Technical details of read and write path?

[Sebastian Gebauer]

Hi,

I'm currently writing my master theses and a main part of it is to compare the read and write paths of mongodb with cassandra.

Is there any page that describes how a read and/or write of an document works in details?

Thank you!

[Keith Bostic]

On Sunday, August 9, 2020 at 1:14:03 PM UTC-7 sebastian...@gmail.com wrote:

 

I'm currently writing my master theses and a main part of it is to compare the read and write paths of mongodb with cassandra.

Is there any page that describes how a read and/or write of an document works in details? 

I would suggest contacting the MongoDB server team by filing a JIRA ticket at https://jira.mongodb.org/. Your question isn't really a WiredTiger question, there's a lot of work that happens in MongoDB before reaching the storage layer.

[Michael Cahill]

Hi Sebastien,

Also consider watching this presentation: It's 10:00 pm: Do You Know Where Your Writes Are?

Kind regards,

Michael.

--
You received this message because you are subscribed to the Google Groups "wiredtiger-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wiredtiger-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/wiredtiger-users/fe300b01-4537-4c72-b8e8-c6c6ecec35afn%40googlegroups.com.

[Sebastian Gebauer]

Thank you both for your input! It's been a honor for me to get a response from both of you.

I would like to get an overall picture, so what I'm searching for is something like this:

https://www.percona.com/sites/default/files/presentations/Monitoring-MongoDBs-Engines-in-the-Wild.pdf (Slide Number 9).

We are trying to figure out how many page reads/writes (in terms of logical and physical reads/writes) are necessary to update a single document. I was trying to understand the logical and physical reads/writes by the WiredTiger Cache Stats (with mongoDb collStats/serverStatus) but I'm having a hard time to understand how that metrics works. The mongodb documentation says that if you want to have an overview of the I/O Operations you should have a look at

• wiredTiger.cache.pages read into cache (I would understand that as physical reads?)
  
• wiredTiger.cache.pages written from cache (This is my understanding for physical writes?)


I drafted a simulation with 5M entities. In the simulation, a new property is added to each document (average object size was 98 before adding the property, 117 afterwards).
But when I'm restarting mongodb (heating up the cache) and execute the simulation the metrics will read as:   

        "wiredTiger.cache.pages read into cache": 4452,
        "wiredTiger.cache.pages written from cache": 7117,
        "wiredTiger.cache.pages requested from the cache": 20165168

How is it possible that there are many more pages requested from the cache (logical reads?) than initially have been read into cache?

Sorry for my bad english - I'm a german native speaker...