A Grain of Salt
-
Comments posted to this topic are about the content posted at http://www.sqlservercentral.com/columnists/sjones/agrainofsalt.
-
ABSOLUTELY!! This one hits close to home, a company I used to work for was 'investigating' out-sourcing the IT dept. As rumors started circulating all over the company about it, mgmt decided it was probably time to actually say something about it to IT staff. They tried to present it as merely as study to quantify and prove the worth and effectiveness of IT dept, since, as one Mgr said 'Numbers don't lie!'.
Good thing I was in a seperate office listening via conf call, with mute on, it's been a while since I had a good laugh like that. Statistics can mean whatever you want them to mean, and if they don't, they'll just ask you to keep working until you find some that do.
-
I've been in this business for over 20 years - way back to the days of punch cards and paper tapes. We've always measured uptime as the availability of the system to our users. We always log ANY activity that prevents users from accessing the system, no matter what the cause. We even break this down to our various network segments. This way we can report that users in location A had 99.9% uptime, while users in location B only had 99.5% uptime.
Now if only those users would just go away, we DBA's could get some work done!
-
There are four (mis)quotes that spring to mind when someone bangs on about statistics.
- He uses statistics as a drunken man uses lamp posts -- for support rather than illumination (Andrew Lang)
- Lies, damn lies and statistics (Benjamin Disraelli).
- The devil takes equal delight in quoting scripture and statistics. (HG Wells).
- If your experiment needs statistics, you ought to have done a better experiment. (Edward Rutherford)
Statistics are supposed to be a tool to determine IF something has worked, not something to proove anything.
My personal feel on uptime is "was the system available when it was needed". Why should I care that the system works between 18:00 and 06:00 when everyone using the system works 09:00 - 17:00!
-
Great point! I really shudder when someone quotes statistics completely out of context.
I come primarily from a mathematics and philosophy backround (before anyone snickers at that, philosophy is, at its core, analytical thinking and problem solving). One of the most interesting graduate seminars I had was working with probability. The point is, there is no such think as "knowledge", only probability and corelation. IT is a perfect example of this. It's important to keep in mind what 99.99 uptime means. It has nothing to do with the likelihood of future uptime -- it is only a calculation of historical data (and a very primative representation of it). Historical uptime has no direct impact on the probability of future uptime -- it's easy to say, "Hey, we're doing great with 99.99%", only to follow it with a major hardware failure that drops the stat down to 89%. Could you have predicted it? No. Does that fact that you're at 89% uptime mean that you're doing a poor job at availability? No. A statistic is just a number, nothing else. It can have value (depending on how it is derived), but it in no way determines or implies truth.
OK, that was up from the soapbox, but I think you've hit a very important point that most people take for granted. Well done.
-
I like statistics
Especially those I have manipulated myself
It depends on whom you present your statistics to. I usually have a totally different argumentation when arguing about the same numbers to internal management or external auditor. And I've made the experience that (except for auditors) no one is really interested in precise numbers, but rather in plausible ones. In some maybe existing ideal world both premise match, but not here...
And as David and James already have mentioned, a healthy mistrust is almost everywhere advisable.
--
Frank Kalis
Microsoft SQL Server MVP
Webmaster: http://www.insidesql.org/blogs
My blog: http://www.insidesql.org/blogs/frankkalis/[/url] -
I like statistics Especially those I have manipulated myself Wot your age
or your IQ 
FrankFar away is close at hand in the images of elsewhere.
Anon. -
Bear with a slow german, Dave. What does this mean?
--
Frank Kalis
Microsoft SQL Server MVP
Webmaster: http://www.insidesql.org/blogs
My blog: http://www.insidesql.org/blogs/frankkalis/[/url] -
What does what mean?
You're confused!!!
Far away is close at hand in the images of elsewhere.
Anon. -
Hey, I'm at the office.
Sure I'm confused right now. I always argued the equities have a higher implicite risk than bonds or the german type called registered bonds. So far, that's nothing too scientific, or?
Well, now management decided in their wondrous divine enlightenment, that we don't have ANY risk in equities. Aargh, I really hate those marketing dictated decision. It doesn't even help that this decision will be well documented as you can't expect management to be able to read. Hm,.. sometimes this ability to have instantaneous memory weaknesses on former decisions must be something you learn on management seminars.
