Competition Example Architectural Decision Entry 2 – Use of RDMs in Standard IaaS Clusters

Name: Chris Jones
Title: Virtualization Architect
Twitter: @cpjones44
Profile: VCP5 / VCAP5-DCD

Problem Statement

VMs require more than 1.9TB in a single disk. The existing virtual environment has LUNs provisioned that are 2TB in size. As these VMs have virtual data disks (VMDKs) that are > 1.9TB in size, alarms are being triggered by the infrastructure monitoring solution and raising Incident tickets to the Virtual Infrastructure support queue.

Assumptions

1. Data within the OSI must reside within the VM and not on some kind of IP based store (like a NAS share).

2. vSphere datastores are presented through FC and not IP based stores (ie. NFS).

3. vSphere Hypervisor is ESXi 4.1.

4. There is no requirement for the VMs to be performing SAN specific functionality or running SCSI target-based software.

Constraints

1. The implemented monitoring solution cannot be customised with triggers and monitoring policies for individual objects within the environment (ie. having one monitoring policy per individual or sub-group of datastores).

2. Maximum vSphere datastore size in version 4.1 is 2TB minus 512 bytes.

3. Unable to upgrade beyond ESXi 4.1 Update 3.

Motivation

1. Reduce the number of incident tickets being raised, thus improving SLA posture.

2. Reduce the requirement to span single Windows logical volumes across multiple VMDKs.

Architectural Decision

Turn the disk into an RDM (Virtual Compatibility Mode) to remove the level of monitoring from the vSphere layer.

Alternatives

1. Create smaller VMDKs (ie. 1-1.5TB disks) and create a RAID0 volume within the guest OS.

2. Change the level of alerting so that tickets are not raised for alerts that trigger beyond 90%.

3. Turn the disk into an RDM to remove the level of monitoring from the vSphere layer.

4. Thin Provision the virtual disks

5. Store the data within the guest on some kind of IP based storage (NAS/iSCSI target).

Justification

1. Option 5 goes against the assumption that data must be local to the VM, so was ruled out.

2. Whilst thin provisioning (Option 4) is an attractive solution, this option is ruled out based on a wider infrastructure decision to thick provision all disks in the environment to reduce the risk to datastores filling up and critical business VMs stopping.

3. Option 1 via smaller VMDKs spread across multiple vSphere datastores will result in these alerts disappearing, however it will create issues when trying to execute a DR recovery for either the individual disks (Active/Passive) or the whole VM (Active/Cold). All that’s needed is for one VMDK not to be replicated and the whole Windows volume will be corrupted, or for the VMDKs to be mounted in the wrong order. Multiple VMDKs to one Windows volume also complicates the recovery of snapshot array-based backups (eg. via SMVI or NetBackup).

4. Option 2 goes against the constraint of the infrastructure monitoring solution not being able to creating individual alerting policies for either a single or sub-group of datastores in the inventory. Should individualised policies be created, we would need to ensure that the affected VMDKs that consume 90-95% of a datastore remain on that datastore as moving from one to another (ie. from Tier 2 to Tier 1) will require a change to the monitoring that has been configured. At this stage, the monitoring solution has no way to track these customised policies, which is most of the reason why global environment wide policies exist.

5. Option 3 and the use of RDMs in Virtual Compatibility Mode will allow the VM to benefit from the features of VMFS, such as advanced file locking for data protection and vSphere snapshotting. The use of RDMs will also allow for VMs to be managed by DRS (ie. can be vMotion’ed) and protected by vSphere HA.

Implications

  1. The RDM mapping will need to be recorded clearly to avoid the lengthy process of discovering from scratch what physical LUN is presented to the virtual machine.

An example of how to map these will be to:

A)    Record the name of the VM that has the RDM.

B)    Record the NAA number of the physical LUN(s) that are presented to the VM.

C)    Record the virtual device node on the virtual disk controller as to where the RDM is mounted.

D)    Record the Windows drive letter that this RDM is mounted to.

2. Additional paths will be consumed, reducing the total number of vSphere datastores that can be presented to the cluster.

Back to Competition Main Page or Competition Submissions

Example Architectural Decision – Datastore (LUN) Sizing with Block Based Storage

Problem Statement

In a vSphere environment, What is the most suitable Datastore (LUN) sizing to use for to support both production & development workloads to ensure minimum storage overhead and optimal performance?

