Here are my notes for the white paper from netapp
These are items that I felt I should review or felt that they are important to know.
Read the white paper for a fuller experience.
The recommendations and best practices presented in this document should be considered as deployment requirements.
80/20 rule,
which is that 80% of all systems virtualized are for consolidation efforts. The remaining 20% of the systems are classified as business-critical applications.
80%: consolidated datasets 20% isolated datasets
CONSOLIDATION DATASETS
VMs do not require application-specific backup and restore agents.
Individually, each VM might not address a large dataset or have demanding IOP requirements; however, the collective whole might be considerable.
served by large, shared, policy-driven storage pools (or datastores).
Consolidated datasets work well with Network File System (NFS) datastores because this design provides greater flexibility in terms of capacity than SAN datastores
Isolated datasets
require application-specific backup and restore agents.
Each individual VM might address a large amount of storage and/or have high I/O requirements.
ideally served by individual, high-performing, nonshared datastores.
SPANNED VMFS DATASTORES
spanned datastores can overcome the 2TB LUN
most commonly used to overcome scaling limits imposed by storage arrays that use per LUN I/O queues.
NetApp does not recommend the use of spanned VMFS datastores.
RAW DEVICE MAPPINGS
RDM: ESX acts as a connection proxy between the VM and the storage array.
thin virtual disk
The process of allocating blocks on a shared VMFS datastore is considered a metadata operation and as such executes SCSI locks on the datastore while the allocation operation is executed.
Although this process is very brief, it does suspend the write operations of the VMs on the datastore.
Data thatneeds to be written must pause while the blocks required to store the data are zeroed out.
STORAGE ARRAY THIN PROVISIONING
The value of thin-provisioned storage is that storage is treated as a shared resource pool and is consumed only as each individual VM requires it.
FAS data deduplication.
With NetApp FAS deduplication, VMware deployments can eliminate the duplicate data in their environment, enabling greater storage use. Deduplication virtualization technology enables multiple VMs to share the same physical blocks in a NetApp FAS system in the same manner that VMs share system memory.
Deduplication runs on the NetApp FAS system at scheduled intervals and does not consume any CPU cycles on the ESX server. (but it does on the storage device)
For the largest storage savings, NetApp recommends grouping similar operating systems and similar applications into datastores,
To recognize the storage savings of deduplication with LUNs, you must enable NetApp LUN thin provisioning. ::
Although deduplication reduces the amount of consumed storage, the VMware administrative team does not see this benefit directly, because its view of the storage is at a LUN layer, and LUNs always represent their provisioned capacity. (seen this is my netapp where even though there is high dedup, the amount used as far as esxi is concered is high)
DEDUPLICATION ADVANTAGES WITH NFS
when deduplication is enabled with NFS, the storage savings are immediately available and recognized by the VMware administrative team.
VAAI
vStorage APIs for array integration :: mechanism for the acceleration of certain functions typically performed at the hypervisor by offloading these operations to the storage array.
To use VAAI, the NetApp array must be running NetApp Data ONTAP version 8.0.1. VAAI is enabled by default in Data ONTAP and in ESX/ESXi.
storage I/O control (SIOC)
enables quality of service control for storage using the concepts of shares and limits
Allows the administrator to make sure that certain VMs are given priority access to storage compared to other VMs, based on the allocation of resource shares, maximum IOPS
limits, and whether or not the datastore has reached a specified congestion threshold.
SIOC is currentlyonly supported on FC or iSCSI VMFS datastores.
SIOC does not take action to limit storage throughput of a VM based on the value of its resource shares until the datastore congestion threshold is met.
STORAGE NETWORK DESIGN AND SETUP
converged networks:: current industry trend is solely focused on multipurposed Ethernet networks ) that provide storage, voice, and user access.
FC storage networks provide a single service, these single-purpose networks are simpler to design and deploy i
primary difference between SAN and NAS is in the area of multipathing. In the current versions of ESX/ESXi, NFS requires manual static path configuration,
VLANs and VLAN tagging simple but important role in securing an IP storage network. restricted to a range of IP addresses that are available only on the IP storage VLAN.
