lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Erick Erickson <>
Subject Re: Multiple servers support
Date Sun, 25 Sep 2011 15:40:17 GMT
Well, this is not a neutral forum <G>...

A common use-case for Solr is exactly to replace
database searches because, as you say, search
performance in a database is often slow and limited.
RDBMSs do very complex stuff very well, but they
are not designed for text searching.

Scaling is accomplished by either replication or
sharding. Replication is used when the entire index
fits on a single machine and you can get
reasonable responses. I've seen 40-50M docs fit
quite comfortably on one machine. But 150TB
*probably* indicates that this isn't reasonable in your

If you can't fit the entire index on one machine, then
you shard, which splits up the single logical index
into multiple slices and Solr automatically will query
all the shards and assemble the parts into a single

But you absolutely cannot guess the hardware
requirements ahead of time. It's like answering
"How big is a Java program?" There are too
many variables. But Solr is free, right? So you
absolutely have to get a copy and put your 2.5M
docs on it and test (Solrmeter or jMeter are
good options). If you get adequate throughput, add
another 1M docs to the machine. Keep on until
your QPS rate drops and you'll have a good idea how
many documents you can put on a single machine.
There's really no other way to answer that question


On Sun, Sep 25, 2011 at 5:55 AM, Raja Ghulam Rasool <> wrote:
> Hi,
> I am new to Solr, and I am studying it currently. We are planning to
> implement Solr in our production setup. We have 15 servers where we are
> getting the data. The data is huge, like we are supposed to keep 150 Tera
> bytes of data (in terms of documents it will be around  2592000 documents
> per server), across all servers (combined). We have the
> necessary storage capacity. Can anyone let me know whether Solr will be a
> good solution for our text search needs ? We are required to provide text
> searches or certain limited number of fields.
> 1- Does Solr support such architecture, i.e. multiple servers ? what
> specific area in Solr do i need to explore (shards, cores etc, ???)
> 2- Any idea whether we will really benefit from Solr implementation for text
> searches, vs let us say Oracle Text Search ? Currently our Oracle Text
> search is giving a very bad performance and we are looking to some how
> improve our text search performance
> any high level pointers or help will be greatly appreciated.
> thanks in advance guys
> --
> Regards,
> Raja

View raw message