Requirements

1. RTO 4hrs
2. RPO 12hrs
3. Support Production and Test & Development Workloads
4. Ensure optimal storage capacity utilization
5. Ensure storage performance is both consistent & maximized
6. Ensure the solution is fully supported
7. Minimize BAU effort (Monitoring)

Assumptions

1. Business critical applications are excluded
2. Block based storage
3. VAAI is supported and enabled
4. VADP backups are being utilized
5. vSphere 5.0 or later
6. Storage DRS will not be used
7. SRM is in use
8. LUNs & VMs will be thin provisioned
9. Average size VM will be 100GB and be 50% utilized
10. Virtual machine snapshot will be used but not for > 24 hours
11. Change rate of average VM is <= 15% per 24 hour period
12. Average VM has 4GB Ram
13. No Memory reservations are being used
14. Storage I/O Control (SOIC) is not being used
15. Under normal circumstances storage will not be over committed at the storage array level.
16. The average maximum IOPS per VMs is 125 (16Kb) (MBps per VM <=2)
17. The underlying storage has sufficient performance to cater for the average maximum IOPS per VM
18. A separate swap file datastore will be configured per cluster

Constraints

1. Must used existing storage solution (Block Based Storage)

Motivation

1. Increase flexibility
2. Ensure physical disk space is not unnecessarily wasted
3. Create a Scalable solution
4. Ensure high performance
5. Ensure high utilization of storage resources by reducing “islands” of unused capacity
6. Provide flexibility in the unit size of partial SRM failovers

Architectural Decision

The standard datastore size will be 3TB and contain up to 25 standard virtual machines.

This is based on the following

25 VMs per datastore X 100GB (Assumes no over commitment) = 2500GB

25 VMs w/ 4GB RAM = 100GB minus 0Gb reservation = 100GB vswap space to be stored on the swap file datastore

25 VMs w/ Snapshots of up to 15% =  375GB

Total = 2500GB + 375GB = 2875GB

Average capacity used per VM = 115GB

Justification

1. In worst case scenario where every VM has used 100% of its VMDK capacity and has 4GB RAM with no memory reservation and a snapshot of up to 15% of its size the 3TB datastore will still have 197GB remaining, as such it will not run out of space.
2. The Queue depth is on a per datastore (LUN) basis, as such, having 25 VMs per LUNs allows for a minimum of 1.28 concurrent I/O operations per VM based on the standard queue depth of 32 although it is unlikely all VMs will have concurrent I/O so the average will be much higher.
3. Thin Provisioning minimizes the impact of situations where customers demand a lot of disk space up front when they only end up using a small portion of the available disk space
4. Using Thin provisioning for VMs increases flexibility as all unused capacity of virtual machines remains available on the Datastore (LUN).
5. VAAI automatically raises an alarm in vSphere if a Thin Provisioned datastore usage is at >= 75% of its capacity
6. The impact of SCSI reservations causing performance issues (increased latency) when thin provisioned virtual machines (VMDKs) grow is unlikely to be an issue for 25 low I/O VMs and with VAAI is no longer an issue as the Atomic Test & Set (ATS) primitive alleviates the issue of SCSI reservations.
7. As the VMs are low I/O it is unlikely that there will be any significant contention for the queue depth with only 25 VMs per datastore
8. The VAAI UNMAP primitive provides automated space reclamation to reduce wasted space from files or VMs being deleted
9. Virtual machines will be Thin provisioned for flexibility, however they can also be made Thick provisioned as the sizing of the datastore (LUN) caters for worst case scenario of 100% utilization while maintaining free space.
10. Having <=25 VMs per datastore (LUN) allows for more granular SRM fail-over (datastore groups)

Alternatives

1.  Use larger Datastores (LUNs) with more VMs per datastore
2.  Use smaller Datastores (LUNs) with less VMs per datastore

Implications

1. When performing a SRM fail over, the most granular fail over unit is a single datastore which may contain up to 25 Virtual machines.

2. The solution (day 1) does not provide CapEx saving on disk capacity but will allow (if desired) over commitment in the future

Thanks to James Wirth (VCDX#83) @JimmyWally81 for his contributions to this example decision.

