ISC Cloud 2013 - Cloud Architectures for HPC – Industry Case Studies
2 likes710 views
This presentation discusses private cloud architectures for high-performance computing (HPC). It begins by describing the use case of using a private cloud for HPC workloads. It then covers the main challenges of deploying private HPC clouds, including flexible application management, resource management at scale, and ensuring application performance. Several case studies of existing private HPC clouds are presented, including those at FermiCloud, CESGA Cloud, SARA Cloud, SZTAKI Cloud, and KTH Cloud. Finally, trends in private cloud adoption by industry are discussed, such as experimenting with ARM architectures and providing hybrid cloud deployments.
ISC Cloud 2013 - Cloud Architectures for HPC – Industry Case Studies
- 1. Private Cloud Architectures for HPC Industry Case Studies ISC Cloud 2013 Heidelberg, September 23rd, 2013 Ignacio M. Llorente Project Director © OpenNebula Project. Creative Commons Attribution-NonCommercial-ShareAlike License
- 2. 2/22Private Cloud Architectures for HPC! Contents Private Cloud Architectures for HPC! This presentation is about: • The Private HPC Cloud Use Case • Main Challenges for Private HPC Cloud • Private HPC Cloud Case Studies • Private Cloud Trends in Industry
- 3. 3/22Private Cloud Architectures for HPC! The Private HPC Cloud Use Case The Pre-cloud Era! LRMS (LSF, PBS, SGE…) AccessProvision
- 4. 4/22Private Cloud Architectures for HPC! The Private HPC Cloud Use Case OpenNebula as an Infrastructure Tool – Enhanced Capabilities! Virtual Worker Nodes LRMS (LSF, PBS, SGE…) AccessProvisionService • Common interfaces • Custom environments • Dynamic elasticity • Consolidation of WNs • Simplified management • Physical – Virtual WNs • Dynamic capacity partitioning • Faster upgrades Service/Provisioning Decoupling!
- 5. 5/22Private Cloud Architectures for HPC! The Private HPC Cloud Use Case OpenNebula as an Provisioning Tool – Enhanced Capabilities! Pilot Jobs, SSH… IaaS Interface AccessProvisionService • Simple Provisioning Interface • Raw/Appliance VMs • Dynamic scalable computing • Custom access to capacity • Not only batch workloads • Not only scientific workloads • Improve utilization • Reduced service management • Cost efficiency
- 6. 6/22Private Cloud Architectures for HPC! Main Challenges for Private HPC Cloud Main Demands from Engineering and Supercomputing ! Flexible Definition of Multi-tier Applications Resource Management Scale-out and Provisioning Application Performance
- 7. 7/22Private Cloud Architectures for HPC! Main Challenges for Private HPC Cloud Using the Cloud – Execution of Multi-tiered Applications ! Management of interconnected multi-VM applications: • Definition of application flows • Catalog with pre-defined applications • Sharing between users and groups • Management of persistent scientific data • Automatic elasticity Front-end Worker Nodes { "name": ”Computing_Cluster", "deployment": "straight", "roles": [ { "name": "frontend", "vm_template": 0 }, { "name": "worker", "parents": frontend, "cardinality": 2, "vm_template": 3, "min_vms" : 1, "max_vms" : 5, "elasticity_policies" : { ”expressions" : ”CPU> 90%”, "type" : "CHANGE", "adjust" : 2, "period_number" : 3, "period" : 10 }, …
- 8. 8/22Private Cloud Architectures for HPC! Main Challenges for Private HPC Cloud Using the Cloud – Performance Penalty as a Small Tax You Have to Pay! Overhead in Virtualization • Single processor performance penalty between 1% and 5% • NASA has reported an overhead between 9% and 25% (HPCC and NPB)1 • Growing number of users demanding containers (OpenVZ and LXC) Need for Low-Latency High-Bandwidth Interconnection • Lower performance, 10 GigE typically, used in clouds has a significant negative (x2-x10, especially latency) impact on HPC applications1 • FermiCloud has reported MPI performance (HPL benchmark) on VMs and SR-IOV/Infiniband with only a 4% overhead2 • The Center for HPC at CSR has contributed the KVM SR-IOV Drivers for Infiniband3 (1) An Application-Based Performance Evaluation of Cloud Computing, NASA Ames, 2013 (2) FermiCloud Update, Keith Chadwick!, Fermilab, HePIX Spring Workshop 2013 (3) http://wiki.chpc.ac.za/acelab:opennebula_sr-iov_vmm_driver , 2013 Overhead in Input/Output • Growing number of Big Data apps • Support for multiple datastores including automatic scheduling
- 9. 9/22Private Cloud Architectures for HPC! Main Challenges for Private HPC Cloud Operating the Cloud – Resource Management! Optimal Placement of Virtual Machines • Automatic placement of VM near input data • Striping policy to maximize the resources available to VMs Fair Share of Resources • Resource quota management to allocate, track and limit resource utilization Management of Different Hardware Profiles • Resource pools (physical clusters) with specific Hw and Sw profiles, or security levels for different workload profiles (HPC and HTC) Isolated Execution of Applications • Full Isolation of performance-sensitive applications
