-- Stephen Braunius Director of Instructional Technology Zeeland Public Schools - http://www.zeeland.k12.mi.us Zeeland, Michigan
B0027863531@mail.webdsi.com">hello,
On Scaling : --> fitting data topology to need and demand.
two seperate factors must be scaled :
(1) Quantity
If I must keep 10,000 100 byte records (A) per 9 week term,
with four measured terms per year
then the volume of the data will grow by
4A per annum.
this is a number all your vendors need.
(2) Accessability
if one hundred users access together the average wait is ~
% fails rises on slope @ with volume to M
you respond to these motivations by publishing agreed to service levels.
this ties service to a transactional cost.
"Within 30 days, late orders 72 hour delivery on notification of tardiness
and so on ..
benefit --> you can observe tradeoffs between cost, effectiveness and public
image. (Response time and failure rate are the major impacts of record keeping
operations on public image of the institution. ( Purely based on
observation, not statistics. )
let's take :
50 folders of 100 bytes each :
5000 bytes - 50 folders by one backup : 10000 bytes - 100 folders
by 40000-400 (four measured terms) annual
for 10 users with 8 record accesses over the total population of folders :
10x8x400 = 32,000 accesses
if all accesses occur exactly on the four term boundaries
on a single 6 hour day in each of these four instances
and an average fetch is one minute user time (at the client / terminal )
each day you can do 360 transactions
but you need to do 8000 in this period
so you can either :
a : drop the transaction time. at one second, it's at 21,600 - overshoot !
b: lengthen the publishing period. at 6 days of 7 hours each 2520 (not good
enough to meet anticipated demand).
these worst case rough cuts establish scalibility goals - the end points.
this approach lets you size the ruler you're going to use.
vendor relationships could start with this number ?
my "hunch" is that growth in accesses for a given folder show high demand for
at most 5 measured terms to span years, then they decline on an exponentially
increasing ( if negativce ? ) slope i.e. people look at newer records more
than older records and that as the distance along that line increases so does
disinterest.
you can design a solution with these numbers.
i would use two brand new wal-mart 1gig athalons
with 2 40 gig drives
and 256 meg memory
and standard redhat 7.2
maybe swapping out some parts
more than enough to do the job, but,
price and value count too :-)
there are better math guys than me on every street corner,
so the idea is a little rough cut, but
... well - you tell me.
mike eschman, etc ...
not just an afterthought
http://www.engima.com
On Wednesday 1 3 March 2002 09:48 am, Stephen Braunius wrote:Thanks for the response! I'll try to clarify my situation by responding
to your questions:
Les Richardson wrote:1) What format are the assessment items in now? Are they in just raw text
or are they marked up in some fashion...Tex, HTML, etc. or are they just
collections of files in proprietary formats like Word, Wordperfect, etc?Currently we are collecting or converting all assessment data into CSV
files using either FMPro or Excel. These assessments are typically
either percents or scales from 1-4. Everything is currently tagged to
our district assigned student id's. These assessments are created as a
part of our curriculum rewrite/revisit process.2) How do you want these items to be assembled into tests?The assessments are actually cumulative scores (kind of like final
grades).3) How are you going to classify these items so that they can be usefully
retrieved?This is what I am struggling with. I don't know how best to store the
data to be meaningful over a long period of time. Is it best to have a
table with characteristics of each assessment (year given, score type,
scale, quantatative/qualitative, curricular area, etc) and a table with
just data and the associated test?4) What kind of output formats do you want to generate?I would like an individual teacher to be able to pull up information
about each student, the average for their class, and the average for the
building in each area.5) Who has access both for retrieval and for entry/editing of existing
items?Currently all entry is coordinated by me using CSV files imported into
MySQL with with fields like this: student id, score1, score2, etc. They
are then associated with a table that has some very general test
information.So, I'm willing to help out with some opinions or code if you wish. I use
a MySQL database and Perl/DBI for my testbank. Markup is either raw text
or TeX for the items.Thanks for the offer. I enjoyed looking at the testbank stuff you have
created. It looks like you have already worked through a lot of the
items for how to collect assessment data. I wish there were some site
out there that has a diagram for how best to create the database tables
for creating an academic assessment warehouse.
Wow, if anyone made it to the end of this email and actually understood
what I was saying I would be amazed. ;^)
I'll see how many people have ideas on this topic and will then probably
contact you outside of this list since the topic probably isn't of
interest to everyone. Thanks again for sharing your work.