For more information regarding the security incident at F5, the actions we are taking to address it, and our ongoing efforts to protect our customers, click here.

Forum Discussion

Renato_166638's avatar
Renato_166638
Icon for Nimbostratus rankNimbostratus
Nov 27, 2015

Cache size, configured vs observed

Hi,

I was observing the web-acceleration profile I've configured and I noted an huge difference between the cache size configured and the observed. Here go some examples:

Configured parameter: "cache-size 225mb"
Value observed by 'show ltm profile web-acceleration': "Cache Size (in Bytes)           1.4G"

Or

Configured parameter: "cache-max-entries 300000"
Value observed by 'show ltm profile web-acceleration': "Total Cached Items            628.1K"

The question: is it normal? is it possible to calculate and predict this difference?

application delivery
BIG-IP
LTM

4 Replies

  • boneyard's avatar
    boneyard
    Icon for MVP rankMVP
    Nov 28, 2015
    the first one i can't explain, but can't the second one be the total number of items it ever cached? go through the manual :)
  • Renato_166638's avatar
    Renato_166638
    Icon for Nimbostratus rankNimbostratus
    Nov 28, 2015
    Ok, but I forgot to mention that the **Eviction was zero**, the memory usage was going crazy and I uploaded two TMM's crash dumps to support team a few hours after posting this. Too many coincidences.
  • theCook's avatar
    theCook
    Icon for Employee rankEmployee
    Nov 30, 2015

    The real max-size should be the configured max times the number of TMMs. So, in this case if you have 8 TMMs, this profile should have a maximum size of 1800mb. I believe the same is true for max-entries. If you haven't dug these up already, here's some useful links regarding this topic:

     

    This solution discusses the ( of tmm instances times configured value = actual value) mentioned above: https://support.f5.com/kb/en-us/solutions/public/8000/700/sol8788.html

     

    Discusses a DB key that limits what you're going to be able to configure for cache sizes https://support.f5.com/kb/en-us/solutions/public/13000/800/sol13878.html

     

    Discusses some cache improvements in 10.1.0. Might be useful. https://support.f5.com/kb/en-us/solutions/public/11000/900/sol11990.html

     

  • Renato's avatar
    Renato
    Icon for Altostratus rankAltostratus
    Dec 01, 2015

    Well, theCook, this is a great point! I really forgot to consider this hypothesis. A lot of information are really stored per TMM instance in order to avoid concurrency and locking between processes on memory access. In this case the costumer asked me to enable cache for one transaction that receives random parameters, it reaches a huge cardinality and the cache hit ratio is not that high.

     

    I don't know if the increase of memory usage was really related to those TMM crashes we observed, but they happened just after enabling the web acceleration profiles and the device was under a high load. The profiles are still enabled and the device is working properly, but we are not facing that high load anymore.