[DOC-9267] Guidance for Search Service Sizing#4078
[DOC-9267] Guidance for Search Service Sizing#4078sarahlwelton wants to merge 9 commits intorelease/8.0from
Conversation
|
|
||
| Based on these variables, the required vCPUs could be either: | ||
|
|
||
| * stem:[290], using a value of 300 in the vCPU calculation. |
There was a problem hiding this comment.
We should have two cases for 150 & 200 respectively instead of 200 & 300 as 150 & 200 are the values of QPS/vCPU that we have documented earlier
|
|
||
| |==== | ||
|
|
||
| Based on these variables, the required vCPUs would be stem:[40], based on the more complex queries needing a higher QPS per vCPU and using a value of 300 in the calculation. |
There was a problem hiding this comment.
Same here, the value should be 200
There was a problem hiding this comment.
@sarahlwelton I've some comments around search index nomenclature and requirements for sizing right.
I think the estimation itself @TusharMadaan04 's done nice work on and you've captured it quite well here.
| Search Service nodes manage Search indexes and serve your Search queries. | ||
|
|
||
| Basic Search indexes are lists of all the unique terms that appear in the documents on your cluster. | ||
| For each term, the Search index also contains a list of the documents where that term appears. |
There was a problem hiding this comment.
It'd be good to use the nomenclature - inverted index here .. which is the primary data structure within the search index that indicates the list of the documents where the term appears.
There was a problem hiding this comment.
Sure, that's easy enough to add
|
|
||
| Basic Search indexes are lists of all the unique terms that appear in the documents on your cluster. | ||
| For each term, the Search index also contains a list of the documents where that term appears. | ||
| These lists inside a Search index can cause the Search index to be larger than your original dataset. |
There was a problem hiding this comment.
This is a possibility, it depends very much on the data though .. in many a situation it could be lesser too because we de-dup terms/prefixes/sub-strings/suffixes in the index and the inverted-index/postings-lists are compressed bitmaps alongside a map for doc keys.
Will defer to your judgement on how to frame all of this :)
There was a problem hiding this comment.
Perhaps this framing works better, just to keep things accurate?
…lude scenarios for numeric and geospatial sizing
…formatting. Repeat that best sizing results come from support for numeric and geo workloads
sarahlwelton
changed the title
simon-dew
left a comment
simon-dew
left a comment
There was a problem hiding this comment.
LGTM 👍
One observation about the rendering of the STEM formulas. This probably falls outside the scope of this PR, and may be a question for @osfameron. Not a blocker.
There was a problem hiding this comment.
reminder to delete before merging
There was a problem hiding this comment.
Was keeping around in case there was any review comments - which there was :)
|
|
||
| For example, if your target QPS is 30,000 and your queries are less complex, divide your total QPS target by 200 to get your required vCPUs: | ||
|
|
||
| [stem] |
There was a problem hiding this comment.
It takes a couple of seconds for the LaTeX to be rendered when you first load the page. As it's rendering, the stem formulas are displayed with large whitespace margins, which makes the page pretty hard to read until rendering is done. Would asciimath any faster?
There was a problem hiding this comment.
I can change these to AsciiMath but I also can't say I had the same problem - they've already rendered for me by the time the page loads up. I wouldn't be able to test this to confirm or not.
There was a problem hiding this comment.
Fair enough — in that case I would say merge this as is, unless there are any other SME comments. The STEM rendering is a slippery problem and not an issue to be fixed in this PR
4 participants
The long-awaited guidance for sizing the Search Service in a Couchbase Server deployment.
Preview URL:
https://preview.docs-test.couchbase.com/docs-server-DOC-9267-fts-sizing/server/current/install/sizing-general.html#sizing-search-service-nodes
You will need the Docs Team credentials on Confluence.