[nycbug-talk] Hypothetical: the end of the sysadmin/systems engineer/DBA?

[George Rosamond]

On Sat, Mar 27, 2010 at 12:11:16PM -0400, Marc Spitzer wrote:
> On Sat, Mar 27, 2010 at 10:57 AM, Matt Juszczak <matt at atopia.net> wrote:
> >> build process and you fix the build process. ?Now doing this right is
> >> a high skill task, trouble shooting a box has become a low skill task
> >> and the middle get squeezed a bit smaller.
> >
> > So where are the lines drawn? ?If someone who has experience only in the
> > things you specify, like running fdisk, debugging broken servers, basic
> > installs, etc., then I agree that type of skillset is going to be less and
> > less in demand. ?But what about people with more knowledge? ?Knowledge in
> > LDAP, CF Engine, Puppet, Apache - those sorts of tools, and even more
> > important, standardizing servers (central authentication, central
> > authorization, sudo in ldap, etc.). ?Are these going to be automated as
> > well?
> 
> Personally I think you want to keep an eye on the system side and not
> so much on the server side.  I expect that other then fans servers
> will have no moving parts fairly soon, 5 years or 10 at the out side,
> as plater drives get replaced by SSD drives, there goes a lot of work.
>  Also properly designed servers, for example HP's, are very fast to
> work on, again there goes some work.  So you will have less to do and
> what you have to do will take less time, any manager will tell you
> that means reduce head count.
> 
> >
> > Think about this: will there be a time when competitors no longer compete,
> > because major providers will automate the instance creation of environments
> > (for instance, Amazon may choose to make a one-click environment that uses
> > puppet for configuration, openldap for authorization/authentication, postfix
> > for email, and a few other tools for things like monitoring). ?Would
> > competiting tools (like exim, for instance) be run out of demand as more and
> > more people go to these "one click" approaches and choose not to replace
> > them?
> 
> Well for postfix, at the time it was written sendmail was the MTA on
> the internet, no not so much.  Good ideas do tent to survive and
> prosper.  But the need to be much better to be worth the pain of
> switching.  Postfix is as long as you can live with a mail switch as
> sendmail is a mail router, switch == internet only and router ==
> internet to other network(ex. decnet).
> 
> Also traditional sysadmin skills include areas like security, security
> is a growth industry.  Disaster recovery, capacity planning, building
> cloud physical plant for people who want the cloud but need to keep
> data private and a bunch of other skills.  And to reply to the exim
> bit, what is stopping exim people from taking the dam thing and
> replacing postfix with exim and uploading it for people to use?
> 
> I think the SA world is going to split along the lines of people who
> build integrated circuits, people who build things with integrated
> circuits and people who put IC's in sockets,  Figure out where you
> want to be and get there.  Two of the three will pay well.
> 

I think the problem with approaching this question is everyone gets very linear in their estimations.

The waves and troughs vary in depth and breadth.

What would your guest-imates be if the internet (really) "Balkanized"?

What if technology goes into something of a stagnation or a decline?  Certainly whole periods of history are full of such happenings.

Even look at the shallow yet explosive technology boom of the late 90's.  From new web browser releases, IDS systems, etc., every six months. .. to a certain amount of stagnation for a few years post-2000.

Assuming there continues to be forward-moving trends, though, I'd agree with Marc about the "less moving parts" idea. . . that is already in flux, and it's hard to believe on desktops/laptops/stand-alone servers/devices that this trend won't continue.  Oh, right, it's kind of like old-school terminals in the terminal-mainframe model.

More generally, what about the cloud?

Well, if someone figures out how it's dramatically different from clustering, network-wide file storage, etc., then it would be worth approaching as a distinctive trend.

However, lots of development in security devices and apps, management tools, control panels,     
 etc., will certainly go hand-and-hand with the (possible) cloud growth.

The 1990's explosion of colocation centers, company and carrier-neutral, certainly opened some deskilling among a layer of staff at those places, and also opened up "differently skilled" technical people into the world of higher uptimes and networks.  This is also the case with various hosting packages, available all over.

That was a more significant trend, IMHO, then the (in the) cloud talk.

g