[OSGeo-Discuss] Enterprise OSDB for OSGIS

Paul Ramsey pramsey at refractions.net
Mon Dec 11 21:00:08 EST 2006


Best be VERRRRRY careful about doing the best tuning possible when  
carrying out these tests, or the results you get back could be  
meaningless.  Different workloads with bad / good tunings can yield  
order-of-magnitude differences in performance.

P

On 11-Dec-06, at 5:41 PM, <chenrg at lreis.ac.cn> <chenrg at lreis.ac.cn>  
wrote:

> Thank you very much, Frank,
> PostGIS is very promising. We've tested it with ArcSDE and Oracle  
> Spatial. Its spatial query performance is equivalent to ArcSDE, and  
> much better than Oracle Spatial. Its spatial processing (such as  
> intersection, union) performance is better than Oracle Spatial, but  
> has a big gap comparing to ArcSDE due to the GEOS's poor  
> performance (The test showed that GRASS is very good at spatial  
> processing).  We're considering to propose a new benchmark for  
> spatial DBMS (The Sequoia 2000 and Paradise benchmark are quite old  
> ones). The effectiveness of the benchmark will be demonstrated  
> using a variety of spatial queries over a 10-100GB spatial data in  
> five example DBMSs including the commercial ones such as Oracle  
> Spatial, IBM DB2 Spatial extender and ESRI ArcSDE and the open  
> source ones such as PostGIS/PostgreSQL and MySQL Spatial  
> extensions. And we have plan to do some DBMS benchmark (TPC-C and  
> AS3AP) tests. The targets will be Oracle, Ingres, PostgreSQL,  
> MySQL, MaxDB and Firebird in Linux. The tools'll include Benchmark  
> Factory for Databases from Quest(www.quest.com) and the ODBC Driver  
> from OpenLink (www.openlinksw.com). If PostgreSQL + PostGIS  
> performs well, we'll carry through our GRIDGIS project based on it.
>
> Best regards,
>
> Chen
>
>
> ----- Original Message -----
> From: "Frank Warmerdam" <warmerdam at pobox.com>
> To: <discuss at mail.osgeo.org>
> Cc: <chenrg at lreis.ac.cn>
> Sent: Monday, December 11, 2006 11:26 PM
> Subject: Re: [OSGeo-Discuss] Enterprise OSDB for OSGIS
>
>
>> chenrg at lreis.ac.cn wrote:
>>> Dear all,
>>> There are many comparisons about OSDBs, such as:
>>> http://www.geocities.com/mailsoftware42/db/
>>> http://www.fabalabs.org/research/papers/FabalabsResearchPaper- 
>>> OSDBMS-Eval.pdf
>>> http://www.virtuas.com/files/osl-osrdb-01.pdf
>>> http://www.osdbmigration.org:8080/osdb/osdb-features
>>> I'm not sure which will be the most promising enterprise OSDB for  
>>> OSGIS.
>>> (1) PostGIS is an excellent one, but its performance depends on  
>>> PostgreSQL;
>>> (2) MySQL Spatial Extension (MyGIS) faces the same problem.
>>> Another solution is to build a Spatial Data Engine (like ArcSDE) for
>>> FireBird or MaxDB or Ingres.
>>> Checked the source codes of several OSDB:
>> ...
>>> It seems that Ingres is more powerful and has more enterprise  
>>> functions.
>>> Further more, it has internal support for spatial extension.
>>> Is it a reasonable solution to choose it to build enterprise  
>>> OSGIS? Any
>>> advice and suggestions?
>>
>> Prof. Chen,
>>
>> I reviewed the helpful document you referenced at:
>>
>>   http://www.osdbmigration.org:8080/osdb/osdb-features
>>
>> And from this it did not seem clear that Ingres was substantially  
>> more
>> powerful than PostgreSQL.  In that matrix it seems that PostgreSQL
>> compared fairly well on the various enterprise features listed.  I'm
>> afraid I did not have time to review all the other documents.
>>
>> You mention that PostGIS is excellent, but that its performance
>> depends on PostgreSQL.  Is that a problem?  It has been my (limited)
>> experience that a well tuned postgres performs well compared to other
>> enterprise class commercial databases, though it sometimes performs
>> less well than very performance focused databases like MySQL.
>>
>> MySQL is a promising player in the geospatial osdb space, and I was
>> pleased to see their addition of limited OGC simple features support.
>> But, in my opinion, MySQL is a stretch to refer to as "enterprise
>> class" in the terms the document above lists.  Also, it's spatial
>> support is quite limited by comparison to PostGIS.
>>
>> I am not really familiar with the other databases listed.
>>
>> My main point though is that usefulness and success in the open  
>> source
>> community is in significant part about the healthiness of the  
>> ecosystem
>> around a project.  That is, the number of users and contributors  
>> and the
>> number of people knowledgeable about the product.  Also, the  
>> helpfulness
>> of on line resources such as mailing lists, irc channels, and web  
>> sites.
>>
>> On that basis it seems to me that the only spatial open source  
>> database
>> with any significant mind-share and community is PostgreSQL 
>> +PostGIS.  I
>> am speaking from the spatial point of view of course.  I'm sure  
>> all the
>> databases have substantial users bases.  In the case of MySQL it also
>> has a substantial number of users in the geospatial space, though few
>> of those users are actually using the spatial extensions to mysql
>> (in my experience).
>>
>> So, if I were wanting to build out an enterprise class open source
>> spatial database system for an enterprise, I think I would start
>> by looking at PostgreSQL + PostGIS and only look further if I found
>> substantial inadequacies for my needs.
>>
>> Best regards,
>> -- 
>> --------------------------------------- 
>> +--------------------------------------
>> I set the clouds in motion - turn up   | Frank Warmerdam,  
>> warmerdam at pobox.com
>> light and sound - activate the windows | http://pobox.com/~warmerdam
>> and watch the world go round - Rush    | President OSGeo, http:// 
>> osgeo.org
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: discuss-unsubscribe at mail.osgeo.org
>> For additional commands, e-mail: discuss-help at mail.osgeo.org
>>





More information about the Mail_discuss mailing list