[LEAPSECS] system engineering best practices

[Rob Seaman]

On Aug 10, 2013, at 11:13 AM, Poul-Henning Kamp <phk at phk.freebsd.dk> wrote:


> Dave Finkleman writes:

> 

>> However, things now broken do not depend on the leap second

>> whereas those that have implemented it correctly obviously do.

> 

> No, there are three classes of systems:

> 

> 1. Those which ignore leapseconds

> 2. Those which handle the leapseconds

> 3. Those which depend on leapseconds

> 

> Estimated ratios:  10,000,000 : 1,000 : 1


A completely ad hoc estimate (with perhaps a slight spin? :-)  No comparative impact ranking.  Not contrasted with the implications should leapseconds continue.  No costing.  No schedule.  No risk analysis.  No systems-of-systems issues.  Etc and so forth.

Leapseconds exist for a purpose.  An inclination to discount that purpose would benefit from well known system engineering best practices to compare the trade-offs.  In the mean time projects that want to ignore leapseconds can choose a different time scale.

Another quote from that data center infrastructure management article:


> The goal is to provide a seamless, real-time view of all elements of data center operations, allowing data center managers to quickly assess a range of factors that impact efficiency and capacity.


Leapseconds are just one of many factors.  For instance, the far more massive impact on Amazon of the nearly concurrent electrical storm:

	http://www.datacenterknowledge.com/archives/2012/06/30/amazon-data-center-loses-power-during-storm/

Rob