The Downside of Real Time

  • Comments posted to this topic are about the item The Downside of Real Time

  • "business rules dictating triggers firing off emails"

    isn't firing off an email for every transaction a core part of being ACID?

    🙂

    ---------------------------------------
    elsasoft.org

  • So I think you saying that SQL Server Central should not be sending us an email every time someone posts to a thread we have subscribed to.

    Instead, it should have some configurable frequency and send us a single email containing links to all of the threads that have had a posting during the last period.

    I am prety much ok with getting emailed repeatedly.

    In many cases, I have given up trying to explain to people that what they have asked for is probably not what they want.

    From a BI perspective, I am careful to get users to split up "operational" reporting and "strategic" reporting. I have been pretty successful convincing users that operational reporting is real-time OLTP reporting and strategic reporting is data-warehouse and scheduled reporting. In most cases, I am then just asking users to tell me if their report has strategic value to the company or if it is being used to manage day-to-day operations.

    It is amazing how many people will accept that their reporting will not be real-time just to be able to say it has strategic value.

  • Personally, I take the attitude that, as a DBA, I have all the rights and none of the authority to do things. In short, I don't believe it matters whether or not I think any request for (in this case) real time reporting is right or wrong; if the business feels it is necessary, it's my job to implement it, not judge.

    Of course, it's also my job to let the business know of any cost or security implications or technical limitations before the business makes its decision. In practice, of course, that means telling someone something along the lines of, "you can have what you want but it'll be expensive and come out of your budget" or, "you can have almost all that you want for next to nothing as long as you can live with updates happening only every 30 mins/1 hour/daily etc.".

    And if, having provided all the information necessary for the business to make the decision, I don't like the outcome, that's my problem and no-one else's.

    IMHO, of course.

    Semper in excretia, suus solum profundum variat

  • I always warned my users ahead of time that if they wanted real-time reporting (or as close as the system would let me) they would need to understand that the report is only good for the timestamp printed on the page. With real-time reporting, you get the data as it happens, and if a user enters something in wrong and a supervisor has to correct it, it's quite possible for the data to be different from when the report was run. And every couple weeks, I'd have at least one manager ask why the report he was looking at didn't match the one another manager had run 5 minutes after his. :doze:

  • My favorite was trying to get people to understand the difference between "real-time" and "real-enough-time". We had one manager who was insistent that his numbers had to be real-time when he ran his report (a huge monster with tons of queries, hitting all sorts of tables and bringing the server to its knees). After much discussion, he realized that if he could pull his data off a server using replication with the data pretty much up to the minute he could get his report just about as accurately and without stopping all work while the report ran.

    Of course, we went around in circles for several weeks trying to figure out why he needed this report to be up-to-the-second for something that would be printed out and displayed, not analyzed or used to predict dangerous trends that needed to be addressed right that second. It ended up being that he really just wanted a report that was reasonably accurate for the time it was generated so he could do some basic trend analysis for the day/week and let the employees know how they were doing at mid-day. Still, trying to work within the requirements of "real-time" would have meant lots of money for little actual value.

  • Been there done that and think I have a t-shirt from a company stating the same thing on it. Fortunately over the years I have been able to get most to undertsand what they want and when we are working thru requests we make sure we always point out that if they say real time what they will see happen. Most times it is a snapshot of the system from the prior day they actually want more than anything. But we try to be very specific in meetings on design what they will see and get along with the common complaints we have heard so we can set expectations. Then we note it in a BRD addendum and say review you documentation when they call us.

  • I agree with that and in an even broader sense it can simply be covered with "too much information" as you said. I once had complaints about a report that took too long to execute. Well, the report would return over 500,000 rows! So of course it may take it a while to execute. My response to them was, "What are you possibly going to do with 500,000 rows? You're not going to ever look at every single row. So why don't you tell me exactly what you are looking for and I might be able to speed up the report." But of course all they could tell me was, "We don't know what we want, but we want the report to run faster."

  • I love long reports. When I get a request for something really outrageous, I tend to break it up or summarize things since that's what people usually want.

    It's interesting to think that the "Cost" of getting things to update and appear in real time is similar to the cost of uptime. Getting closer and closer to real time means a larger cost, which often isn't worth it.

    I've always wanted to build a dashboard that would update in real time, kind of like the speedometer, tech, etc. in a car, showing the managers what's really happening now. That alone would probably capture their attention and prevent them from asking for more reports 🙂

  • Steve Jones - Editor (12/12/2007)


    ....That alone would probably capture their attention and prevent them from asking for more reports 🙂

    Optimist.

    Semper in excretia, suus solum profundum variat

  • Ah, no a Realist. Build a nice Fisher Price interface, similar to this one - http://www.fisher-price.com/img/product_shots/L3480_d_1.jpg, big buttons, lots of color, things moving, and you'll get a manager's attention everytime

    :w00t::D

  • Hey, Steve

    How did you know that managers tend to look over my shoulder whilst I'm working!

    Nice work environment, though, eh?

    Semper in excretia, suus solum profundum variat

  • Steve Jones - Editor (12/13/2007)


    Ah, no a Realist. Build a nice Fisher Price interface, similar to this one - http://www.fisher-price.com/img/product_shots/L3480_d_1.jpg, big buttons, lots of color, things moving, and you'll get a manager's attention everytime

    :w00t::D

    Bet they still couldn't tell you want they wanted to see from that.:w00t:

  • It's always fun.

    Who cares what they want. Change the label and color, and they're fine.

    Heck, I could probably remove the scale, just put some indicators and lines for "relative" positioning and hook the thing up to the thermostat and they're watch it all day. 😛

Viewing 14 posts - 1 through 13 (of 13 total)

You must be logged in to reply to this topic. Login to reply