Re: [one-users] Queries regarding Storage Backend for DataStore
Hi Robert, Am 14.01.15 um 15:08 schrieb Robert Foote: > wever, the initial deployments of VMs now, are quite bogged down, and can > take upwards of an hour or more to deploy a single VM with a 100Gb drive. We > believe this is now due to some NFS bogging. We are currently in the process > of moving to iSCSI with our OpenNebula + Nexenta deployments, and are > expecting a boost in performance as it relates to the initial deployment of > VMs. Do *NOT* use iSCSI on Nexenta. We used it in the past and run into different problems. F.e. it caused: a complete freeze/panic of nexenta, reaching connection limit and blocking until reboot of nexenta. Maybe your problem is not related to NFS. Do you use ZFS dedup or did you use ZFS dedup in the past? Did you move your ZIL to SSD? How old are this SSD? Fabian ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Queries regarding Storage Backend for DataStore
We are using glusterfs-fuse as our storage backend. It outperforms single NFS servers and scales better, we have looked into qemu-glusterfs native integration however its not quite ready for prime time deployments. The key is to have well balanced storage hardware and fast networks. For our storage we like to use few high performance cores attached to a DAS with some SSD caching. Then we scale out as needed and let glusterfs / network hardware balance out the performance. Opennebula actually makes it really easy to scale out as well with multiple system data stores. We don't necessarily have to add more bricks to gluster volumes, we just add more volumes and then create more system data stores for i/o load balancing. ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Queries regarding Storage Backend for DataStore
Neelesh: Just a quick FYI – We are also running a Nexenta SAN with our OpenNebula deployment. At the moment, we are utilizing NFS mounts from the Nexenta SAN for our datastores, over a 10Gb switched storage network. This has worked well until we got into the ~250 VM mark. Running VM’s are still maintaining great storage performance, and hitting upwards of 8k iops when needed. However, the initial deployments of VMs now, are quite bogged down, and can take upwards of an hour or more to deploy a single VM with a 100Gb drive. We believe this is now due to some NFS bogging. We are currently in the process of moving to iSCSI with our OpenNebula + Nexenta deployments, and are expecting a boost in performance as it relates to the initial deployment of VMs. Robert Foote Chief Technical Officer Phone | 1-800-611-8763 Ext. 201 www.bpsnode.com From: Users [mailto:users-boun...@lists.opennebula.org] On Behalf Of Jaime Melis Sent: Tuesday, January 13, 2015 11:32 AM To: ngur...@neeleshgurjar.co.in Cc: users Subject: Re: [one-users] Queries regarding Storage Backend for DataStore Hi, - Very first is how actually these datastore works in comparison with Cinder and Glance of OpenStack ? The best thing to do here is to read the documentation and let us know if there's something you don't understand. http://docs.opennebula.org/4.10/administration/storage/sm.html - Which Datastore I can put on faster connectivity and which on slower? I don't understand the question, the only thing I can say is that the storage subsystem is something OpenNebula relies on. You will get the same performance on your VMs with or without OpenNebula, as it depends on the hypervisor hardware and storage. So when thinking about this, think about it in terms of general performance. - Suppose I launched 2 VMs from one image, where the images are stored of Both VMs? In OpenStack I need to create Volume and attached as a RootDisk to instance, how should I go ahead in OpenNebula? You register an image, which will be stored in the datastore of choice, and then you create a template that references that image. When you instantiate it, OpenNebula will clone that image, and depending on the datastore driver it will clone it somewhere. Read the guide for each datastore type for more info. - How to calculate required IOPS to decide Storage Backend eg. Number of Drives, RAID level, Connectivity, etc. Best thing to do here is to run some benchmarks, as it largely depends on the whole setup - I was planning to use iSCSI with Nexenta based storage. Is that ok or should I go with Ceph ? will Ceph or NFS give performance ? Yes, Ceph will have better performane than NFS. - Any suggestions to design the storage backend for DataStore. It depends on each driver - Do I need fiber connectivity between Compute Nodes and Storage ? No, but the better the connectivity is, the less chances you have of having a bottleneck. Regards Neelesh Gurjar ___ Users mailing list Users@lists.opennebula.org <mailto:Users@lists.opennebula.org> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org -- Jaime Melis Project Engineer OpenNebula - Flexible Enterprise Cloud Made Simple www.OpenNebula.org <http://www.OpenNebula.org> | jme...@opennebula.org <mailto:jme...@opennebula.org> ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] Queries regarding Storage Backend for DataStore
Hi, - Very first is how actually these datastore works in comparison with > Cinder and Glance of OpenStack ? > The best thing to do here is to read the documentation and let us know if there's something you don't understand. http://docs.opennebula.org/4.10/administration/storage/sm.html > - Which Datastore I can put on faster connectivity and which on slower? > I don't understand the question, the only thing I can say is that the storage subsystem is something OpenNebula relies on. You will get the same performance on your VMs with or without OpenNebula, as it depends on the hypervisor hardware and storage. So when thinking about this, think about it in terms of general performance. > - Suppose I launched 2 VMs from one image, where the images are stored of > Both VMs? > In OpenStack I need to create Volume and attached as a RootDisk to > instance, how should I go ahead in OpenNebula? > You register an image, which will be stored in the datastore of choice, and then you create a template that references that image. When you instantiate it, OpenNebula will clone that image, and depending on the datastore driver it will clone it somewhere. Read the guide for each datastore type for more info. > - How to calculate required IOPS to decide Storage Backend eg. Number of > Drives, RAID level, Connectivity, etc. > Best thing to do here is to run some benchmarks, as it largely depends on the whole setup > - I was planning to use iSCSI with Nexenta based storage. Is that ok or > should I go with Ceph ? will Ceph or NFS give performance ? > Yes, Ceph will have better performane than NFS. - Any suggestions to design the storage backend for DataStore. > It depends on each driver > - Do I need fiber connectivity between Compute Nodes and Storage ? > No, but the better the connectivity is, the less chances you have of having a bottleneck. > > Regards > Neelesh Gurjar > ___ > Users mailing list > Users@lists.opennebula.org > http://lists.opennebula.org/listinfo.cgi/users-opennebula.org > -- Jaime Melis Project Engineer OpenNebula - Flexible Enterprise Cloud Made Simple www.OpenNebula.org | jme...@opennebula.org ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
[one-users] Queries regarding Storage Backend for DataStore
Hi, I am planning to setup OpenNebula 4.10 for Cloud Hosting. In beginning I will be hosting 50 VMs of 2 vCPU with 40GB SSD Storage and OS will be Linux. - Very first is how actually these datastore works in comparison with Cinder and Glance of OpenStack ? - Which Datastore I can put on faster connectivity and which on slower? - Suppose I launched 2 VMs from one image, where the images are stored of Both VMs? In OpenStack I need to create Volume and attached as a RootDisk to instance, how should I go ahead in OpenNebula? - How to calculate required IOPS to decide Storage Backend eg. Number of Drives, RAID level, Connectivity, etc. - I was planning to use iSCSI with Nexenta based storage. Is that ok or should I go with Ceph ? will Ceph or NFS give performance ? - Any suggestions to design the storage backend for DataStore. - Do I need fiber connectivity between Compute Nodes and Storage ? Regards Neelesh Gurjar ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org