Scaling

Diogo Thomaz shared this question 2 years ago
Answered

Dear support,


Which architecture do you recommend for scaling?


a) one big server (one YF license) or (b) multiple medium-sized servers (enterprise license for clustering)?


For example:

a) one server with 8 cores and 32GB of memory

b) two servers where each one has 4 cores and 16 GB

c) four servers where each one has 1 core and 8 GB


And why?


License-wise the end result will much different on each scenario?


Thank you in advance,

Michalis ARGYRIOU

Aquanetix CTO

Comments (5)

photo
1

Hi Diogo,

Speaking generally, if you have one big server then that is a single point of failure whereas the more servers there are in a cluster then the higher the availability of Yellowfin. And Yellowfin offers Session Replication and so a user's session can seamlessly be transferred to another node by a load balancer, so in the event of hardware failure of one of the nodes the user won't be aware anything has happened.

Regarding licencing, yes to set up clustering you will need a special clustering licence that lists all of your nodes, however, whether the costs are equivalent or different for each scenario you will definitely have to discuss this with your account manager as this is not a responsibility of support.


regards,

David

photo
1

Dear Dave,


Thank you for the answer. Fault-tolerance wise you have covered me. How about scaling and performance-wise - which approach is better? For example, due to JAva Garbage Collector until how many GB may I use? Can you forward this question to L2 support?


Kind regards,

Michalis

photo
1

Hi Diogo,

That's no problem. We don't really have such a strict regime of levels here, sometimes I am looking at what are generally considered L3 cases and other times looking at L2 or L1 just depending on what is required, however I will forward your request to another support colleague as requested.


In the meantime while you are waiting for one of my colleagues to take over this question you may like to read our documentation on Server Specifications.

Also, from reading your question "due to JAva Garbage Collector until how many GB may I use?" I think you may be interested to read Oracle's documentation on said subject matter because you will discover that in fact the GC is highly configurable.


regards,

David

photo
1

Hi Diogo,

Dave is right on the money here. As far as scaling goes, the option you choose is going to be highly dependent on your scenario.

If you desire fault tolerance, you will want multiple nodes.

As far as scaling resources, the only consideration for multiple nodes is how much activity the server is going to have. This should not only include concurrent logins, but also broadcast and task traffic. If you plan on an environment that will host many users at the same time, as well as consistent broadcasts throughout the day, you'll want to cluster your environment. This allows you to offload the tasks and broadcasts to a specified node.

Does this address your question?

Thanks,

Ryan

photo
1

Dear both,


The servers specification is a good start since it provides some high level guidelines.

Thank you for your time,


Kind regards,

Michalis ARGYRIOU

Aquanetix CTO