Comparing FA to vSAN for VCF Workload Domains
The purpose of this document is to share how the Pure Storage FlashArray can enhance the capabilities of the VMware Cloud Foundation (VCF) suite. VCF is an integrated stack comprised of vSphere, vSAN, NSX, SDDC Manager, and the vRealize suite. VCF can be deployed on-premises, or as VMware Cloud via a VMware Cloud Provider.
For questions on this document, please reach out to Vaughn Stewart.
Executive Summary
Pure Storage can help enterprises and service providers exceed SLAs and extend the capabilities of the VCF / vSphere deployments in the following key areas:
- Optimize data center resources. Reduce the infrastructure spend for compute & storage and eliminate hardware-defined silos.
- Provide storage fabric flexibility. Investment protection by enabling VCF on fibre channel fabrics, non-disruptive migration between storageprotocols (i.e. FC & iSCSI), and a future proof architecture that supports next-gen storage protocols (i.e. NVMe-oF & vVols).
- Simply and scale operations. Remove the need for mass data evacuations, data re-hydrations, and data reconstructions that occur with server maintenance, hardware expansions and refresh and server failures.
- Automate and protect the infrastructure. Increase the availability, scale, performance and business continuity of a a VCF deployment through robust integrations.
Pure has achieved a Satmetrix certified NPS score of 86.6 (top 1% of B2B CSAT) and has been identified as a leader in Gartner’s Magic Quadrant for Solid State Array for 5 years. Collectively these data points help communicate why Pure Storage FlashArray and VMware integrations provide the ideal foundation for extending the capabilities of VCF deployments.
VCF Storage Options
VCF deployments introduce the concept of domains; this is the physical assignment of hardware to support a set of virtualized applications and workloads.
• The Management domain. This supports all VMware management tools including SDDC Manager VM, a vCenter VM for each workload domain, NSX VMs, vROPs VMs, vRNI VMs, HCX Manager VMs, etc. As of VCF 4.0 the management cluster must reside on vSAN.
• Workload Domains. These support VMs and containers and may reside on either vSAN or 3rd party storage (which must be supported by vSphere). Workload domain storage types are classified as either Principal or Secondary. Both provide identical management capabilities within SDDC Manager; however, the former can be provisioned within SDDC Mgr and later must be manual provisioned and imported into SDDC Mgr. VMware has a roadmap for promoting protocols from secondary to principal so this document may be out of date when it comes to this classification.
This document is explicitly focused on presenting the capabilities of an on-prem VCF deployment based on vSphere 6.7 and VCF 3.9 with a FlashArray running Purity OE 5.3 and vSAN 6.7.
The data referenced in the following pages was sourced from the following sources:
https://storagehub.vmware.com/export...nical-overview
https://docs.vmware.com/en/VMware-vS...otingguide.pdf https://docs.vmware.com/en/VMware-vS...ment-guide.pdf
https://docs.vmware.com/en/VMware-vS...tion-guide.pdf
https://kb.vmware.com/s/article/2108740
https://blogs.vmware.com/virtualbloc...ntenance-mode/
https://docs.vmware.com/en/VMware-vS...37E61ECE4.html
https://docs.vmware.com/en/VMware-vS...76E46B35F.html
VMware is a strategic partner of Pure Storage. This document is intended solely for educational purposes and not for competitive use.
Core VMware vSphere Functionality
The following list covers core VMware capabilities when vSphere is deployed with vSAN and Pure Storage.
