What kind of server is required for this volume?

[Bruno Mannina <bm...@free.fr>]

Hi,

 

I know that it depends of several parameters to have a good answer for this
question,

But maybe with these information you can show me the right direction.

 

I have actually : 

* Hardware: Intel Xeon-E 2288G - 8 c / 16 t - 3.7 GHz / 5 GHz - 128GB -
SoftRaid 3x4To SATA

* SOLR 8.7 Standalone (not solrcloud)

* 19 collections (185Go, 100Mo, 465Go, 453Go, 54Go, 56Go, 27Mo, 136Ko,
437Mo, 146Mo, 14Mo, 2Go, 403Mo, 25Ko, 19Go, 613Go, 656Go, 36Ko, 695Go)

Big collections (xxxGo) have from 2.5M docs to 131M docs, docs size are from
1ko to 50ko max 

* Xmx = 30Go

* Around 250 requests by day (generally not at the same time)

* Databases are updated 2 times by week (new and amend doc) around 200 000
docs by week for big collections.

* Around 35 fields (stored/indexed) for all big collections, fields are
string, text_en, pdate, text_general

* Requests doesn't use faceting (never). Biggest requests contains 50 terms
and Boolean operators. Sometimes proximities.

* Q and FQ are often used.

* Clustering query also (carrot2 v3 (not v4))

* Sometimes we use multi collection requests

 

Actually it works fine without OOM but queries takes more than 15s to get an
answer. (and few ms if we re-execute the same query)

Query can answer from 1 to 50 000 doc but we get only the first 25 docs.

Actually solrconfig params are by default.

 

I'm trying to increase the solr query speed and I'm looking for a new
server.

For this kind of volume, I don't think it's possible to increase speed just
by software? Right?

 

Is it necessary to have more than 128Go RAM ? if most of collections are
bigger than 128Go.  

 

Many thanks for your help to guide me to the right configuration.

 

Regards,

Bruno

 



-- 
L'absence de virus dans ce courrier électronique a été vérifiée par le logiciel antivirus Avast.
https://www.avast.com/antivirus