Thank you four your responses. Andrew I'll check the link out today...
2011/4/8 Dave Caroline <dave.thearchivist@stripped>
> You will get far more performance by design and tuning of indexes and
> queries than you will by changing to ndb
I need scalability, so I think that ndb is the only way. My question, in the
end, is "Why could be the cluster giving worse performance than the
standalone mysql, if I have only selects, inserts, updates and deletes?" I
have tried configuring ndb and mysql but the performance doesn't change at
> 5 qps is a very low load for a database server normally,
Remember that I said 5qps per user, if I have 1000 users (possible) it will
> I followed your link at the bottom of your email, are you putting
> video data through mysql? best not to do that else use an engine that
> is designed for the job where the user connects direct to the stream.
Oh, no, I'm not doing that, the video is only a little part of the system
i'm working on right now. The webpage you visited is a very small tool, only
video and chat. Interesting link, though.
> Most performance problems can be addressed by correct system tuning
> and query indexing,
> this can give a 1000 times improvement to a badly performing query
> where other methods may only double performace and give you only a
> temporary improvement while you fix the real design problems
That's just i'm looking for! double or triple performance to have temporary
improvement, but with the cluster I'm having poorer performance. I would
like to feel that with 2 servers mysql do the job twice as fast, that's it.
As I said, I know that I have to improve my queries, but I really have to
get into it because I'm not an expert on the subject, so i would like to
have easy scalability through a mysql cluster. Sadly, there is no much
documentation about mysql cluster tuning, and i need scalability.
I guess that i will start to read about indexing and improvement of queries
> Dave Caroline
> On Fri, Apr 8, 2011 at 4:08 AM, Ivo Perich <iperich@stripped> wrote:
> > Hi, I have an application working on Amazon EC2, on a "Large (m1.large)"
> > instance, that is a 64-bit platform with 7.5GB 4 EC2 Compute Units (2
> > virtual cores with 2 EC2 Compute Units each one, whatever that means)
> > I have a web app with a lot of users, every one making about 20-25 querys
> > every 5 seconds, this is 4-5 querys a second. I know that its maybe too
> > much, but I dont have enough time to re-design the app because it is very
> > complex stuff, almost a web OS. Besides, EC2 instances are cheap. In the
> > beggining it was intended for up to 100 concurrent users, but that number
> > increasing fast. With the biggest EC2 instance i can handle about 300
> > I have only flat data, selects, updates, inserts, deletes, and that's it,
> > joins , no complex stuff. I installed mysql cluster 7.1, with 1 ndb
> > and 2 data nodes, for test all the stuff. The app runs ok, i can see the
> > data from both nodes, but the performance is worse than using normal
> > (the users updates overall time is about 2-3 times the normal). I have
> > configuring [mysqld] sections in the cluster manager and the nodes, but
> > performance doesn't change at all. I have read a lot of info about this,
> > I know that is not a simple problem to configure correctly the cluster.
> > my question is if the mysql cluster will work for this kind of stuff, and
> > what kind of configuration can I try for testing, I'm a little lost here.
> > is all about the config files and optimize my queries? can I try another
> > thing?
> > Other question, if I use a data node as a manager node, it is bad in
> > of performance? Just for know.
> > Thanks in advice.
> > --
> > Regards
> > Ivo Perich
> > www.ubiqq.com