MetaCentrum Cloud

From MetaCentrum
Jump to navigation Jump to search

Czech version

MetaCentrum Cloud or simply „MetaCloud“ - is [1] ("Infrastructure as a Service") cloud administered by OpenNebula software.

MetaCloud can utilize all MetaCentrum or e-Infrastructure CESNET users, so all academic and employees and students and also The Czech academy of science employees. The OpenNebula web interface is https://cloud.metacentrum.cz/.

Hardware

MetaCentrum run the cloud in a cooperation with CERIT-SC Masaryk University in Brno. Available physical devices are listed in http://metavo.metacentrum.cz/pbsmon2/cloud/ and covers:

  • CERIT-SC clusters
    • hda - situated in Brno, 48 machines with 12 CPU, 90 GB RAM, 2x 600GB HDD
    • hdb - situated in Jihlava, 32 machines with 8 CPU, 128 GB RAM, 2x 600 GB HDD
    • hdc - situated in Jihlava, 112 machines with 16 CPU, 128 GB RAM, 2x 600 GB HDD
  • CESNET clusters
    • dukan - situated in Brno, 10 machines with 12 CPU, 96 GB RAM, 2x 600 GB HDD
    • warg - situated in Plzeň, 13 machines with 16 CPU, 64 GB RAM, 4x 900 GB HDD

All zegox, zigur a zapat cluster nodes are actually virtual machines maintained by PBS scheduler and allocated on physical clusters hda, hdb a hdc and their size may change in time when neccessary.

Similarly as Amazon EC2 cloud „availability zones“, also MetaCloud is divided into parts, so called "clusters" in OpenNebula. These parts behave as separate sections, and only one cluster can be choosen during image loading or network addresses specification. Ordinary users do not see these clusters in OpenNebula web interface, but it can be distinguished according to "datastore" where are disc images saved. Clusters also influence physical device, where virtual machine will run. For example, if you upload your own disc image to datastore cerit-sc-jihlava, virtual machines may physically run in Jihlava only.

Comparison of MetaCloud and other clouds

MetaCloud is an IaaS (Infrastructure as a Service) cloud built using OpenNebula v4.14.2. Pilot cloud installation provided by CESNET and CERIT-SC.

MetaCloud is similar to e.g. Amazon Elastic Compute Cloud (EC2), Microsoft Azure or Google Compute Engine. When familiar with Amazon EC2 or Google Compute Engine, following table can help you in terminology:

clouds compared
MetaCloud Amazon EC2 Google Compute Engine
web interface URL https://cloud.metacentrum.cz/ https://console.aws.amazon.com/ec2/ https://console.developers.google.com/
hypervisors KVM Xen KVM
provided disk images Debian, CentOS, SciLinux, MS Windows Amazon Linux, Ubuntu, RedHat, Suse, MS Windows Debian, RHEL, CentOS, SLES
disk image upload formats qcow2, vmdk, vdi, raw raw disk.raw in tar.gz
term for VM VM instance instance
term for VM spec (CPUs,RAM..) template AIM + instance type image + machine type
terms for VM suspend, stop and delete suspend, stop, undeploy -, stop, terminate -, -, delete
file storage service Cumulus (S3 compatible) S3 Google Cloud Storage
persistent disk storage persistent disk images, copy on start and stop Elastic Block Store, mounted over network persistent disk images

Comparison of MetaCloud and MetaCentrum

MetaCentrum (scheduler PBS) MetaCloud
Machines are administered by MetaCentrum administrators. Machines are created, run and administered by users.
MetaCetrum administrators are responsible for security of machines. Users are responsible for virtual machines security, especially for regularly software updates.
A user has rights as ordinary Unix user in all MetaCentrum machines. A user has root rights in his virtual machine.
Applications are installed by MetaCentrum employees, users can install applications only in their /home directory. Applications are installed by a user only in his virtual machine.
Jobs scheduled by PBS scheduler. No jobs at all, a user can run virtual machines and applications on them any time he wants.
Only particular combinations of CPU number and memory size are possible depending on hardware configuration. A user creates virtual machine exactly as he needs.
After script termination, data must be moved from /scratch to /home directory in the front end. There are many possibilities to save data in cloud, for example, use persistent image.

Interactive jobs in MetaCentrum may be slightly comparable to use of MetaCloud, but you can not influence time in PBS scheduler when machines will be available to you.

Whereas PBS scheduler schedule jobs to the MetaCentrum machines, users run their own virtual machines in MetaCloud and they are fully responsible for their conditions. They can also freely choose operating system and install applications.