Hi, We are a startup, that is adopting several enablers : Cosmos, CKAN, Keyrock, POI, Wirecloud, and we are currently using virtual machines from Fiware labs, for CKAN, Keyrock & POI. But for cosmos we are using the fiware cloud servers (cosmos.lab.fiware.org). We are using currently a shared webhosting, but we need more hardware and grants to run our future system.
We have calculated requirements hardware for the enablers:
RAM: 32Gb (Cosmos) + 4Gb (Ckan) + 8GbRam (SpagoBi) + ¿0.5Gb? (keyrock) + ¿0.5Gb? POI + 1Gb Wirecloud = 32+4+8+0.5+0.5+1Gb=55Gb RAM. HD : 512 GB (Cosmos) + 80Gb (CKAN) + X (SpagoBi) + X (KeyRock) + 40Gb (POI) + 10Gb(Wirecloud)=642+XGb CPU: 3 CPU (Cosmos) + 4CPU (CKAN) + 2CPU (SpagoBI) + 0CPU (KeyRock) + 2CPU (POI) + ¿1CPU? (Wirecloud) = 12CPU
Those numbers are based on the requirements found in documents from the enablers at the bottom. We expect low user quantity interaction with the platform within the next 6 months. Below 300 users.
We need to know which hardware to use, the minimum based on the requirements of the enablers. We have little knowledge in enablers, and can't make a plan without experiences in the field.
Do you think a single machine with 32gb with RAM could, 8 core, 1Tb would fit?
May be it's better to use more machines with lower specs. What do you think?
I'm sure this question is difficult to answer without knowing more information about the project. If I had a contact from someone from fiware to obtain more information, it would be helpful, but I don't know who to contact.
Thanks for your answers in advance
Cosmos platform requires the following 3 machines:
The master node machine, supporting the APIs and all the management logic of the platform. The Infinity namenode machine, supporting the naming of all the files within the permanent storage of Cosmos. The Infinity datanode machine, supporting the real data of the files managed by the Infinity namenode machine. In addition, at least 2 slave machines are required, in order to host the cluster's nodes. The exact number of slave machines depends on your needs, the many slaves you deploy, the many clusters you will be able to create.
These machines may be physical (recommeded) or virtualized on top on some hipervisor-based virtualization tool such as VirtualBox or VMWare. Anyway, the following hardware is recommended:
32 GB RAM At least 500 GB HDD When formating the file system of the machines (except the one acting as master node), take into account that a specific partition for OpenVZ (/vz) is desirable (format it to ext4).
For a city level portal with lower traffic levels:
2 * Servers with 2GB of RAM (Web and DB/solr) 80GB hard drive on both. (You may want more storage on the Web server if you intend to do a lot of file storage) Dual core processors