Skip to content


Winblows, Sqlways, ntfs performance with thousands files.

I'm partially involved in a data migration project from Sqlserver to Oracle.

The product being used for the conversion is called Sqlways.
It is ok (I didn't see what it was up against).
There are a number of things I hate about it:

  • it isn't scriptable, wizard only, it stores the last set of user entered criteria in 2 separate files so you might be able to hack it.

  • Poor user interface, particularly on one of the main forms where you have to know which level in a tree control to click on to get important conversion parameters.
  • Atleast it supports creation of unix scripts so you can do your import on *nix.
  • Not scalable - large scale migrations would want the benefit of multiple cpu 's/machines to run in parallel to reduce the migration window. I guess you could do it in multiples partitioned by tables, but try syncing up a program that doesn't isn't scriptable out of the box.
  • It doesn't fail fast, during import, not sure about export.

The database to be transferred is pretty small yet the times were really blowing out on tables with blobs.

There is an option to actually save individual blobs in separate files (per column and per row). This doesn't go down well with NTFS file system when you have a million files even if they are small in the same dir.

So if you ever find yourself using this or a similar product:

  • Inline blob's where you can, one example was a reduction in export time from 87 mins to 3mins, with no affect on import performance. This would purely be due to the ntfs issues. Nice one Danny.

  • use direct path load feature of sqlldr on imports.

Posted in General.

0 Responses

Stay in touch with the conversation, subscribe to the RSS feed for comments on this post.

Some HTML is OK

(required)

(required, but never shared)

or, reply to this post via trackback.


Twitter links powered by Tweet This v1.8.1, a WordPress plugin for Twitter.