running time too long

[nsuarezg]

hello, I am looking for help regarding the use of KrakenUniq. I am planning to run samples containing between 20 and 40.10E6 reads (150PE) using the MicrobialDB. According to my first trial and estimations, each sample will take between 10 and 20 days to be classify. The same samples only need a few hours run when using Kraken2. I am copying the DB to the local disk before the run and use the script below. Can you help me in order to decrease the calculation time? Thank you in advance.
Nathalie

[salzberg]

You probably don't have enough RAM on your computer to load the DB, which forces it to swap. This can take forever (or 100s of times longer). First make sure you have enough RAM (should be larger than the DB) and also try adding the parameter --preload to your krakenuniq command. That will force it to load the whole DB in memory.

[nsuarezg]

Thank you for your reply. I have enough RAM. The DB is copied and extracted without any problem (takes 2h). But then the classification takes forever (for instance 1h to classify 2,4Mb files (80.000 reads)). I tried the --preload, but it did not improve the process.

[salzberg]

this should only take a minute to classify. You might have enough RAM but your OS might be set so that you can only use 50% of it for 1 process - this is the default for many Unix installations. You could turn off swap (requires a reboot). Or try using --preload-size and only give it, say, much less than 50% of the RAM you think you have. I'd recommend trying 32G first, see how that works. You didn't say how much RAM you have so I'm guessing.

[nsuarezg]

Thank you again. I will try and come back to you :-)