- 10. 10/22Private Cloud Architectures for HPC! Main Challenges for Private HPC Cloud Operating the Cloud – Scale out and Provisioning! Multi-tier Deployment • Management of multiple cloud instances that may be hosted in different sites Provide VOs with Isolated Cloud Environ • Automatic provision of Virtual Data Centers Hybrid Cloud Computing • Cloudbursting to address peak or fluctuating demands for no critical and HTC workloads
- 11. 11/22Private Cloud Architectures for HPC! Private HPC Cloud Case Studies One of Our Main User Communities! Supercomputing Centers Research Centers Distributed Computing Infrastructures Industry
- 12. 12/22Private Cloud Architectures for HPC! Private HPC Cloud Case Studies FermiCloud! Nodes KVM on 23 nodes (1 TB RAM - 368 cores) Koi Computer Network Gigabit and Infiniband Storage CLVM+GFS2 on shared 120TB NexSAN SataBeats AuthN X509 Linux Scientific Linux Interface Sunstone Self-service and EC2 API App Profile Legacy, HTC and MPI HPC http://www-fermicloud.fnal.gov/ Typical Workloads • Scientific stakeholders get access to on- demand VMs • Developers & integrators of new Grid applications
- 13. 13/22Private Cloud Architectures for HPC! Private HPC Cloud Case Studies CESGA Cloud! Nodes KVM on 27 nodes (0.5 TB RAM – 216 cores) HP ProLiant Network 2 x Gigabit (1G and 10G) Storage ssh from remote EMC storage server AuthN X509 and core password Linux Scientific Linux Interface Sunstone Self-service and OCCI App Profile Individual VMs and virtualised computing clusters Typical Workloads • 103 users • Genomic, rendering… • Grid services on production at CESGA • Node at FedCloud project • UMD middleware testing http://cloud.cesga.es/
- 14. 14/22Private Cloud Architectures for HPC! Private HPC Cloud Case Studies SARA Cloud! Nodes KVM on 19 HPC nodes (256 GB RAM 608 cores) Dell PowerEdge and 10 “light” nodes (64 GB RAM 80 cores) Supermicro Network 4 x Gigabit (10G) with Arista switch Storage NFS on 400 GB NAS for HPC and ssh for “light” AuthN Core password Linux CentOS Interface Sunstone and OCCI App Profile MPI clusters, windows clusters and independent VMs ww.cloud.sara.nl Typical Workloads • Ad-hoc clusters with MPI and pilot jobs • Windows clusters for Windows-bound software • Single VMs, sometimes acting as web servers to disseminate results
- 15. 15/22Private Cloud Architectures for HPC! Private HPC Cloud Case Studies SZTAKI Cloud! Nodes KVM on 7 nodes (1.8 TB RAM – 448 cores) DELL PowerEdge Network 2 x Gigabit (1G and 10G) Storage iSCSI on DELL storage server 72 TB shared AuthN X509 Linux CentOS Interface Sunstone Self-service, EC2 and OCCI App Profile Individual VMs and virtualised computing cluster http://cloud.sztaki.hu/ . Typical Workloads • Run standard and grid services (e.g.:web servers, grid middlewares…) • Development and testing of new codes • Research on performance and opportunistic computing
- 16. 16/22Private Cloud Architectures for HPC! Private HPC Cloud Case Studies KTh Cloud! Nodes KVM on 768 cores (768 GB RAM) HP ProLiant Network Infiniband and Gigabit Storage NFS and LVM AuthN X509 and core password Linux Ubuntu Interface Sunstone self-service, OCCI and EC2 App Profile Individual VMs and virtualised computing cluster http://www.pdc.kth.se/ Typical Workloads • Mainly BIO • Hadoop, Spark, Galaxy, Cloud Bio Linux…
- 17. 17/22Private Cloud Architectures for HPC! Private Cloud Trends in Industry Experimenting with ARM for the Private Cloud! Why? • Decrease power consumption, reduce costs, simplify solutions… • Mostly managing bare metal and early experiences with virtualization Tiniest Cloud Ever! (by Citrix and Linaro at LCU 2013) Ubuntu on Versatile Express Cortex-A15 Dual core Ubuntu on Arndale Board Cortex-A15 Dual core http://www.youtube.com/watch?v=xZP9YKv3P_E
- 18. 18/22Private Cloud Architectures for HPC! Private Cloud Trends in Industry Cloud for Mission-critical Applications! Availability and redundancy to keep it running in case of failure • Cloud services availability => HA Architectures • Application availability => Failover Solutions Service Continuity (by European Aeronautic Company) OpenNebula 4.0 Automatic failover and recovery within 1 minute KVM
- 19. 19/22Private Cloud Architectures for HPC! Private Cloud Trends in Industry Hybrid Cloud Deployments ! Transparent and automatic access to the public cloud • Dev&testing to the public cloud • Security and performance sensitive workloads on the private cloud Cloudbursting Deployment (by Telecom Company) Public Cloud 1 Public Cloud 2 Local data center OpenNebula Private Cloud Cloud API is not relevant
- 20. 20/22Private Cloud Architectures for HPC! Try it Out! OpenNebula Sandboxes! ● OpenNebula pre-installed in a VM: VirtualBox, KVM, VMware, Amazon
- 21. 21/22Private Cloud Architectures for HPC! Join Us at OpenNebulaConf 2013!
- 22. 22/22Private Cloud Architectures for HPC! Thanks to People and Organizations that Provided Info to Prepare this Presentation! Questions? OpenNebula.org @OpenNebula