lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shawn Heisey <>
Subject Re: Need Support - Apache Solr - 20180915
Date Sat, 15 Sep 2018 17:22:47 GMT
On 9/14/2018 9:51 PM, senthil wrote:
> We are beginners to Apache Solr and its implementations. We need the 
> following basic clarifications regarding Apache Solr usage and 
> implementing with MS-SQL server database.

I don't know what you think of Erick's answers, but he's right on the 
money with everything he said.  Here's my contribution.  Just more detail.

> 1. Our MS-SQL server database having the data table which contains 20 
> columns with billions of data.

MS SQL probably means your environment is Windows Server. If you can, 
run Solr on something other than Windows.  Solr can run just fine on a 
Server edition of Windows, but it will run better on something else.  
Open source operating systems will serve you very well.

> 2. How to implement Apache Solr in the particular above table to 
> increase search capability?

Setting Solr up to import from a database is not terribly difficult.  
Where you will probably spend the most time is perfecting your field 
analysis in your schema.  Getting that right can take a lot of 
experimentation, rebuilding the index every time you change something.  
You probably don't want to import your whole database table every time 
while you work on this step.

There are certain gotchas when using the DataImport Handler with 
SolrCloud.  You'll be happier with Solr if you can build your own 
program to transfer data from your database into Solr.  With a 
multi-threaded indexing application, you can achieve import speeds far 
greater than DIH can.

> 3. Is there any way to call the data which is distributed across 2 
> shards/node of Apache Solr at a time?

As Erick said, this is where SolrCloud shines.  You can do sharded 
indexes without SolrCloud, but it is much more difficult to manage.

> 4. Is there any performance difference between search the data in a 
> single shard/node and multiple shard/node?

I'm not sure how to approach this question - mostly because I cannot 
tell exactly what you're asking.  Are you asking about multiple shards 
per node, or multiple shards in general?

The short answer is yes in either case.  And if all you want to know is 
whether a performance difference EXISTS, then the answer is yes.  The 
long answer, like MANY questions about Solr, is "it depends."  If, in 
addition to whether a performance difference exists, you want to know 
which way has better performance, the answer is still "it depends."

If your query rate is VERY low, splitting into multiple shards on the 
same node can actually perform BETTER than a single node on the same 
machine.  As your query rate grows, you'll want those shards to be on 
separate machines, or query performance will suffer.


View raw message