In 89S8252 datasheet it is mentioned that,
2k Bytes EEPROM
- Endurence:100,000 write/read cycles.

Does it mean if I write or read EEPROM 100,000 times
it may fail?

Hallo Mak,

there's an error in your post: It's not 'write/read cycles' but 'write/erase cycles'. Don't worry about read cycles. There's no danger, at all.
When an EEPROM cell is erased or written, a quantum mechanical tunneling effect is used to put a handful electrones on highly isolated gate. Because extremely high electric field strengths are used for this, crystal erodes a little bit every time. So, after a certain number of cycles a failure can occur, means, at least one cell does not allow any further reprogramming.
Another topic is 'data retention', This mechanism is similar. But here, EEPROM cell looses information after some time.

Dealing with terms like 'data retention' and 'endurance' requires the use of statistical concepts. So, endurance means, that when analyzing behaviour of an infinite number of identical EEPROM devices, a certain percentage of this infinite number of devices shows a failure after 100,000 cycles.
Here a high potential of arbitrary is hidden, because endurance specification highly depends on this percentage. You will get total different endurance specification if you say 1% or 5%. And because mathematics behind is rather comlicated, it's not so easy to verify manufacturers data. If I remember correctly, it has to do with chi-square distribution.
As consequence manufacturer can fabricate an as arbitrary specification as he wants.
Another big factor is temperature. Rising temperatur shortens endurance! This must be kept in mind.

For me, the data '100,000' is purely psychologically motivated. Keep in mind, that 100,000 cycles with 1% failure rate or 10,000 cycles with 0.1% failure rate or 1,000,000 cycles with 10% failure rate, can correspond to the SAME reliability. (Given data here is only for demonstration purpose, actual data depends of course on statistical material!!)
Would be 10,000 good enough for you? No, 100,000 is better, of course, isn't it? And 1,000,000? No, who the hell is changing EEPROM content a million times?
Did you get me?

There is one other very unfortunate point. Statistics does not allow you to make any prognosis of performance of YOUR INDIVIDUAL device. So, you cannot say 'ok, I have changed EEPROM 80,000 times, 20,000 cycles I have remaining'. This sight of view is totally wrong, when dealing with statistics. It gives only information about performance of a real big number of devices. So, if you plan mass production, statistics gives you a powerful tool, to estimate relaibility. But if you want to have some concrete information about ONE individual device, forget it!

Away from this, you will probably never have any problems with your EEPROM, especially if you reprogram it only a few times and keep it running at room temperature. But don't be surprised if though!
It's wise to use some redundance technique, when working with EEPROMs: Store your data at more than only one location. Then you can detect any failure. If you use three different locations for each data, e.g., you can even fix the error and warn the user, that a failure has occured. But, of course, you would need three times more memory space.

Kai