Can't really be answered. For instance, stored data is held in *.fdt
files and is largely irrelevant to searching since that data is only
consulted for returning stored fields of the top N docs. So if your
index consists of 90% stored data it's one answer, if 10% it's totally
another. the stored data can be swapped in and out of the OS memory
space ('cause it's MMapped) with vastly less impact on your system
than other parts of the index. Certainly if you could fit it all in
memory it'd be as fast as possible, whether enough faster to justify
any extra cost is the question.
Plus you'll want to understand how much data on your particular system
is "too much" and take proactive actions when you approach that limit.
So yeah, you'll have to test. Here's a long blog on the subject:
https://lucidworks.com/2012/07/23/sizing-hardware-in-the-abstract-why-we-dont-have-a-definitive-answer/.
Skip to the section "Prototyping: how to get a handle on this problem"
Best,
Erick
Best,
Erick
On Wed, Nov 14, 2018 at 7:28 AM David Hastings
wrote:
>
> Was wondering if anyone has an idea of the ratio size of indexed only vs
> stored and indexed in solr 7.x. I was gong to run some testing myself
> later today but was curious what others have seen in this regard.
> Thanks,
> David