TYPES OF VMWARE DATASTORES

TYPES OF VMWARE DATASTORES

An introduction to storage in virtual infrastructure

Vmware ESX supports three type of storage configuration when connecting to the shared storage array:

VMFS: virtual machine files system Datastore.

NAS: Network attached storage Datastore.

RDM: Raw device mapping Datastore.

The shared storage is required for the HA (high –availability), DRS (distributed resource scheduler), Vmotion and fault tolerance.

The 80/20 rule:

This rule is well known rule when we design virtual data center. This 80/20 rule means that the 80% of all system virtualized are of consolidation efforts. The remaining 20% of the system are classified as the business critical application. Although these applications can be virtualized successfully, they tend to be deployed on shared storage pools but in what we refer to as isolated dataset.

THE CHARACTERISTICS OF CONSOLIDATION DATASETS

Consolidation datasets have the following characteristics:

• The VMs do not require application-specific backup and restore agents.

• The dataset is the largest in terms of the number of VMs and potentially the total amount of storage

-addressed.

• Individually, each VM might not address a large dataset or have demanding IOP requirements; --however, the collective whole might be considerable.

• These datasets are ideally served by large, shared, policy-driven storage pools (or datastores).

THE CHARACTERISTICS OF ISOLATED DATASETS (FOR BUSINESS-CRITICAL APPLICATIONS)

Isolated datasets have the following characteristics:

• The VMs require application-specific backup and restore agents.

• Each individual VM might address a large amount of storage and/or have high I/O requirements.

• Storage design and planning apply in the same way as with physical servers.

• These datasets are ideally served by individual, high-performing, nonshared datastores.

Consolidated datasets work well with Network File System (NFS) datastores because this design provides greater flexibility in terms of capacity than SAN datastores when managing hundreds or thousands of VMs. Isolated datasets run well on all storage protocols; however, some tools or applications might have restrictions around compatibility with NFS and/or VMFS.

Unless your data center is globally unique, the evolution of your data center from physical to virtual will follow the 80/20 rule. In addition, the native multiprotocol capabilities of NetApp and VMware will allow you to virtualize more systems more quickly and easily than you could with a traditional storage array platform.

 VMFS DATASTORES

The Vmware VMFS is high-performance clustered file system that provides datastores, which are shared storage pools. VMFS Datastore can be configured with logical unit numbers (LUN) accessed by FC, iSCSI and FCoE. VMFS allows traditionally LUN accessed simultaneously by every ESX server in cluster.

Applications traditionally require storage considerations to make sure their performance can be virtualized and served by VMFS. With these types of deployments, NetApp recommends deploying the virtual disks on a datastore that is connected to all nodes in a cluster but is only accessed by a single VM.

This storage design can be challenging in the area of performance monitoring and scaling. Because shared datastores serve the aggregated I/O demands of multiple VMs, this architecture doesn’t natively allow a storage array to identify the I/O load generated by an individual VM.

SPANNED VMFS DATASTORES

Vmware provides the ability of VMFS extents to concatenate multiple LUN into a single logical Datastore, which is referred as a spanned datastore. Although the spanned Datastore can overcome the 2TB lun size limit, but it will affect the performance of the lun, because each size lun have the capacity to handle the I/Ops.

NetApp does not recommend the spanned datstores.

NFS DATASTORE.

vSphere allows customer to leverage enterprise-class NFS array to provide datastores to concurrent access to all of the node in ESX cluster. The access method is very similar to one with VMFS.

Deploying VMware with the NetApp advanced NFS results in a high-performance, easy-to-manage

Implementation that provides VM-to-datastore ratios that cannot be accomplished with other storage protocols such as FC. This architecture can result in a 10x increase in datastore density with a correlating reduction in the number of datastores. With NFS, the virtual infrastructure receives operational savings because there are fewer storage pools to provision, manage, back up, replicate, and so on.

SAN RAW DEVICE MAPPING

ESX gives VMs direct access to LUN for specific –use case such as P2V clustering or storage vendor management tool, this type of access is called raw device mapping and it support FC,iSCSI,FCoE protocol, In this design the ESX act as a connection proxy between the VM and storage array. RDM provides direct LUN access to the host so that they can achieve high individual disk I/O performance and can be easily monitored for the disk performance.

RDM LUNS ON NETAPP

RDM is available in two modes that are physical and virtual. Both mode support the key vmware features such as vmotion and can be used in both HA and DRS cluster.

NetApp enhances the use of RDMs by providing array-based LUN-level thin provisioning, production-use data deduplication, advanced integration components such as SnapDrive, application-specific Snapshot backups with the SnapManager for applications suite, and FlexClone zero-cost cloning of RDM-based datasets.

Note:  VMs running MSCS must use the path selection policy of Most Recently Used (MRU). Round

Robin is documented by VMware as unsupported for RDM LUNs for MSCS.

Datastore supported features

Capability/Feature	FC/FCoE	iSCSI	NFS
Format	VMFS or RDM	VMFS or RDM	NetApp WAFL
Maximum numbers of Datastore or LUNs	256	256	64
Maximum Datastore size	64TB	64TB	16TB or 100TB*
Maximum LUN/NAS file system size	2TB minus 512bytes	2TB minus 512 bytes	16TB or 100TB*
Optimal queue depth per LUN/file system	64	64	N/A
Available link speeds	4 and 8Gb FC and 10GbE	1 and 10GbE	1 and 10GbE

*100TB requires 64 bit aggregates.