Flow control low-level process for managing the rate of data transmission between two nodes to prevent a fast sender from overrunning a slow receiver.
Can be configured on ESX/ESXiservers, FAS storage arrays, and network switches.
NetApp recommends turning off flow control For modern network equipment, especially 10GbEequipment, and allowing congestion management to be performed higher in the network stack.
NetApp recommends older equipment, typically GbE configuring the endpoints, ESX servers, and NetApp arrays with the flow control set to "send."
Spanning Tree Protocol (STP) : network protocol that provides a loop-free topology for any bridged LAN.
allows a network design to include spare (redundant) links to provide automatic backup paths if an active link fails, withoutthe danger of bridge loops or the need for manual enabling or disabling of these backup links.
Bridgeloops must be avoided because they result in flooding the network.
ROUTING AND IP STORAGE NETWORKS
NetApp recommends configuring storage networks as a single network that does not route. This model helps to provide good performance and a layer of data security.
SEPARATE ETHERNET STORAGE NETWORK
NetApp recommends separating IP-based storage traffic from public IP network traffic by implementing separate physical network segments or VLAN segments.
VMware best practice for HA clusters is to define a second service console port for each ESX server.The network use for IP-based storage is a convenient network that you can use to add this second SC port.
NetApp recommends not allowing routing of data between the storage or VMkernel and other networks. In other words, do not define a default gateway for the VMkernel storage network. Withthis model, NFS deployments require defining a second service console port on the VMkernel storage virtual switch within each ESX server.
virtual network interface (VIF) or EtherChannel
mechanism that supports aggregation of network interfaces into one logical interface unit. Once created, a VIF is indistinguishable from a physical network interface. VIFs are used to provide fault tolerance of the network connection and in some cases higher throughput to the storage device.
Cisco Nexus 1000V
Be aware that it is composed of two components, the Virtual Supervisor Module (VSM) and the Virtual Ethernet Module (VEM).
The VSM runs as a VM and is the brains of the operation with a 1000V.
Traffic continues if the VSM fails;however, management of the vDS is suspended.
VSMs should be deployed in an active-passive pair. These two VSAs should never reside on the same physical ESX server. This configuration can be controlled by DRS policies.
The VEM is embedded in all ESX/ESXi hosts, and it is managed by the VSM. One exists for each host in the cluster that is participating in the Cisco Nexus 1000V vDS.
NetApp and Cisco recommend making sure that all service console and VMkernel interfaces (vswif and vmknic) reside on a system VLAN. System VLANs are
defined by an optional parameter that can be added in a port profile.
Do not configure a VM network as a system VLAN.
storage availability
Deployed storage design that meets all of these criteria can eliminate all single points of failure::
purchasing physical servers with multiple storage interconnects or HBAs,
deploying redundant storage networking and network paths
leveraging storage arrays with redundant controllers.
data protection requirements in a virtual infrastructure are greater than those in atraditional physical server infrastructure.
NetApp RAID-DP
advanced RAID technology that is provided as the default RAID level onall FAS systems.
protects against the simultaneous loss of two drives in a single RAID group.
overhead with default RAID groups is a mere 12.5%.
safer than data stored on RAID 5 and more cost effective than RAID 10.
Aggregate
is the NetApp virtualization layer,
abstracts physical disks from logical datasets that are referred to as flexible volumes.
Aggregates are the means by which the total IOPS available to all of the physical disks are pooled as a resource.
This design is well suited to meet the needs of anunpredictable and mixed workload.
NetApp recommends using a dedicated two-disk aggregate.
By default, the root aggregate is composedof three disks due to the overhead of RAID-DP. To reduce the disk total from three to two, you must
modify the RAID type from RAID-DP to RAID 4.
If small number of disk drives. aggregate is not recommended.
