The New Rogue Database Platforms
It would appear that my editorial on “nobody ever got fired for going with Big Blue” has struck a chord with many of you. Today we have three people who have sent links or written blogs on the topic.
One thing of which I continue to be reminded is the strength of the PosgreSQL platform, especially running on Linux. Although PosgreSQL and MySQL are confinuing to grow in popularity as the platform with the best TCO in the relational database space, they are not new products. They are both quite mature with a large following.
Here are some comments from our readers.
Bastien:
MySQL does support partitioning as of 5.1.6
http://dev.mysql.com/doc/refman/5.1/en/partitioning.html
It capabilities are actually very good. And let’s not forget Postgre SQL, it also supports partitioning, is open sourced and has similar performance to MySQL and is often referred to as the open source of Oracle.
http://www.wikivs.com/wiki/MySQL_vs_PostgreSQL
Jim:
I have been around the block for awhile (34+ years of IT). I started as a programmer on “Big Iron”. The changes, while dramatic, almost seem to come full circle. Distributed processing was supposed to supplant centralized processing, but now consolidation, VM, and “dumb” end-user devices are the focus.
When the IT world was bitten by the distributed bug, the dream was to exploit the low utilization of all the PCs on the network – unlimited MIPs and the ultimate of parallel processing. While the laptop/desktop world is much more standardized, the performance and availability demands from a suite of disparate applications today demand ruthless standardization (so much for the “open” hardware and software concept). At a previous company I worked at, we had to delay for months a critical upgrade in our Unix environment because we couldn’t find the root cause of a calamitous performance problem when testing the upgraded HW/SW. The problem ended up being a mismatch in the software for failover capability. While this happened in a shop which employed a high degree of standardization, just with four or five vendors they can pass the buck. The watchword for cloud implementations is to be careful of this. If the vendor isn’t practicing ruthless standardization, you might want to dust off your SLA and find out what the penalties are for loss of service.
Adding more processors also doesn’t guarantee a linear growth in performance. The mainframe discovered this long ago. The workloads for most operational applications cannot scale (Google is an exception since it is a high transaction, read-only, and small I/O model). Adding more processors just adds more memory management overhead and you will find the knee in the performance curve very quickly. The key is to know what application workloads can scale and which cannot. You will probably be surprised that most do not scale that well.
Duke:
It is were "just" the DBMS that mattered (and I’m almost enough of a database bigot to believe so 🙂 then postgreSQL on Linux should rule. I think MS’ play is the integrated platform for BI: http://www.informationweek.com/software/information-management/microsoft-sql-server-2012-vs-oracle-cust/232800476?pgno=2
I’ve blogged on this topic http://it.toolbox.com/blogs/data-ruminations/oracle-dbms-where-market-share-is-weakest-53301
Thanks for the great input.
Cheers,
Ben
$$SWYNK$$
Featured Article(s)
Troubleshooting problems with SQL Server 2008 R2 System Stored Procedures
In this article, Alexander Chigrik explains some problems that you can have when you use SQL Server 2008 R2 system stored procedures.
Featured White Paper(s)
Encryption & Key Management for Microsoft SQL Server 2008/2012
Simplify encryption and key management on your SQL Server. Data thieves are targeting SMB companies because of their inadequa… (read more)