System requirements (RAM/CPU) for INCEpTION

[Serge Sotnyk]

Hi all!

I've trying to organize the annotation process for my company. I use INCEpTION 0.15.2.

We need only in NER layer right now. There are 7 equal projects with 109 documents (2.3 MB of UTF-8 plain texts), 1 project per entity type.

When I was testing the installation, the INCEpTION was working well. But our annotators (max simultaneous worked annotators can be 4 people) complain that their working is too slow. Pauses between page switching and an entity marking can reach 20 seconds.

I've read the Administrator Guide (system requirements), but I have not found the recommendations about minimal and recommended hardware. It contains recommendations about software only.

Right now, we use AWS instance with 2cpu 4GB. java -xmx2700m

Could anyone give me advice about recommended resources? Or maybe I should look in another direction?

Thank you!

[Richard Eckart de Castilho]

Hi,

> On 3. Jun 2020, at 15:18, Serge Sotnyk <serge....@gmail.com> wrote:
>
> Right now, we use AWS instance with 2cpu 4GB. java -xmx2700m
>
> Could anyone give me advice about recommended resources? Or maybe I should look in another direction?

The resource usage depends on a number of factors. My guess is though that it may not be memory or
CPU that is limiting here, but rather latency and bandwidth. There are a few optimizations underway.

If you are brave and like living on the edge, you might care to grab a recent CI build of the
INCEpTION master branch to see if that improves your experience:

https://zoidberg.ukp.informatik.tu-darmstadt.de/jenkins/job/INCEpTION%20(GitHub)%20(master)/

Mind that we are in a phase right now were we pump in a few systemic improvements that haven't
seen much testing so far. They will make things better, but it can well be that other things
turn out to be broken and still need to be fixed - even for CI builds that come up green.

If you try, please let us know.

Cheers,

-- Richard

[Serge Sotnyk]

Hi Richard,

Thank you for your answer.

No, I'm not so brave to update the projects right now :)

I'll try to finish the current annotation with this stable version, but next time I'll certainly use the fresher version of INCEpTION.

Regards,

Serge.

[Serge Sotnyk]

Hi Richard,

It looks, like 4 cpu and 8 GB RAM is enough for our tasks (4 annotators + 1 curator simultaneous).

Regards,

Serge.

[Jake Foobar]

Hi Serge,

Have you tried to scale beyond 5 concurrent users? If so, is the platform stable with  good performance? We are looking at scaling up eventually to 100+ concurrent users. 

Thanks in advance!