| # |
Aug 9th 2008, 05:50 |
AD7six |
bbl |
| # |
Aug 9th 2008, 05:49 |
Howard`` |
alright then |
| # |
Aug 9th 2008, 05:49 |
AD7six |
or rather make it work then make it work better.678 |
| # |
Aug 9th 2008, 05:48 |
AD7six |
Howard``: if not don't over think - forget sharding, forget everything. make it owrk and then just add indexes - which make finding info (find all messages to this user) faster |
| # |
Aug 9th 2008, 05:48 |
Howard`` |
the idea can work i just need to figure out how to put everything together |
| # |
Aug 9th 2008, 05:48 |
Howard`` |
no it's really just me and a friend who helps me run servers |
| # |
Aug 9th 2008, 05:48 |
AD7six |
Howard``: got a dba on the team? |
| # |
Aug 9th 2008, 05:47 |
Howard`` |
sorry i'm just a little bit over my head, trying to make sense of it all, databases aren't my strong suit |
| # |
Aug 9th 2008, 05:47 |
AD7six |
also I pointed out indexes and you asked about primary keys - indexes aren't primary keys |
| # |
Aug 9th 2008, 05:47 |
AD7six |
Howard``: I'm giving you examples and you're asking about app logic |
| # |
Aug 9th 2008, 05:46 |
Howard`` |
would sync occur on a cron or immediately? |
| # |
Aug 9th 2008, 05:45 |
AD7six |
well it doesn't matter so long as it's unique and can/will never change (don't use a username). uuids are useful to share info across dbs. i.e. allowing slaves to accept registrations to later sync with the master (if there is one) and other slaves. |
| # |
Aug 9th 2008, 05:44 |
Howard`` |
I noticed virtual worlds tend to do the uuid |
| # |
Aug 9th 2008, 05:44 |
Howard`` |
what would you use for indexes on large databases? an auto_increment field, a username, or a uuid? |
| # |
Aug 9th 2008, 05:43 |
AD7six |
and it depends what you're doing and how you set it up (indexes!) |
| # |
Aug 9th 2008, 05:43 |
Howard`` |
oh ok |
| # |
Aug 9th 2008, 05:43 |
AD7six |
Howard``: a million rows isn't a lot |
| # |
Aug 9th 2008, 05:42 |
Howard`` |
isn't there a performance gain if you have 10 databases of 100k rows rather than 1 database of 1 million rows? |
| # |
Aug 9th 2008, 05:42 |
AD7six |
premature optimization otherwise. |
| # |
Aug 9th 2008, 05:42 |
AD7six |
Howard``: indexes -> dedicated db server -> clusters -> sharding. |
| # |
Aug 9th 2008, 05:41 |
AD7six |
Howard``: that's imo pretty much pointless. |
| # |
Aug 9th 2008, 05:41 |
Howard`` |
initially, after the first week or month i plan on spanning across a minimum of two servers |
| # |
Aug 9th 2008, 05:40 |
AD7six |
Howard``: are you thinking of putting (e.g.) 26 dbs, split on the same server? |
| # |
Aug 9th 2008, 05:40 |
Howard`` |
What I want to avoid is having a database with so much information that pulling 50 specific rows is going to strain the entire server |
| # |
Aug 9th 2008, 05:40 |
AD7six |
Howard``: might be directly possible have a look at schemas although I (atm) assume they expect a db to exist beforehand |
| # |
Aug 9th 2008, 05:39 |
AD7six |
Howard``: via query at a guess. |
| # |
Aug 9th 2008, 05:39 |
Howard`` |
Databases |
| # |
Aug 9th 2008, 05:39 |
Howard`` |
not tables.. |
| # |
Aug 9th 2008, 05:39 |
Howard`` |
They are, is it possible to get cake to generate tables on the fly by the way? |
| # |
Aug 9th 2008, 05:38 |
AD7six |
if arbitary shards are the goal taht is |
| # |
Aug 9th 2008, 05:38 |
Howard`` |
yeah I'm trying to understand all of this performance stuff, it's a lot to take in really |
| # |
Aug 9th 2008, 05:38 |
AD7six |
Howard``: use something like the last char (or last 2 chars) of the numeric/random id. |
| # |
Aug 9th 2008, 05:38 |
AD7six |
Howard``: using something that is none-linear to shard on will give you poor performance |
| # |
Aug 9th 2008, 05:38 |
achew22 |
AD7six: did you know that prager from Revision 3 writes in cake? |
| # |
Aug 9th 2008, 05:37 |
AD7six |
Howard``: I wouldn't set a fixed idea on the number of dbs - use a formula and factor to define the limits. so that you can say "well, 1 db isn't enough - change $n to 2, resync dbs (move things around) and off we go again" |
| # |
Aug 9th 2008, 05:36 |
Howard`` |
it's a Joint Venture as they refer to it |
| # |
Aug 9th 2008, 05:36 |
achew22 |
AD7six: damn me... cascade huh. Thanks |
| # |
Aug 9th 2008, 05:36 |
Howard`` |
These people only make money based on referrals and are tight on who they'll brand apparently |
| # |
Aug 9th 2008, 05:35 |
Howard`` |
even if the conversion rate was 1% that would be a pretty significant load from the getgo |
| # |
Aug 9th 2008, 05:34 |
Howard`` |
They're expecting to reach out to about 500k people |
| # |
Aug 9th 2008, 05:34 |
Howard`` |
well the people i'm going to be marketing through are only going to be making money through referrals |