Forum Discussion

Wynand_van_Nisp's avatar
Icon for Nimbostratus rankNimbostratus
May 19, 2011

Webservices auth proxy cache with millions of connectgions




We have been asked by our client if its feasible to use a LTM as a auth cache for their one app. The requirement is to cache all auth requests and take the load of the back end server. Each auth request is about 30k in size and the record that need to be cached is 100k. They get between 5-6 million transactions per day of which they recon 1 million is unique and need to live in cache for 12 hours.



The idea was to create an irule to get the requests and reply with the cached auth entries if stored in the table.



Can tables be that large and what size box would in theory be required to story a million entries of 130k in a table?





Thanks in advance.





3 Replies

  • Hi Wynand,



    Do you need to store the request and response? If so, I don't think it's practical to try to store 1mil x 130Kb of data in a session table as that's ~124Gb of data. Our biggest platforms aren't in that range of memory per TMM instance. I suggest contacting an F5 SE or partner to try to get more specific sizing information and recommendations on how you could approach this project.



  • Hi Aaron,



    Thanks for the reply. If we are asked to go ahead there will be more investigation. Is there a record count limit for tables. If we can drop the record size to something like 4k will it be possible to store up to a million entries in a table?





  • I think the practical limit is based on the amount of memory allocated to each TMM. You can check this by running tmstat from the CLI and seeing how much memory is allocated to one TMM:



    Memory Allocated


    33,777,728 / 1,814,036,480



    4Kb x 1 million is still 3.8 Gb. On an 8900, each TMM has ~1.8Gb of memory available.