Re: Oracle on a laptop - how to support large-scale sales demos

"Jim Smith" <jim_at_jimsmith.demon.co.uk> wrote in message news:BG9VSuYN7LeBFwf5_at_jimsmith.demon.co.uk...
> In message <%0Udd.19460$jo2.13612_at_twister.socal.rr.com>, Runs with
> Scissors <tgb_at_REMOVEsoftitect.com> writes
>>I am currently working on a project for a Business Performance Management
>>application. The initial goal of the project is to develop a demonstrable
>>application that our sales force can show to potential clients. The
>>targeted clients will be higher echelon retailers (read as very large data
>>sets). One requirement of the project is to deploy, onto a single laptop,
>>the application server/web server, analytic engine, and database -
>>including
>>a realistic data set (i.e. a few/several gigabytes of data) in order to
>>provide realistic demonstrations. Note that in this case, several
>>gigabytes
>>of data represent a very small sample of the real-world product solution,
>>say 1-5%. The primary reason for the single box deployment is to allow an
>>anytime/anyplace demonstration. No network access required. They will
>>not
>>be demonstrating multi-users capabilities.
>>
>>
>>
>>Basically, They (aka Management) want a magical solution.
>>
>>
>>
>>The deployment will be Oracle 9i or 10g (Enterprise Edition), Web Sphere
>>5.0.2, and an analytic engine I'm not free to specify at this time. The
>>BPM
>>application will include several star-schema data marts covering things
>>like
>>sales, inventory, planned sales, purchases, planned inventory, etc. The
>>application must also emulate a real-time data feed, so the database will
>>be
>>a "general" database doing both transaction processing and data
>>warehousing
>>functions.
>>
>>
>>
>>I'm assuming that we will be able to get top-of-the-line laptops running
>>some flavor of Linux (probably Red Hat), but still, a single CPU and
>>single
>>drive smells like a performance disaster to me, not to mention a 2gig
>>memory
>>limit. I'm looking for alternatives in the deployment model. One thing
>>that comes to mind is a portable USB RAID solution with 5 or more drives
>>in
>>a RAID0 or RAID5 configuration. Another might be a three-laptop
>>deployment
>>where each laptop supports a specific server, or, possibly a combination
>>of
>>the two.
>>
>>
>>
>>Does anyone have similar experiences they would care to share?
>>Alternative
>>ideas on how to get enterprise performance out of a single CPU? Know any
>>good magic spells, or rules-of-thumb to quantify the performance expected
>>in
>>a real-world deployment?
>>
>
> Its not reasonable to expect a laptop demo to demonstrate the performance
> characteristics of a full scale system, and I don't think your (potential)
> customers would expect it. A salesman demo is generally just to
> illustrate functionality.
>
> If your customers are interested in the product then the sales process can
> move on to bench marking if necessary
> --
> Jim Smith
> Because of their persistent net abuse, I ignore mail from
> these domains (among others) .yahoo.com .hotmail.com .kr .cn .tw
> For an explanation see <http://www.jimsmith.demon.co.uk/spam>

Thanks for the response Jim.

There is most certainly a difference between what a laptop can demonstrate versus what a full-scale system can provide. I think the key words here are demonstrate and provide. Ideally, we would be able to reduce the data along appropriate axis in an attempt to scale down to "laptop" size. Granted there are more aspects than simple CPU/hard drive quantities to consider, but we should be able to demonstrate the functionality at some quantifiable level (e.g. benchmark) without completely throwing performance out the window.

Something we, as a company, have been experiencing is that demonstration performance is a go/no-go decision point for many of the prospects. It shouldn't be this way, but when showing an application intended to replace something that exists in a spreadsheet, scalability lands on unhearing ears. "Ours runs this fast on my laptop, why can't yours?" is a generalization of many responses. I can never decide if the prospect is genuine in the response, or if they are simply establishing a negotiating point. Many times these same people will tell us they don't need a data warehouse (sometimes they are correct) and arguing the technical merits of a data warehouse vs. transaction system vs. operational data stores, are not the sales forces strong suits.

I was interested in hearing other peoples experiences, and if this is your experience, I thank you for your reply.   Received on Fri Oct 22 2004 - 15:40:29 CDT