  • Decrease in delay time : reduction in memory through efficient data compression, reduction in cost resulting from work delay through fast search and response

  • Connection with other system : portability with applied SW in the system, smooth connection with other system

Do you need wings for your application?

© 2017 Copyright by Sunjesoft. All rights reserved

LINKS
ABOUT