QVS Cache Warming - usage efficiency and scenarios

[schtierlietz_1]

Hi there,

I read on community about the features of using cashe warming, but in the last years there have been no publications on the topic. 

Perhaps it is because the feature is not that efficient and there are new tools for QlikView Server apps?

Is someone using cashe warming on QVServer for perfomance now? Does it exist feature like out of the box? For information security reasons, we cannot use macros/PowerShell on a production server. Does anyone have a similar experience?

[rwunderlich]

Perhaps a reason you don't hear much about cache warming is that it is a technique that should rarely (in my opinion) be used.  Better to fix the underlying problem of application performance or server resource allocation and let the cache algorithm manage the cache based on dynamic usage patterns. 

I have only used cache warming as a short term band-aid to provide time to systematically correct the underlying problem(s).  Left in place or improperly implemented, cache warming can create new problems. 

-Rob 

[Maria_Halley · Accepted solution]

@schtierlietz_1 

QlikView has a function called preload. It means that the document get loaded after every restart and stays in memory. This can be set per document.

Is that what you mean?

[schtierlietz_1]

Thank You.

No, I mean that the QlikView engine takes the results of calculations in a cache, which is used by other users. As a result, calculations are performed for them faster, since the result is taken directly from the cache.

[Maria_Halley · Accepted solution]

@schtierlietz_1 

QlikView already does this. It saves cached results until the Working set limit is reached, then it starts clearing out the oldest values.

But there are no way that I know to save the cache when restarting the service, if that is what you mean

[schtierlietz_1]

Yes, it is known. The task looks like this - we make 90% of the most requested selections by a robot,  QLikView will automatically create a cache for most possible user actions. Thus, the result of most user actions will be obtained from the cache. After clearing the cache/ apps reloading, we repeat the steps to fill the cache - warming up.

[Maria_Halley · Accepted solution]

@schtierlietz_1 

I thought of another this to maybe consider. 

It might be a good idea to set the documents as preloaded anyway. A preloaded doc does not get unloaded from memory when document timeoute is reached like "normal" documents. The cache will not be cleared when the document is unloaded, but it will still be slower for the user to load the doc.

I'm guessing these docs are used enough to not reach the document timeout but I just wanted to mention it.

Also QlikView will start deleting cache, starting with the oldest unused information, if the working set limit is reached.

So it is important to make sure that the working set limits are correct, default the are 70% and 90%, but id the Server have a lot of RAM these should be adjusted.