Open Source Databases Won't Fly

"Though

my heart is with open source, my mind suggests that it is unlikely..."

It is with a mixture of anticipation, trepidation, and even a hope of being

proved wrong, that I write this article. Database technology, especially

relational database technology, with its full regalia of query and transaction

processing smarts, is a key to setting free the full potential of information

systems. Having indicated these mandatory cliched phrases, it is nonetheless

accurate to say that information storage, retrieval and management is the most

important aspect of the information revolution that we are living through.

However, the details–and recall where the devil is–are not easily grasped,

implemented or deployed.

Now certainly, you will not disagree that commercial databases have been

wildly successful and time-tested. And in doing so, various problems have been

recognized and solved or resolved (eg consider data independence). While open

source systems are far newer and lithe, one still encounters problems in using

them; eventually, they too will come to be similar in spirit to the commercial

efforts: old wine in new bottles, or vice versa? And then there is the issue of

installed base of legacy systems.

Significant is my lament as regards the lack of rigor in coding and

programming. While there are relatively few well-accepted coding standards, at

least the importance of issues such as careful documentation, checking for the

logic in the programs written, maintenance of the versions, all contribute to

errors.

Now, let’s discuss something I do know. And so, let us quickly deal with

the typical reasons why open source databases are regarded as being inadequate,

and with which I agree. The first is the issue of relatively weak query

processing support in such systems. Second, there is the larger issue of

transaction processing support. And transactions are admittedly more complex

than even query processing.

Even open source proponents believe commercial database systems will prevail.

I share the view that open source databases will co-exist with their commercial

counterparts. To assess the likelihood, consider the points that are said to

serve as the accelerators for open source systems...

First, systems that need to be reliable, scalable, and stable (i.e., that

would be every system from my perspective!), are expected to be amenable to open

source development. A second factor is that the applications be such that

verifying their correctness is difficult. The third is that in cases where

"control" over the software, presumably to manage and modify it, is

needed, open source systems are suitable.

A fourth factor suggests that where the common computing and communications

infrastructure is supported by software, open source systems will prevail.

Fifth, and finally, areas in which key methods are well known will fare

better with open source development. And if the reader is still paying

attention, he will realize that my argument has been that database technology is

quite otherwise.

Time for some reconciliation. Having said all of the above, let me also

mention that I believe open source database systems will serve to provide the

means to run pilot projects, teach people, as well as for relatively small

projects. They will be wonderful in these cases, where one would not want to use

a heavyweight commercial system: where a flyswatter would work well, a cannon

would be disastrous Besides, it would be vastly cheaper–it always boils down

to the money, I am afraid. And now, I must find myself a bomb shelter…

Prof Nandit Soparkar is a faculty member at the University of Michigan. His

research includes transaction processing, datamining and analyses technology and

logic-enhanced memory