Now Playing Tracks

主流列式数据库评测Ingres VectorWise

 【IT168 专稿】在上一系列文章中,我们先后介绍了Gbase 8a,Infobright,InfiniDB, MonetDB,SybaseIQ等5种列式数据库的功能,利用tpc-h模型scala=1和10的数据作了测试,并和传统行存储的Oracle数据库作了对比,得到了一些结论。本文将介绍另一种重要的新型列存储数据库Ingres VectorWise的功能特点,同样采用tpc-h scala=10数据评估它的性能,为用户数据库选型提供参考。

  我们先

  Ingres是关系数据库软件中的元老级产品,70 年代开始于加利福尼亚大学伯克利分校的一个研究项目,它的代码使用BSD许可证。从 80 年代中期起,在Ingres 基础上产生了很多商业数据库软件,包括 Sybase、Microsoft SQL Server、Informix 和许多其他的系统。在 80 年代中期启动的后继项目 Postgres,产生了 PostgreSQL,在开源数据库市场份额仅次于MySQL。由于Ingres有它自己的查询语言Quel,导致它对SQL语言的支持比较晚,从而慢慢淡出了主流用户的视线,但Ingres仍在继续发展,先后发布了Ingres 2006等产品,目前最新版是10.0。Ingres公司和荷兰的CWI研究院合作推出的VectorWise是基于Ingres10.0和MonetDB/X100这2种各自领域的领先产品,可以说是系出名门。2010年6月发布1.0版,2011年3月31日,发布了获取了tpc-h的100GB数据量的第一排名的1.5版。详细信息见http://www.tpc.org/tpch/results/tpch_perf_results.asp。本文即针对VectorWise 1.5版展开测试。 

       系列文章回顾:

四款主流列式数据库横评

        主流列式数据库评测:Sybase IQ

       主流列式数据库评测:InfiniDB和MonetDB

       主流列式数据库评测之Infobright

       主流列式数据库评测:南大通用GBase 8a

  一、测试平台

       本次测试基于Intel Xeon 7550*8的PC服务器上用VMWare VSphere 4.1管理的虚拟机,虚拟机的逻辑CPU个数是8,内存100GB,存储为8个300GB SAS本地磁盘,采用一块512M缓存RAID卡,按RAID5方式组成磁盘阵列操作系统采用和RHEL 5相同的核心级别的RedFlag Asian Linux Sever 3.0 x64。

  二、安装

  (一)VectorWise的安装步骤

        VectorWise的评估版安装包可从网站http://www.ingres.com/downloads/vectorwise获取,Linux x86-64平台同时提供了rpm和非rpm二种安装包,前者简便易行,适合于初学者,后者提供更多的定制选项,适合于熟练用户。下面分别介绍这2种安装包的安装步骤。rpm安装包文件名是ingresvw-1.5-141-NPTL-eval-linux-x86_64.tgz,大约38MB,包含了服务器端核心和基于字符终端的命令行工具。这里NPTL 是Native POSIX ThreadingLibrary 的缩写。

         http://esd.ingres.com/上提供了详细的用户手册,内容包括概念、安装、管理和维护、基准测试、功能特性和系统限制等方面,基本涵盖了用户日常使用的各方面。至于更详细的SQL命令和网络配置步骤,需要参考Ingres 10.0的相关文档。

  1. rpm安装

contact vectorwise research publications talks varia 

Contact

To contact me, please write to marcin at vectorwise dot com.

VectorWise

In 2008 the MonetDB/X100 project described further on this page has been converted into VectorWise, a company which I currently manage.

VectorWise cooperates with Ingres, the leading open-source database company, on the Ingres VectorWise product.

Research

Between 2001 and 2008 I was a member of the MonetDB development team. My initial work with MonetDB concentrated on parallel query execution, which resulted in my master’s thesis (pdf).

From 2003 I was working on the MonetDB/X100 project under the supervision of Peter Boncz and Martin Kersten.

This project resulted in my PhD thesis and was a base for the VectorWise company

System Error 127 = The Specified Procedure Could not Be Found

When adding an ODBC data source for a new vnode and selecting the Database, it gives an error: 
Connection failed.
Specified driver could not be loaded due to system error 127(Ingres).

This can occur with Hummingbird BI(Business Intelligence) Query, but the problem is not specific to that.

Vectorwise represents a step-change in database technology. By unlocking the capabilities of the modern chip, our customers and partners report 10x–70x better performance benefits for analytic workloads. Those queries which currently take you minutes can be cut to seconds. The answers you’ve been told are too expensive to compute can be delivered using commodity servers and don’t require expert tuning. Vectorwise can power your business in four easy steps: install, load data, gather statistics and start running queries. This skips over complex, time-consuming steps such as complex schema, data load design and expert tuning that you would normally expect to take hours or even days.

To Tumblr, Love Pixel Union