Drew, 

Playing the "uncertainty" card is defintely a very good move.  As you
mentioned, "higher ups" and "owners" don't like to deal with the "unknown",
and sometimes this can be a good approach. 


Celeste
-----Original Message-----
From: Drew Falkman [mailto:[EMAIL PROTECTED]]
Sent: Monday, April 01, 2002 10:39 AM
To: JRun-Talk
Subject: RE: load testing (was RE: Difficult To Reproduce Session Issue)


I have not, quite frankly, worked in a corporate environment. But as a
consultant I often face the same issues - not just with load testing, but
also with general code tuning and upgrading the applications I have created.
It's almost an impossible sell to convince a customer they need to spend
more money and not be able to "see" the results. Celeste, this is probably
old hat to you, but for others lurking, I have found some good approaches to
convincing upper management that code tuning is important (these would be
the same issue for load testing):

1. Track regular performance during regular traffic. Maybe take screenshots
of perfmon or log performance heuristics so you can easily create graphs of
performance throughout the day. IF you have SNMP monitoring tools, use them
(some are pretty reasonably priced). Sometimes simply illustrating the
fragility of an environment will make upper management afraid and want to
know more. You could probably put together some of this preliminary stuff in
a couple hours.

2. Present the above along with "what-if" scenarios (like what if a hard
drive blows, or what if we get 500 users simultaneously), and explain that
without load testing you simply cannot know what will happen.  Most upper
management types simply don't deal well with uncertainty (think of how you
are always needing to give accurate assesments of the time it will take to
do project X). If you illustrate that you don't know what will break first
under heavy loads or what the breaking point is, that will often help.

3. The reality is that given the two points above most people will agree
that load testing *would* be a good idea, but will often say that they
simply don't have the time or the money to invest in it. "Let's do it next
year." is not an uncommon response. The only way to get around this is to
speak their language: ROI. Return on Investment. Calculating ROI, especially
with something like this, is not easy and often arbitrary. But basically you
need to put together the cost of doing load testing (hours + software, if
necessary). Then, put together the costs for a couple scenarios. One should
be the cost of the applications server going down. Estimate the amount of
time it would take you to rebuild your environment. Now, add in dollars for
the amount of sales lost while your server is down. Add more for the
confidence lost by customers who tried to visit your site during downtime
(this can be large, imagine if 100 customers never return at
$100/yr/customer). Give a cost scenario during a promotion.  Again, explain
that this is pretty much inevitable if you don't know at what point your
application will break and where. It might be good to do one that is a
little less severe (people tend to think that the tragedies won't happen to
them, but might believe that the server will get bogged down and you will
have to spend time to troubleshoot and tune code and waste a lot of time
because it is less drastic).

4.  Lastly, sometimes use a promotion or current project to spur it. Simply
asking what kind of traffic is expected from an upcoming promotion opens the
door to you to say: "Are you sure we can handle that much?". If you follow
that up fairly quickly by saying you could find out by doing some load
testing will illustrate that you know what you are doing and also that you
are nervous about the upcoming promotion (which should make them nervous, as
well).

Just some thoughts.

Drew Falkman
Author, JRun Web Application Construction Kit
http://www.drewfalkman.com/books/0789726009/


-----Original Message-----
From: Haseltine, Celeste [mailto:[EMAIL PROTECTED]]
Sent: Monday, April 01, 2002 8:11 AM
To: JRun-Talk
Subject: RE: load testing (was RE: Difficult To Reproduce Session Issue)


Charles,

My response wasn't directed at just "what to do when the bosses don't
listen", but also how to "not push your ideas to the point that you alienate
the bosses you are trying to convince".  Again, unless you are an officer of
a company, or are the owner, you are often NOT in a "decision making" role.
Therefore, the only thing you can do is to try and convince people to test a
product before deployment.  But you can't force your bosses to do something.
And if you try to do so, you will risk your job and your professional
reputation.

Again, manpower issues and money issues are the key here.  Even if the tools
are free, it takes the use of "x man-hours" to test a product, when those "x
man-hours" could be used towards something viewed as "more profitable" or
more important to the company by upper management.  This is especially true
when you are in a very small shop as I am.  I can push all I want, but if I
push too much, I run the risk of alienating the very people I am trying to
convince.  And if I do alienate my "bosses", what have I gained for my group
and "my cause"?  Nothing.  In fact, by pushing too far, I could end up
hurting my group and my "cause" more, by not giving my management team a
"graceful out" once they realize that I was probably right all along.  This
is called politics.  And the higher you move up as an IT professional, the
more important the "how well you play the political game" becomes vs. "how
often you are right and your boss is wrong".

Perhaps someone else on this list has a better approach to convincing their
management team for taking the time and the money to do things such as
testing.  I'm always interested in hearing other/better approaches.  But for
me, this is the way I have approached all "opposition" to any IT
recommendation that may be viewed as "not necessary, too costly, etc.",
testing included.  As I mentioned before, people and companies both make
mistakes, and hopefully learn from them.  Sometimes you just have to let
your upper management make what you know is a decision that will probably
come back to haunt them, and then just move forward from that point on.

Celeste

-----Original Message-----
From: charles arehart [mailto:[EMAIL PROTECTED]]
Sent: Saturday, March 30, 2002 1:10 AM
To: JRun-Talk
Subject: RE: load testing (was RE: Difficult To Reproduce Session Issue)


Well, those are all good thoughts, Celeste, and surely helpful to many. But
my point wasn't "what do we do if the bosses don't listen" so much as "why
aren't we pushing testing (especially load testing) more". I think there's a
sense in many that it's not important, or it's too difficult, etc.

While it's true that many sites may not get thousands (or tens of thousands)
of hits in a day, it's reasonable for one getting just hundreds to get
several at once, and there are clearly issues (like the one that touched off
this thread) that may be triggered with just such a low level of load.

If I seem to be on a bit of a bandwagon for load testing, even for "smaller
shops", I am. And as has been pointed out, cost of tools isn't really the
issue. I'm curious for the kinds of arguments against it that have either
kept people from pursuing it or, even in the face of these assertions of the
low costs and good benefits, why they would still say "nah, not going to
bother".

/charlie

> -----Original Message-----
> From: Haseltine, Celeste [mailto:[EMAIL PROTECTED]]
> Sent: Friday, March 29, 2002 10:55 AM
> To: JRun-Talk
> Subject: RE: load testing (was RE: Difficult To Reproduce Session Issue)
>
>
> Charlie,
>
> I just try to point out the business cases for testing.  With many
> owners/managers, it comes down to money and time, with time also being
> related to money.  As a professional, I ALWAYS write up my
> business case in
> a 1 page summary, which hits the highlights as to why ANY
> software product,
> internet or otherwise, should be tested.

<snip>




______________________________________________________________________
Your ad could be here. Monies from ads go to support these lists and provide more 
resources for the community. http://www.fusionauthority.com/ads.cfm
Archives: http://www.mail-archive.com/[email protected]/
Unsubscribe: http://www.houseoffusion.com/index.cfm?sidebar=lists

Reply via email to