Ready, Set, Go! Converged Infrastructure and OpenStack Deployments Q&A
Q: What is converged cloud? A: Converged infrastructure is the hardware and infrastructure software (ie, not cloud, necessarily), and a cloud offering on converged infrastructure could be defined as converged cloud.
Q: In term of operations, if storage and compute are on the same node, when you have a outage you lose both a storage and compute node. How are your customers handling that? It's 2 hits instead of 1 when non-converged. A: In this architecture, storage nodes (running 3x redundant Ceph) are isolated and separate from other nodes. Compute are also isolated and only used for compute. Foundation nodes are 3 nodes providing a virtualized control plane of all management functions.
Q: Would you qualify converged as turn-key vs hyper-converged as multi-role nodes? A: While confusing, definitions can vary. Converged does not necessarily mean completely turnkey (look at details of a given solution). Hyper-converged is turnkey in terms of it is out of the box, but most do not offer Cloud out of the box (again, look at details). What we are talking about here is a completely turnkey OpenStack cloud rack-based solution.
Q: Do you have anything that spells out the economic advantages of OpenStack+converged infrastructure? I don't hear much beyond technical detail. What makes the case besides technical advantages? Can you be specific about TCO or other economic benefits? A: This was touched on a bit in the actual webinar. Mirantis does have TCO comparison tools for analysis against AWS, as well as traditional virtualization. These tools are for use by Certified Rack Partners and/or Mirantis Sales. Your TCO calculations will vary compared to another customer, based on your costs and uses of cloud.
Q: What kind of customers could benefit from converged infrastructure? What should traditional customers consider? How about the migration effort from AWS for customers who adopt this for on-premise cloud? A: At a high level and to answer this as simply as possible, there are two sides to the equation of benefits here. If a customer wants to move from a legacy datacenter environment to save on the high costs of VMware, (or from AWS for that matter,) Mirantis has the experience to assist our customers in workload migration from AWS to an OpenStack private cloud environment. It is known that in an operational state, a private OpenStack cloud costs roughly 40% less to operate than a similar AWS public cloud.
Q: Murano is available for which release of Fuel?
A: Murano is an element of OpenStack. Murano is the application catalog that contains all of the different apps that you can utilize very quickly in you OpenStack Cloud. Murano and Fuel are independent of each other. Murano is in all (modern) releases of OpenStack. The Mirantis Unlocked Appliance ships with the full release of Mirantis OpenStack and therefore has Murano and all of the relevant apps within it.
Q: If I am the developer, how could this help me to work on my projects with less waiting for IT resource allocation? A: This is IaaS on day one, and an ability to move to DevOps tools and applications later in the afternoon. That enables developers to get going very rapidly in a Private Cloud that provides self-service once a services catalog is populated with tools and applications.
Q: How long does it take to get the all the pre-configured, pre-validation to be ready for customers to get the real systems? A: The longest part of the wait time is hardware leadtime. Rack N' Stack is generally a large rack in a day, automated software load is a few hours, validation test harness is three hours or so. Ignoring hardware leadtime, a partner can generally easily create a rack in 2 days, ready to ship. Onsite installation takes less than a day, with about an hour of uplink integration, rerunning validation testing, doing a demo, and handing over the credentials to the cloud.
Q: So once the systems are shipped, how are they maintained as the openstack software and system firmware go through their release cycles? A: There are SW images updates that are also validated. Performing the upgrade requires coordination with Mirantis.
Q: Could you provide approximate costing per rack with all the infrastructure? A: There is a range of prices between the smallest populated Unlocked appliance and the fully loaded Unlocked appliance. We would be happy to set up a call to discuss the pricing around the size of the Unlocked appliance that would best fit your needs.
Q: Can you please provide detailed hardware requirements for this solution? A diagram will help. A: You can fin diagrams and information at https://mirantis.com/appliances.
Q: Is this completely based on the CAPEX model, or do you also offer an OPEX model? A: This is a CapEx, upfront purchase, which does include support for hardware and software (generally thought of as OpEx). The Certified Rack Partners may offer programs and means of moving this to an OpEx model.
Q: Can you give an example of a minimal OpenStack converged infrastructure deployment? A: The smallest Mirantis Unlocked Appliance would be 6 compute nodes, 3 storage nodes and 3 foundation nodes in addition to the top of rack switches. When the customer receives the rack at their location (on/off prem) the typical time to deploy and stand up their OpenStack cloud is less than one day.
Q: What is done to improve on Ceph performance and/or monitor aggregate IOPS? A: The deployment of Ceph by Fuel, in conjunction with the automation tools, will leverage the use of SSD's for the Ceph Write journals to improve upon Ceph write performance. Also, the deployment will take into consideration the number of OSD's within the deployment to define an optimized setting the Ceph Placement Group(s). For monitoring, the Zabbix plugin is part of the deployment for the Unlocked appliance OpenStack deployment.
Q: If we have a live OpenStack cloud running the Icehouse version without Fuel then is it possible to migrate to Fuel 8? A: Fuel is the deployment tool for Mirantis OpenStack. Fuel for MOS 8.0 supports the Kilo version of OpenStack. Thus, it would require an upgrade path to MOS 8.0 to be able to leverage the Fuel deployment tool.
Looking for more information? Please check out the full webinar.