VMkernel swap
ESX servers create a VMkernel swap or vswap file for every running VM. ie: 4gb ram = 4gb-8GB ram swap
NetApp recommends relocating the VMkernel swap file for every VM from the VM home directory to a datastore on a separate NetApp volume dedicated to storing VMkernel swap
Reason: this data is transient in nature and is not required in the case of recovering a VM either from a backup copy or by using Site Recovery Manager. (SRM)
NetApp recommends creating either a large thin-provisionedLUN or a FlexVol volume with the autogrow feature enabled.
This design removes theneed to micromanage the swap space or to reduce the usage rate of the storage.
If you undersize the swap space, the VMs fail to start;
The datastore that stores the vswap file is a single datastore for an entire cluster. (single point of failure here)
NetApp does not recommend implementing local datastores on each ESX/ESXi host to store vswap because this configuration has a negative impact on vMotion migration times.
OPTIMIZING WINDOWS FILE SYSTEM FOR OPTIMAL I/O PERFORMANCE
If your VM is not acting as a file server do this in the windows vm:
disable the access time updates process in (NTFS). This changereduces the amount of IOPS occurring within the file system.
In command prompt :: fsutil behavior set disablelastaccess 1.
Defrag
VMs stored on NetApp storage arrays should not use disk defragmentation utilities
the WAFL filesystem is designed to optimally place and access data at a level below the guest operating system file system.
If a software vendor advises you to run disk defragmentation utilities inside of a VM, contactthe NetApp Global Support Center before initiating this activity.
OPTIMAL STORAGE PERFORMANCE
Guest OS: align the partitions of the virtual disks to the block boundaries of VMFS and the block boundaries of the storage array.
When aligning the partitions of virtual disks for use with NetApp FAS systems, the starting partition offsetmust be divisible by 4,096.
Windows 2008, Windows 7, or Windows Vista don’t need alignment. Windows 2000, 2003,and XP do
Datastore: NetApp systems automate the alignment of VMFS when you select the LUN type"VMware" for the LUN.
Snapshots:
NetApp Snapshot technology can easily be integrated into VMware environments,
This is the only Snapshot technology that does not have a negative impact on system performance.
VMware states that for optimum performance and scalability, hardware-based Snapshot technology ispreferred over software-based solutions.
The shortcoming of this solution is that it is not managed withinvCenter Server, requiring external scripting and/or scheduling to manage the process.
vSphere installation bundle (VIB),
support for adding partner and other third-partysoftware components.
Host profiles
To avoid conflicts where VSC sets a parameter one way, rendering the ESXi host noncompliant with its host profile, it is important that the reference host Be configured with NetApp best practices using the VSC Monitoring and Host Configuration panel before creating the profile from the reference host.
Storage DRS
provides smart VM placement across storage by making load-balancing decisions based upon the current I/O latency and space usage and moving VMDKs
nondisruptively between the datastores in the datastore cluster (POD)
Note:Snapshot copies cannot be migrated with the VM.
datastore cluster (POD)
collection of like datastores aggregated into a single unit of consumption from anadministrator’s perspective.
enables smart and rapid placement of new virtual machines and VMDKs and load balancing of existing workloads.
Best Practices storage DRS and Datastore clusters: (NETAPP)
Set SDRS to manual mode and to review the recommendations before accepting them.
All datastores in the cluster should use the same type of storage (SAS, SATA, and so on) and have the same replication and protection settings.
SDRS will move VMDKs between datastores, and any space savings from NetApp cloning or deduplication will be lost when the VMDK is moved. rerun deduplication to regain savings.
Do not use SDRS on thinly provisioned VMFS datastores due to the risk of reaching an out-of-space situation.
Do not mix replicated and nonreplicated datastores in a datastore cluster.
All datastores in an SDRS cluster must either be all VMFS or all NFS datastores.
storage I/O (SIOC)
provide I/O prioritization of virtual machine running on a cluster of VMware ESX servers
prioritizes VMs’ access to shared I/O resources based on disk shares assigned to them.
*Storage DRS works to avoid I/O bottlenecks.
*SIOC manages unavoidable I/O bottlenecks.
0 comments:
Post a Comment