Automatic Visibility/Observability

Dear community,

I am a few month in with Instana and have a question regarding the automatic aggregation of short calls. Here is the short documentation:

While I understand the intent of Instana, in my use case, this may lead to loss of important information or potentially misleading labelling. Within my use case, I observe Webservice request against my company's software. In these requests I might have 10-20 database calls for each, which are mostly below 1 ms in duration and their time between invocation will be often less than 10 ms (since these Webservice equests take normally less than 60 ms). So I often have a database call, which batches 10-20 database calls and I just get the longest sql statement (with table name) as the label of these batched calls.

This means two things for me:

• It may be misleading to some, that ALL of the batched calls belong to the label (and the label may be only one call of the batch, mostly its longest) and that the number of calls to the table of the label may be inflated.
• The information of the subcalls ist mostly lost (apart of how many). I just get aggregated numbers. That means, if I wanted to analyse to which tables a high amount of sql calls goes, it is not possible anymore. Admittedly, they are then very fast, but if I want to clamp down on their number, it does not help.

Am I correct in my two assumptions? Or is the data of subcalls still stored somewhere else and I can access it in another view?

Thanks a lot for help!

Thomas

Hi Thomas

thank you for reaching out. 

In short you are correct and no, the data is not stored anywhere (doing so would mean we have to get the data after all which imposes the perf penalty we try to avoid in the first place).

Most (all?) measurements in general have the opposite goals of precise vs non-impacting. Think also about a classical profiler as example. The higher the sample rate, the more precise the outcome but also the bigger the overhead.

Instana is not designed to give you ALL answers you might have regarding DB performance. It gives you a lot but for example execution plans and other very technical details are always hidden from Instana. Is your DB a "classical" SQL derivative like Oracle, mysql, postgres etc? If yes, have a look at DBMarlin, https://www.dbmarlin.com/ , one license is free for all Instana customers ;-). It is a sleek and very effective tool to drill deep into DB perf issues. Plus they integrate with Instana.

Cheers

Tom

Dear community,

I am a few month in with Instana and have a question regarding the automatic aggregation of short calls. Here is the short documentation:

While I understand the intent of Instana, in my use case, this may lead to loss of important information or potentially misleading labelling. Within my use case, I observe Webservice request against my company's software. In these requests I might have 10-20 database calls for each, which are mostly below 1 ms in duration and their time between invocation will be often less than 10 ms (since these Webservice equests take normally less than 60 ms). So I often have a database call, which batches 10-20 database calls and I just get the longest sql statement (with table name) as the label of these batched calls.

This means two things for me:

• It may be misleading to some, that ALL of the batched calls belong to the label (and the label may be only one call of the batch, mostly its longest) and that the number of calls to the table of the label may be inflated.
• The information of the subcalls ist mostly lost (apart of how many). I just get aggregated numbers. That means, if I wanted to analyse to which tables a high amount of sql calls goes, it is not possible anymore. Admittedly, they are then very fast, but if I want to clamp down on their number, it does not help.

Am I correct in my two assumptions? Or is the data of subcalls still stored somewhere else and I can access it in another view?

Thanks a lot for help!

Thomas