Writing Free Documentation Just Got Sooo Much Easier

Posted by Martin P in Open Source      No comments

May
23

In the formation of user guides and documentation, there are inevitably points where background information is not only appropriate but essential.  In the case of a user guide for DCM4CHEE, background sections on DICOM, HL7, and XML/XSLT are good examples – a user guide without some explanation of such concepts would be incomplete.

Where would one go for a briefing outside of the context of a user guide?  There is only one place to start of course and that is the ever-increasingly useful Wikipedia.  That unfortunately is where it stops when building a user guide, because content in Wikipedia is licensed under the GNU Free Documentation License. GFDL has a number of critics (including the very freedom-oriented Debian project) but the main problem is that it makes it virtually impossible to take snippets of different documents (in the case of Wikipedia, articles), and collate them into a separate document.

One of the reasons this is so hard is that the in GFDL, the full history of any source documents must be included in full in the resulting document.  To include the history from even a handful of Wikipedia articles would end up in a horrific document.

But now, SlashDot reports that the Wikimedia Foundation has resolved to relicense its content with the Create Commons Attribution Share-Alike 3.0 license.  This is an order of magnitude simpler than GFDL and effectively means that content can be reused with just a link back to the original.

A major step for increasing the quality of free software documentation.

Not having a great involvement in Wikipedia, I cannot say if the WikiMedia software has the abilty of alerting subscribers when edits are made.  This would be an important element.  To properly maintain any free resource, one must be in touch with upstream developments.

Archiving PACS data to DVD

Posted by Martin P in PACS General      No comments

May
22

Having re-read this post I thought maybe the postscript point is worth re-iterating in its own post:

If you are using DVD-R disks for archiving, beware of the shelf-life.  Many commodity DVD-Rs – even from apparently reputable brands – can become unreadable after just a couple of years (me typing here with burned fingers!).  Make sure you get archive-quality platters (we use Taiyo Uden). Here for more information on platter quality and Care and Handling of CDs and DVDs (Fred R Byers) (pdf).

Believe me, its a real risk.

Cost-effective storage just gets cheaper and cheaper…

Posted by Martin P in Infrastructure      1 comment

May
22

I recall when I started in the world of PACS (over 10 years ago) the idea of specifying a PACS with 10 years worth of near-line storage was lunacy.  That, of course was in the days when data volume was nowhere near the levels of today, and 10 years equated (in the instance I recall) to 10TB.  The cost of that much storage was astronomical, and even back-of-the-envelope calculations could not find adequate Return on Investment (RoI) to justify.  There was a clear inflection point at around 3-4 years where the savings in Radiologist and Radiographer time ceased to justify the added cost of nearline vs hierarchical-storage.  But I know even then many disagreed.

As both storage technologies and data volume have advanced apace, the stories have both changed and stayed largely the same.  SAN technology with real-time replication between multiple sites (invaluable for both performance optimisation and disaster tolerance) has, for the last 4 or 5 years, been the no-brainer, but that still comes with a chunky price tag. 10-20TB of that level of SAN protection will cost anywhere between €150k and €400k depending on vendor (that is, storage vendor, not PACS vendor) – and that is after firm negotiation.

But is the story changing?  For smaller installations, it may well be.  It is possible to achieve massive storage capacities on a shoestring budget.  Recently, announcements from Iomega and LaCie illustrate a distinct warming of the marketplace around SME storage.

Of course, these products aren’t nearly as robust, reliable or performant as the SAN equivalent.  But then, as long as they come with redundant, hot-swap disk and redundant power and NIC, then does it matter?  With judicious use of software replication (being mindful of the limitations) such as DoubleTake for Windows, or GlusterFS on Linux, and multiple units of less-expensive hardware (indeed isn’t that the basic idea behind RAID?), an installation of 10s of TB is achievable with very little capital investment.

It does (generally) require other investment.  It needs time, care and attention.  Thee are elements that are often factored out of a solution by paying more up front.  In these trying times, perhaps a little more time investment is worth the financial savings.

With this kind of solution, it is crucial to understand and navigate around the limitations of whichever solution is pursued.  Whether the solution is a periodic synchronisation via rsync, or a high-end distributed architecture like the open source CleverSafe, one should always be aware of the limitations, and the swings, and the roundabouts.

But then.. isn’t that always the case?  IMHO no amount of capital expense absolves one from that responsibility.

Anyway, looking to the future, and bearing in mind the 10TB example at the top of the page, in years to come we may see 10TB DVDs.

On a completely unrelated point on DVDs (and CDs for that matter).  If you are using DVD-R disks for archiving, beware of the shelf-life.  Many commodity DVD-Rs – even from apparently reputable brands – can become unreadable after just a couple of years (me typing here with burned fingers!).  Make sure you get archive-quality platters (we use Taiyo Uden). Here for more information on platter quality and Care and Handling of CDs and DVDs (Fred R Byers) (pdf).