Consider dropping the content_cache_* and core_item_markers tables

[Diskusjon]

Hello,
we're currently working on reducing the MySQL-traffic from our forum. So we're thinking about rewriting the IPSContentCache class to use the value in the $INFO['use_[memcache|diskcache|*'] variable, in our case memcache.

We're also considering moving the core_item_markers to memcache, since everything is stored in core_item_markers_storage anyways, since that table have huge amounts of traffic.

Is this something that would be relevant for future releases of IPB? Have anyone done anything similar?


Kind regards

[bfarber]

Hello,

we're currently working on reducing the MySQL-traffic from our forum. So we're thinking about rewriting the IPSContentCache class to use the value in the $INFO['use_[memcache|diskcache|*'] variable, in our case memcache.

We're also considering moving the core_item_markers to memcache, since everything is stored in core_item_markers_storage anyways, since that table have huge amounts of traffic.

Is this something that would be relevant for future releases of IPB? Have anyone done anything similar?

Kind regards

If you do end up doing this, I'd be extremely curious how well it works out. I would like to expand caching to an extent in a future version, however there are always tradeoffs. For instance, content_cache_* can get rather big, so you'd need a lot of space in memcache to store all of it. Same with core_item_markers table.

Still, if you make such changes let me know how they perform and how much memcache space you needed. To my knowledge, this has not yet been done.

[Diskusjon]

It's probably gonna require a whole lot of space from memcache. But we are running like 7 memcache-servers, so I'm thinking this solution is for larger boards only (10 million - 20 million posts).

But I will give you guys a heads up when we have tested this.

Tt seems like it requires alot of rewriting in the forums-application modules, since it uses IPSContentCache for joining tables.