The Vertica Forum recently got a makeover! Let us know what you think by filling out this short, anonymous survey.

Choosing more less powerful nodes or fewer powerful ones

What is the thinking behind this? Is it cheaper to run a cluster of many commodity nodes? What are the cons/pros to think about? What did you find in your own production environments? Let's say an example if we're choosing between 2 nodes with 12 cores and 6 nodes with 4 cores. Would my hardware be cheaper? Would performance on both be similar or the powerful ones would win because they would need to transfer less data over the network? Would the more powerful cluster be more problematic as there's less machines to manage and there's fewer failures long-term assuming same quality of components. Any answers appreciated!

Comments

  • Hi, First of all to build a Vertica Cluster you need atleasty 3 nodes. Few points to always remember 1. Vertica is higly linearly scalable 2. Vertica relies on the Distributed computing power of all its machines So Firstly, If you have a cluster of 6 nodes with quad core each. that would be better. Try to balance the cluster by keeping the configuration of all nodes similar, this ensures the query respose time to be minimum as all your nodes work on the query in the same manner. So there will be no lag from any node to fetch the result set. Also Cores matter whn Vertica deals with its COPY function. Secondly, The data transfer in your cluster is taken care by the Custer network which is dependent on the Switch you use for your cluster and the speed supported by each of your nodes. Mostly now a days the nodes come with 10 GB support speed, So make sure you have a healty cluster network. Always remember "the more the nodes the faster the query" Hope this answered you query
  • Hey Navin, Thanks for your answer. It helps me understand that more nodes is better but I still don't understand the technical reasoning behind it. Why exactly the 6 node cluster would be better than a 3 node cluster which has the same total IO, CPU power of a 6 node cluster? Thanks!
  • Hi, There will a better parallel execution of queries in more number of nodes then in less.

Leave a Comment

BoldItalicStrikethroughOrdered listUnordered list
Emoji
Image
Align leftAlign centerAlign rightToggle HTML viewToggle full pageToggle lights
Drop image/file

Can't find what you're looking for? Search the Vertica Documentation, Knowledge Base, or Blog for more information.