High RAM machine


I’m aware that can’t have too many CPUs assigned to one instance due to the size of the host machines, but is there a way of increasing the RAM for a machine to e.g. 256 GB? While leaving the number of cores the same.



We’re somewhat limited by the available RAM on the hypervisors that host the instances.

The standard flavours of VM that are currently available (4 vCPU / 16Gb RAM; 8 vCPU / 32 Gb RAM) are designed to balance the CPU/RAM resources of the hypervisors, so that they nicely divide up into a whole number of instances, while still providing a relatively powerful machine.

An 8 vCPU / 256Gb RAM flavour would be unbalanced in favour of RAM and would mean we could fit less instances onto our computing resources, so we don’t plan to offer this as a standard thing.

That said however, we do have extra capacity, outside of our regular hypervisors, for high-memory VMs. We are aiming to provide access to these on a time-limited, request basis.

You’re the first requester, so we don’t really have a system in place for this yet, but if you send me a PM we can work it out!



I am about to embark on a comparative study of classification tools for metagenomic analysis, to find the most suitable for a study conducted on sterile sites. I was wondering if it would be possible to be temporarily assigned a high RAM machine so that I can build databases for each of the programmes I will be testing. (I am assuming that once they are built I will be able to use them on a lower RAM machine)



I’m sure that we could help with this, please PM me and we can start to work the specifics out.

We’re also building a variety of databases for metagenomic classifiers ourselves, which we would make available to download by the community. We currently have databases of varying sizes and Kingdoms for Centrifuge, Kraken, and GOTTCHA, but it would be great to be able to provide more for use by the community.

If you could let me know which classifier(s) you’re looking at using, and what taxa you’d like your database to contain, we could build and host the database files for your experiment and the rest of the community.


Hi Matt,

I am trying to build custom kraken databases. I saw this thread where you mentioned that databases for kraken have been set up already. Could you please give me some more information about this.



Hi Anu - I’m building Kraken databases at the moment.

I’m hoping to build a database of all RefSeq Human, Bacteria, Viruses, Archaea and Protozoa built using some tweaked versions of Mick Watson’s scripts. I’m going to try to shrink it small enough to fit into the RAM of a CLIMB group instance (<64 Gb).

I’ll update this post when I’ve got a database finished.


Thats great to hear! But is it possible to add fungal genomes to this database as well?



Hi Anu - yes, my mistake, missed fungi off the original list!

You’ll find that some of those branches have quite poor representation in complete RefSeq genomes, so be careful when interpreting results because, as an LCA technique, Kraken is susceptible to how complete your database is.

I recommend trying Centrifuge with the nr database in parallel to Kraken to make sure you can replicate your results with two techniques.