In this post I will try to explain what Clustered ONTAP is, and how it works.
First the basics.
There are several parts that make a Clustered ONTAP system. It all starts with a cluster.The cluster operates as a single entity, but it has the underlying concept of individualized components.
A cluster can be build with several HA pairs of nodes (max. 8 nodes with SAN, 24 nodes with NAS). All the nodes deliver physical resources to the cluster that can be used for functionality.
Functionality will be given by Storage Virtual Machines (SVM’ s)
Servers and clients connect to Logical InterFaces (LIF’ s) that are used by a SVM.
Volumes, shares, LUNS and mount points are the places where the data is stored.
The protocol to connect (CIFS, NFS, iSCSI, FC/FCoE) to the data is configured on the SVM.
To protect data we can use snapshots and mirror/vault technics.
Something else that is very cool is the possibility to move resources/functionality without disruption. You can move data (example from SATA to SAS or SSD disks) or LIFs (example from 1Gb to 10Gb) without to reconfigure and disruption. This makes it possible to scale up/out the cluster on an easy way. Also maintenance like updating your system is possible on this manor without downtime. For that reason the cluster is also called an immortal cluster.
Now take a look at the two parts you need.
First we have the hardware.
You need a storage controller that is supported by ONTAP.
All FAS8000 and 2500 systems are supported, but also some older systems like the 6000/3000 are. For a complete list check the HardWare Universe (HWU)
Here you also find the supported shelves and disk drives that you can use.
Another part is the network. You always need a fronted switch, but in some cases you also need a backend switch. If you have a cluster with 2 nodes, they can use a direct connection (back-to-back). But if you have more then 2 nodes in your cluster, they need to connect to each other on a dedicate network. You can see this as a backbone network where all nodes communicate to each other and all the cluster traffic will flow.
As front-end connectivity you can use all common protocols. So you can use ethernet (1 and 10Gb), fibre channel or fibre channel (FC)over ethernet (FCoE). You have the choice of using dedicated interfaces or make use of Converged Network Adapters (CNA), adapters that can do more protocols. The use of CNA’s make you more flexible and will let you use the best solution without investments in your storage system. On the other side, you need a infrastructure that supports this. But the mostly used infrastructure switches of today like Cisco Nexus support this.
Next we have the software.
Each node needs an Operating System to function. The nodes use ONTAP for that.
In a cluster, all the nodes use the same version of ONTAP.
You can say that there are 2 parts that are being used, a part that let the node boot and deliver resources and a part that deliver services to the cluster.
To deliver resources and services, ONTAP uses databases and applications.
Each node in the cluster has its own copy of a replicated database with the cluster and SVM configuration information. Additionally, each node has its own set of user space applications that handle cluster operations and node-specific caches.
To store data on disk, it uses WAFL (Write Anywhere File Layout). WAFL is a proven filesystem that is used by NetApp ONTAP for several years and let companies build high-performance enterprise storage systems.
In a next post, I will go deeper in the software layer, the use of SVM’ s and the other parts that make it a Clustered ONTAP system.