Yes, you can! In fact, pairing IBM Spectrum Scale and the Nutanix Enterprise Cloud OS makes good sense. Spectrum Scale cluster file system provides the global namespace and the Nutanix Enterprise Cloud OS provides the invisible infrastructure. With Spectrum Scale on Nutanix, you no longer need a separate multi-tier IT infrastructure to run analytics applications such as SAS Grid, Informatica, and Ab Initio, or to use products that embed Spectrum Scale, such as IBM Db2 PureScale.
The Nutanix Acropolis Distributed Storage Fabric (DSF) is well suited to supporting global namespace operations, as it transparently manages storage across the Nutanix Enterprise Cloud Platform. DSF knows where reads and writes occur and places data intelligently for optimal performance and reliability, even spreading the work when I/O demands exceed a single node’s capability.
Running Spectrum Scale on Nutanix requires no special processes. You deploy some virtual machines (VMs), define a set of shared LUNS using a volume group, and then create your Spectrum Scale cluster. For details on Spectrum Scale support in virtual environments, see the Spectrum Scale FAQ.
Once you have created the vDisks, there are two ways to access them using Spectrum Scale. First, if you are using AHV, the native Nutanix hypervisor, you can attach the vDisks directly to one or more VMs. In this case, all you need to do is attach the volume group to the Spectrum Scale VMs using Prism, at which point the vDisks are accessible and you can start creating Network Shared Disks (NSDs). Alternatively, you can use iSCSI to access the vDisks if you are using ESXi or AHV. For details on configuring iSCSI, check out Enabling Acropolis Block Services, which is part of the Prism Web Console Guide.
Once you have attached the vDisks to the VMs, you can use Spectrum Scale in the same manner as with other storage. How you architect your Spectrum Scale cluster depends on the way you want to utilize the shared namespace.
Option 1: Using Shared vDisks Directly
The Nutanix Acropolis Distributed Storage Fabric (DSF) allows VMs shared access to the vDisks in a volume group. In this configuration, every node in the Spectrum Scale cluster has block device access to the vDisks. This is the simplest configuration, as the volume group appears to Spectrum Scale as a shared disk SAN environment. As with any shared disk Spectrum Scale cluster, it is a good idea to keep these clusters under twenty members. This configuration works well for workloads with light to medium I/O demands.
Option 2: Using Spectrum Scale NSD
The Spectrum Scale Network Shared Disk (NSD) model can improve I/O performance for some workloads by allowing DSF to improve cache usage and data locality. When Spectrum Scale NSD servers are serving data to NSD clients only, they maintain exclusive access to a subset of the vDisks, thereby allowing DSF to optimize Nutanix Controller VM (CVM) cache usage and data locality, which in turn improves read performance. This configuration optimizes memory usage because DSF caches the data for any single vDisk in only one CVM; it improves locality because DFS automatically migrates data, enabling local storage to satisfy read operations and saving network bandwidth.
Once you have defined the NSD servers, you can access the data from VMs running within your Nutanix cluster or from nodes outside the cluster. Use this architecture if you have a large cluster, are processing high demand I/O, or need to access data from a node outside the Nutanix cluster.
The good news is that there is not much tuning to do. For example, the storage extent size on Nutanix is 1 MiB and thus, for some workloads, it may work better to use a Spectrum Scale file system block size of 1 MiB. However, in testing there was no measurable difference between a 256 KiB and 1 MiB file system block size for sequential read or write I/O operations. For the best performance, configure at least four vDisks per CVM and use a fast network to support write replication operations.
Nutanix invisible infrastructure, with advanced storage features like DSF, makes it easy to deploy your applications on a virtualized Spectrum Scale cluster, freeing you to devote your time toward extracting value from your data, rather than managing infrastructure complexity.
Disclaimer: This blog may contain links to external websites that are not part of Nutanix.com. Nutanix does not control these sites and disclaims all responsibility for the content or accuracy of any external site. Our decision to link to an external site should not be considered an endorsement of any content on such site.
2018 Nutanix, Inc. All rights reserved. Nutanix, the Enterprise Cloud Platform, and the Nutanix logo are registered trademarks or trademarks of Nutanix, Inc. in the United States and other countries. All other brand names mentioned herein are for identification purposes only and may be the trademarks of their respective holder(s).