Y26-050 - Performance improvements for the GBS pipeline

[StephenHulme]

Describe the Housekeeping
GbS pipeline:

• Clicking "Create" in the "add an empty GBS PCR1 plate" page takes a long time
• Selecting the "Add an empty GBS PCR2 plate" takes a long time to load and creating the plate
• Clicking "Add an empty GBS PCR2 Pool Stock tube" takes a long time to load, and then once the plate you wish to use has been scanned in, it takes a long time for that to be verified

Blocking issues
None

Additional context
Raised in the performance survey

Can be split out into smaller stories if required

[KatyTaylor]

Analysis of steps

(using master branches, and whatever was in UAT at the time - no fixes done)

Testing in local using Int Suite:

Customer-described step	Int Suite step	Time	Comment
"add an empty GBS PCR1 plate" - creating the plate	QuadrantStamp creation	14.58s
"Add an empty GBS PCR2 plate" - clicking the button	Trigger child GBS PCR2 plate creation	4.19s	Not that slow
"Add an empty GBS PCR2 plate" - creating the plate	Create tag plate	17.31s
"Add an empty GBS PCR2 Pool Stock tube" - clicking the button	Trigger child GBS PCR2 Pool Stock tube creation	14.37s
Once the plate you wish to use has been scanned in, verifying it	Fill in the pool page	21.17s	Found this step the most frustrating personally. Feels like should be quick. (And is also the slowest)

Testing in UAT using Int Suite:

Customer-described step	Int Suite step	Time	Comment
"add an empty GBS PCR1 plate" - creating the plate	QuadrantStamp creation	21.65s	1.5x slower than local
"Add an empty GBS PCR2 plate" - clicking the button	Trigger child GBS PCR2 plate creation	13.01s	3x slower than local
"Add an empty GBS PCR2 plate" - creating the plate	Create tag plate	24.30s	1.5x slower than local
"Add an empty GBS PCR2 Pool Stock tube" - clicking the button	Trigger child GBS PCR2 Pool Stock tube creation	33.34s	2x slower than local
Once the plate you wish to use has been scanned in, verifying it	Fill in the pool page	43.45s	2x slower than local

[KatyTaylor]

1st fix - "Add an empty GBS PCR2 Pool Stock tube"

Looking at the last two steps in the above table, I noticed a bug in the left hand section of the page. There is only one plate, repeated 100 times. There should be 3 distinct plates brought back from this query. The issue was in the in_state scope in Sequencescape. Fixing this reduced the time to load the page locally from ~14 seconds to ~5 seconds.

[BenTopping]

One thing to note, that we haven't on the story, is that production plates for the GbS pipeline might be much larger than the local and UAT representation of the pipeline via the int suite. Im not sure on the best way to reproduce production like data but it might be needed to help realise where some of the issues are.

[KatyTaylor]

One thing to note, that we haven't on the story, is that production plates for the GbS pipeline might be much larger than the local and UAT representation of the pipeline via the int suite. Im not sure on the best way to reproduce production like data but it might be needed to help realise where some of the issues are.

True, it may well be even slower in prod. However, all the bullet points mentioned in the description are already slow in local and UAT - so I think it's a good starting point to improve those, and easy enough to test without touching prod. Customers can report back if they have further issues.