Now, what I wanted to know is how I can translate WOT?
Looking at a dictionary yields wide-open throttle, but what is the meaning?
--
Frank Kalis
Microsoft SQL Server MVP
Webmaster: http://www.insidesql.org/blogs
My blog: http://www.insidesql.org/blogs/frankkalis/[/url] -
Ah sorry
wot is 'Eye dialect for what. Chiefly British.' Now don't ask me what 'Eye dialect' means
I will try to write proper English next time
But.....
There would no fun
Far away is close at hand in the images of elsewhere.
Anon. -
Frank,
You know when you are trying to teach your kids to behave properly part of it is getting them to speak properly (and the rest is teaching them when to shut up)?
If they speak properly they should not say WOT they should say WHAT!
I have to say that one thing that always humbles me as an Englishman is the ability of other nations to speak my language when I am unable to return the compliment.
My wife has taught me a few German swear words but that isn't much use in a restaurant.
-
You know when you are trying to teach your kids to behave properly part of it is getting them to speak properly Yes but when children are young they are taught phonetics and later on are told that phonetics is not the way to pronounce words. Besides what is 'speaking properly', the so called 'Queens English' with (proverbial plum in mouth). I prefer diverse accents, which I reckon will be forced out due to PC (Political Correctness before Frank asks). The classis being the town of Towcester pronounced toaster.
I'll get of the soapbox now
Wow I have really gone off topic now
As for the topic itself, I couldn't agree more with the comments so far. How can you justify 100% uptime when the user is off on holiday for two weeks.
Far away is close at hand in the images of elsewhere.
Anon. -
You know when you are trying to teach your kids to behave properly part of it is getting them to speak properly (and the rest is teaching them when to shut up)?
Yes, I do! But once they go to the Kindergarten (my elder one is 4.5 years) I guess that's in vain and becomes a never ending story (not to say battle). And what I observe from friends with older kids it doesn't get better.
Here in school they try to teach something called 'BBC english'. Hope I remember correct. At university I attended some english courses, but soon gave up when I realized that anything else is done there, but speaking english
Actually, I don't mind knowing some of this less formal english aka slang. And I ask when I don't understand something. Funny side-effect posting to this forum, is that I sometimes dream in english. Should I worry about that?
As for dialects, I guess that's the same in every country. Here I only need to go 30 kilometers to Cologne and there is the right places I feel like a foreigner
. And that's getting worse the more I move south.--
Frank Kalis
Microsoft SQL Server MVP
Webmaster: http://www.insidesql.org/blogs
My blog: http://www.insidesql.org/blogs/frankkalis/[/url] -
Personnally, I love this one. I have seen and been a part of manipulating the data to provide the best potential statistics possible. Most companies usually drop the lowest and highest numbers (they call these extremes and they give up the highest so they can discard that big mistake at the bottom) usually this gives a nice balance of a normal situation. Knowing this everytime a company makes a statement such as 99.9% uptime I promptly ask for their lowest and highest figures back in the data, if not willing to provide (sorry we don't have those) or say there are none I promptly state "I thank you for your time but we are looking in another direction". I want true quantities not bogus averages.
Dealt with a group once whose SLA was 99.9% uptime but over a weekend the servers were unreachable the entire time. However when reported at the end of the year it was 99.9% up as agreed in the SLA. When questioned on it they claimed the servers themselves are redundant and were available even thou the provided process (the COM objects that made the connections) did not roll themselves to the backup locations as was expected by us (the end user). My problem with this is they provided the COM objects themselves, not a third party company and my group did not use an API set to write our own. In essence it was their issue but they avoided the numbers for the COM object since it upsat their claim.
All in all statistics can be manipulated to suite anyones needs including those outsourcing companies use in order to get your business. Ask to speak with other customers and if all you talk to are people who had minimal issues at all stages of use then remember the old saying "If it sounds to good to be true it probably is.". If however you specifically ask to speak with a group who did run into issues and maybe choose not to go with the product and it is provided then you may want to consider that company as they take pride in being upfront with you, but I am betting this won't happen in 99.9% of the cases.
Viewing 15 posts - 1 through 15 (of 21 total)
You must be logged in to reply to this topic. Login to reply