Related Articles

1. Datastore (LUN) and Virtual Disk Provisioning (Thin on Thick)

2. Datastore (LUN) and Virtual Disk Provisioning (Thin on Thin)

3. Virtual Machine vSwap Location

CloudXClogo

 

Example Architectural Decision – Datastore (LUN) and Virtual Disk Provisioning (Thin on Thin)

Problem Statement

In a vSphere environment, What is the most suitable disk provisioning type to use for the LUN and the virtual machines to ensure minimum storage overhead and optimal performance?

Requirements

1. Ensure optimal storage capacity utilization
2. Ensure storage performance is both consistent & maximized

Assumptions

1. vSphere 5.0 or later
2. VAAI is supported and enabled
3. The time frame to order new hardware (eg: New Disk Shelves) is <= 4 weeks
4. The storage solution has tools for fast/easy capacity management

Constraints

1. Block Based Storage

Motivation

1. Increase flexibility
2. Ensure physical disk space is not unnecessarily wasted

Architectural Decision

“Thin Provision” the LUN at the Storage layer and “Thin Provision” the virtual machines at the VMware layer

(Optional) Do not present more LUNs (capacity) than you have underlying physical storage (Only over-commitment happens at the vSphere layer)

Justification

1. Capacity management can be easily managed by using storage vendor tools such eg: Netapp VSC / EMC VSI / Nutanix Command Center
2. Thin Provisioning minimizes the impact of situations where customers demand a lot of disk space up front when they only end up using a small portion of the available disk space
3. Increases flexibility as all unused capacity of all datastores and the underlying physical storage remains available
4. Creating VMs with “Thick Provisioned – Eager Zeroed” disks would unnessasarilly increase the provisioning time for new VMs
5. Creating VMs as “Thick Provisioned” (Eager or Lazy Zeroed) does not provide any significant benefit (ie: Performance) but adds a serious capacity penalty
6. Using Thin Provisioned LUNs increases the flexibility at the storage layer
7. VAAI automatically raises an alarm in vSphere if a Thin Provisioned datastore usage is at >= 75% of its capacity
8. The impact of SCSI reservations causing performance issues (increased latency) when thin provisioned virtual machines (VMDKs) grow is no longer an issue as the VAAI Atomic Test & Set (ATS) primitive alleviates the issue of SCSI reservations.
9. Thin provisioned VMs reduce the overhead for Storage vMotion , Cloning and Snapshot activities. Eg: For Storage vMotion it eliminates the requirement for Storage vMotion (or the array when offloaded by VAAI XCOPY Primitive) to relocate “White space”
10. Thin provisioning leaves maximum available free space on the physical spindles which should improve performance of the storage as a whole
11. Where there is a real or perceved issue with performance, any VM can be converted to Thick Provisioned using Storage vMotion not disruptivley.
12. Using Thin Provisioned LUNs with no actual over-commitment at the storage layer reduces any risk of out of space conditions while maintaining the flexibility and efficiency with significantly reduce risk and dependency on monitoring.
13. The VAAI UNMAP primitive provides automated space reclamation to reduce wasted space from files or VMs being deleted

Alternatives

1.  Thin Provision the LUN and thick provision virtual machine disks (VMDKs)
2.  Thick provision the LUN and thick provision virtual machine disks (VMDKs)
3.  Thick provision the LUN and thin provision virtual machine disks (VMDKs)

Implications

1. If the storage at the vSphere and array level is not properly monitored, out of space conditions may occur which will lead to downtime of VMs requiring disk space although VMs not requiring additional disk space can continue to operate even where there is no available space on the datastore
2. The storage may need to be monitored in multiple locations increasing BAU effort
3. It is possible for the vSphere layer to report sufficient free space when the underlying physical capacity is close to or entirely used
4. When migrating VMs from one thin provisioned datastore to another (ie: Storage vMotion), the storage vMotion will utilize additional space on the destination datastore (and underlying storage) while leaving the source thin provisioned datastore inflated even after successful completion of the storage vMotion.
5.While the VAAI UNMAP primitive provides automated space reclamation this is a post-process, as such you still need to maintain sufficient available capacity for VMs to grow prior to UNMAP reclaiming the dead space

Related Articles

1. Datastore (LUN) and Virtual Disk Provisioning (Thin on Thick)CloudXClogo