Feature | vSAN | FlashArray |
Support DRS | Yes | Yes |
Support RBAC | Yes | Yes |
Support vMotion | Yes | Yes |
Support Storage vMotion | Yes | Yes |
Support vSphere Replication | Yes |
Yes |
Support Thick VMDKs | Yes | Yes |
Support EZ-Thick VMDKs | Yes | Yes |
Support Thin VMDKs | Yes | Yes (recommended) |
Space Reclamation of data deleted with in VMDKs | Yes | Yes (vVols and VMFS) |
Space Reclamation for datastores | No | Yes |
Support vRealize Automation | Yes | Yes |
Support vRealize Orchestrator | Yes | Yes |
Support vRealize Log Insight | Yes | Yes |
Support vRealize Operations Manager | Yes | Yes |
Support VMware PowerCLI | Yes | Yes |
Support vSphere Update Manager | Yes | Yes |
Supported by SDDC Manager | Yes | Yes |
Supports VCF management domain |
Yes |
No |
Supports VCF workload domains |
Yes |
Yes |
Support datastore provisioning in SDDC Manager |
Yes |
Yes (with principal storage. Secondary storage provisioned in vCenter and imported) |
Support non-disruptive updates of ESXi by SDDC Manager and Update Manager |
Yes |
Yes (without any data evacuation or loss of data protection) |
Supports migrating ESXi nodes between workload domains |
Yes (data rebalancing required in source and destination domains) |
Yes (without data rebalancing) |
Increase infrastructure availability and SLAs
Availability is a fundamental component within an infrastructure. The FlashArray global install base has achieved greater than six-nine’s of availability, which includes planned and unplanned downtime. In addition the FlashArray delivers 100% performance, through failure and maintenance, with non-disruptive updates delivered via Pure1. The following list covers the core tenant of availability when vSphere is deployed with vSAN and Pure Storage.
vSAN |
FlashArray |
|
Support Basic VMs |
Yes |
Yes |
Support VMs with FT |
Yes |
Yes |
Support VMs with MSCS |
Yes (requires iSCSI LUNs mapped to guest initiators) |
Yes (natively) |
Disaster Recovery with SRM |
Yes (requires vSphere replication infrastructure) |
Yes (via native array replication) |
VM Granular Snapshots |
Yes (for ephemeral purposes) |
Yes |
VM snapshots with zero performance impact |
No |
Yes |
Datastore Snapshots |
No |
Yes |
Support vSphere HA |
Yes |
Yes |
Support vMSC HA |
Yes |
Yes (via ActiveCluster) |
Instant deployment of VMs from a template |
No (VSAN requires persistent VMs to be copied from a template) |
Yes (via VAAI XCopy) |
Support Data at Rest Encryption AES-256 |
Yes |
|
Secure erasure of customer data |
No |
Yes |
Automatic Noisy Neighbor QoS |
No |
Yes |
IOPs Limiting QoS |
Yes |
|
Bandwidth Limiting QoS |
No |
Yes |
100% I/O performance thru storage failures and maintenance |
No |
Yes |
Achieved 99.9999+% uptime across entire install base |
No |
Yes |
Non-disruptive Software Updates Delivered via the Cloud |
No |
Yes |
Non-disruptive upgrade (to VMs) of ESXi hosts via VMware SDDC Manager |
Yes |
Yes |
Predictive Support Analytics |
No |
Yes |
Company NPS Score |
39 |
86.6 (top 1%) |
Provides sub-millisecond latency, with erasure coding, FFT=2 data protection, data encryption and data reduction enabled |
No |
Yes |
Simplify Operations at Scale
The benefits of Pure Storage are most obvious in the area of simplified operations at scale. The FlashArray is designed to be successfully operated by any member of the IT staff – be they a member of the VMware or a storage team. This unique benefit is achieved in part by the architecture and by removing storage configuration options and associated trade-offs found in storage arrays and hyper-converged platforms. With FlashArray, all data is automatically protected with N+2 erasure coding, is encrypted with AES-256 encryption, includes automatic noisy neighbor QoS, is deduplicated and compressed for cost savings while delivering consistent sub-millisecond latency through normal operations, failures and maintenance. This allows the FlashArray to simultaneously meet the requirements of performance-centric and cost-sensitive workloads, without any configuration requirements or the need for separate deployment silos. The only storage configuration options requiring administrative effort is configuring data protection policies, bandwidth and /or IOPs limiting QoS policies, naming and sizing volumes, and connecting storage to hosts –the latter can be automated via either the Pure vCenter plug-in, vRO or vRA.
