Sign In Register

How can we help you today?

Start a new topic
Answered

Storage per user

I'm trying to understand the "storage per user" metrics - it's way higher than I would expect. It's been sitting at 1MB for a while and recently jumped to 10MB (probably because I cleared out a bunch of dummy users from the player collection).


But even 1MB seems excessive - we store highscore across a handfull of leaderboards and a small JSON struct on each player in addition to the players default properties. 


If I were to make a worst case estimate, I would have guessed maybe 5K per player.


Is this simply because we're still in test mode with only a handfull of users and the base overhead of the data storage mess up the per-player numbers?


If so, is there any way to separate the "static" base storage number from the "dynamic" per-player number so I can get an actual usable figure?



Best Answer

Hi Niels,


Yes, removing players would result in you seeing an increase in the per-player storage, and in fact the storage is calculated per active player so if you were to even stop using players over the period you are viewing the storage would look skewed as the number of active players dropped.


There is a base overhead but it is really very small, even over a relatively small number of players it would be almost negligible.  The storage is calculated as the data from your collections in your game database, any uploadables and downloadables that you may add, and then there is also the analytics data that we store.  With the analytics data we try and balance being able to provide useful metrics against the volume of data that we store and this is something we are constantly refining.  Additionally, we do schedule periodic cleanup jobs for this analytic data that aims to remove data that is no longer relevant (we've manually kicked this off for you to prune some of this down, so you should see a drop shortly, but generally this would occur as a BAU task anyway).


Hopefully this answers your question, let us know if you have any further question,


Dave


Answer

Hi Niels,


Yes, removing players would result in you seeing an increase in the per-player storage, and in fact the storage is calculated per active player so if you were to even stop using players over the period you are viewing the storage would look skewed as the number of active players dropped.


There is a base overhead but it is really very small, even over a relatively small number of players it would be almost negligible.  The storage is calculated as the data from your collections in your game database, any uploadables and downloadables that you may add, and then there is also the analytics data that we store.  With the analytics data we try and balance being able to provide useful metrics against the volume of data that we store and this is something we are constantly refining.  Additionally, we do schedule periodic cleanup jobs for this analytic data that aims to remove data that is no longer relevant (we've manually kicked this off for you to prune some of this down, so you should see a drop shortly, but generally this would occur as a BAU task anyway).


Hopefully this answers your question, let us know if you have any further question,


Dave

What I'm looking for is some kind of metrics to use as a guide during development. I.e. something that also works when I have only a handful of test users. Right now the average is not very useful because I'll only know if I have a problem when I have 100.000 users at which point it might be too late to fix it.


Right now all my server-side development is feature driven, with no consideration for performance, because I have only a generic idea of how the things I add affects performance. (As in: "searching for stuff is expensive", "collections that can grow infinitely is a bad idea", etc.)... I could very easily do something really stupid and not know about it.


The average response times helps a bit in understanding code performance, but I'm pretty much in the dark on memory usage.


Maybe it would help to have the information as "average per collection"?

Hi,

A year later I have the exact same question. I see sometimes 80Mb of storage per player and normally is 15Mb when seeing hourly, but if I set the date range daily I see  around 3Mb. I don't understand two things from this:

  1. My estimate per player is around 500k
  2. Why if I select a date range of one day only I get different values if I see it hourly or daily
Kind regards

Login to post a comment