0

I have Solr running on a large Amazon EC2 Ubuntu instance. There are a number of other Rails services running on this EC2 instance too. Today, Solr choked and had to be restarted. Here is some more information:

  • Number of documents: + 5 million
  • Throughput: + 100 rpm
  • Avg. Response time: 60 ms (and getting slower as we add more documents)
  • Slowest Response time: 2-3 seconds (this did not happen until recently - perhaps due to too Solr not coping with the amount of data)
  • Number of Solr Instances running: 1
  • On dedicated server: NO - it's shared with some other Rails services that we are running too.

My question is this: If I move Solr to a dedicated large Amazon instance and configure everything just perfectly, what performance can I then expect in terms of max. throughput and max number of indexed documents?

Thanks in advance

Darren

jensendarren
  • 383
  • 2
  • 12

1 Answers1

1

TL/DNR Version:

The only way to answer this is for you to set it up and benchmark it.

Long Version:

This question is impossible to answer for the following reasons:

  • There are multiple server levels of EC2 that you can purchase
  • EC2 is dynamic (the "E" is for "Elastic"), so your speed will vary from time to time
  • The nature of networks, especially over the WAN will dramatically impact performance
  • What does "configuring everything perfectly" mean?
  • Ubuntu is designed to be easy-to-use, and thus gives up some performance, so attempting to tune everything "perfectly" could be said to be moot

Even if you took the most egregious parts of incalculability away (namely the Cloud server in place of a physical box on your LAN), the only way to answer your question is to load up a server exactly like yours with your SOLR instance and DB, and then benchmark it.

gWaldo
  • 11,887
  • 8
  • 41
  • 68