usage of cloud platform in datacenter bitp for resource and energy efficiency
DESCRIPTION
Usage of cloud platform in datacenter BITP for resource and energy efficiency. Oksana Shadura , V.Pelykh , S.Svistynov. Bogolyubov Institute for Theoretical Physics Ukraine. Current status. Small Tier 2 site Supporting ALICE, local scientific jobs, medical information processing - PowerPoint PPT PresentationTRANSCRIPT
Usage of cloud platform in datacenter BITP for resource and energy efficiency
Oksana Shadura, V.Pelykh, S.Svistynov
Bogolyubov Institute for Theoretical Physics Ukraine
Current status
• Small Tier 2 site• Supporting ALICE, local scientific jobs, medical
information processing• UA-BITP also supports all core services for
NGI_UA• Heterogeneous cluster: from old Xeon to E5-
26xx• Now 260 cores in total
Reasons for cloud?• Solution of problem:”3 different clusters for
different needs”– Easier support for IT team
• Lack of flexibility of management of resources for different projects
• Low efficiency of usage resources• Usage for old hardware for special needs
(educational, testing, small local jobs)• Reduce energy consumption?
Components of UA-BITP infrastructure
BITPData
Center
ARCcluster
ALICEcluster
Trainingcluster
Local cluster
Used utilities• RDO - community of people using and deploying OpenStack on Red
Hat and Red Hat-based platforms)– Packstack -> Puppet
Packstack is a utility that uses Puppet modules to deploy various parts of OpenStack on multiple pre-installed servers over SSH automatically. Supported by Fedora, Red Hat Enterprise Linux (RHEL), Scientific Linux and other RHEL based systems.
• Ceilometer The Ceilometer project is infrastructure to collect measurements within OpenStack. It`s primary targets are monitoring and metering, but the framework is expandable to collect usage for other needs– Plugins: KWAPI Kwapi is in fact infrastructure that connects wattmeters to
OpenStack Ceilometer.
Components of Openstack GrizzlyOpenstack
Nova Cinder Keystone
Glance Ceilometer (with plugins)
Nova-scheduler, nova-compute, nova-network (linux-bridge), nova-api
Glance-api Ceilometer-api
Cinder -api Keystone-api
Implementation
3 Virtual nodes
68 core
Implementation
20 tasks in queue68 tasks running
Automaticalstart of new VM
Implementation
69 tasks running
New VM with running task
Implementation
69 cores registered in grid
Openstack Ceilometer
• Collect data needed for accounting, includes:– Compute - Nova:• instances (type) - hourly usage• cpu - hourly usage• ram - hourly usage
– Network (data in/out) - hourly bytes + total bytes• differentiate between internal and external end-points• External floating IP - hourly bytes + total bytes
– Image usage in Glance, and etc.
Plugins:KWAPI 1/2
We need strategy which aims to reduce global energy consumption in cloud architectures:
• Accurately measuring of the power consumption of all devices in the cloud.
• Analysis of collected values, to place virtual machines at optimal location in the cloud.
• Turning off unused machines.
High Availability For Openstack
Main task is to prevent of availability and data loss problems
How we can do this?– Redundant instances of services and requests are
load balanced using virtual IPs and load balancer HAProxy.
– Replacement of failed services by bringing up online backup/detection of services using monitoring by Pacemaker application,
Scheme of Openstack (with HA deployment)
Plans for future development• Update to Havana release (supporting
Neutron, Ceilometer modules and etc.)• Move ALICE cluster to cloud platform (by
CernVM tools or organizing own scheme)• Improve Puppet modules/manifests for own
usage• Deploy process of contextualization for VMs• Analysis received results considering resource
and energy efficiency
Summary• Deployment of cloud is easy and fast…• It is good possibility to small sites to organize
better management for resources;• We can dynamically change resource limits for
different projects;• We can try to play with better allocation of
VM, and as a result, to receive lower energy consumption;
• Usage of virtual resources with same A/R.
Thank you for your attention!