VSAN |
FlashArray |
|
Manage Storage 100% within vSphere Web Client |
Yes |
Yes (via plug-in) |
Number of Options to Configure & Deploy a Datastore |
12 Options - Auto Claim Storage - Data Reduction - Support Reduced Redundancy - Set Fault Domain - Number of Stripes - Cache Reservation - Number of Failures to Tolerate - Failure Balance Method - IOP limits - Disable Checksums - Force Provisioning - Space Reservation |
2 Required Options - Datastore Name - Datastore Capacity - Optional: Data Protection Policies o Local Snap Schedule o Replication Schedule |
Number of Options to Configure & Deploy VVols |
N/A |
2 Options - Register VASA Provider - Create VVol Datastore |
Grow a Datastore |
Yes |
|
Automate Datastore Capacity Increases |
Yes (add new ESXi host or add disks to existing hosts) |
Yes (via vRO) |
Shrink a Datastore |
Yes (eject ESXi host or disks) |
No (VMFS cannot shrink but all datastores are thin provisioned and only consume capacity of data stored) |
Support Multiple Datastores |
No |
Yes |
Automatic Data Reduction |
No |
Yes |
Automatic FTT=2 Data Protection |
No |
Yes |
Automatic Data at Rest Encryption |
No |
Yes |
Complete vSphere Update Manager without reducing Data Protection or Migrating data |
No |
Yes |
Complete Storage SW/FW Upgrade / Update w/o Reducing Data Protection |
No |
Yes |
Complete Storage SW/FW Upgrade / Update w/o Reducing I/O Performance |
No |
Yes |
Provides REST API Interface |
Yes |
Yes |
Supports PowerCLI |
Yes |
Yes |
Zero configuration required to increase storage performance |
No (change policy and rewrite data or create new cluster and migrate data) |
Yes (non-disruptive, in place controller upgrade) |
Zero configuration required to increase storage capacity |
No |
Yes (insert a data pack) |
Zero Performance impacting snapshots and replication |
No |
Yes |
Data can be cloned for access by Physical Servers and other hypervisors |
No |
Yes |
Reduce TCO via Storage Efficiency
Pure Storage is widely recognized as offering the most robust set of data reduction technologies in the industry, commonly delivering 3X to 4X greater data reduction and up to 8X greater total storage efficiency than HCI platforms. Efficiency spans beyond data reduction to include UNMAP and storage architecture overheads like data protection (RAID), caches, failover and slack space capacity reservations. At the end of the day storage efficiency directly translates into cost savings be it in measured in terms of virtual machines (VMs) per rack unit (RU), TBs per RU, GBs per dollar.
Feature |
VSAN |
FlashArray |
Additional hypervisor node required to add capacity |
Yes |
No |
All Inclusive Storage Software? |
No (vSphere Replication is extra) |
Yes |
Native Replication |
No |
Yes |
10% Storage Capacity Required for Cache? |
Yes |
No |
30% Storage Capacity Required for Slack Space? |
Yes (Cluster wide data rebalancing when any host hits 80% utilization) |
No |
Storage Capacity Required for Host Failures? |
Yes (percentage is calculated as N/N-FTT) |
No |
RAID Overhead |
33% (FTT=2 Erasure Coding) |
~12.5% (RAID-HA) |
Global Data Reduction |
No (per host disk group) |
Yes |
Data Deduplication |
||
Deduplication Granularity |
4 KB |
512 Bytes |
Inline Data Compression |
Yes (optional due to performance impact) |
Yes (always on) |
Multi-Algorithm, Background Compression |
No |
Yes (always on) |
Zero Removal |
No |
Yes (always on) |
Pattern Removal |
No |
Yes (always on) |
VMDK UNMAP / TRIM |
Yes |
Yes |
Datastore UNMAP / TRIM |
No |
Yes (automated via vRO) |
Reduce TCO via Server Efficiency
Expanding beyond storage efficiency, one needs to consider the cost placed on server CPU and memory resources when deploying VMs on a hyper-converged platform like vSAN. The benefit of the FlashArray can be measured in terms of an increase in VMs per hypervisor node. This results in fewer servers and software licenses; greater VMs per RU, VMs for Rack/floor tile, and VMs per network switch port.
Feature |
VSAN |
FlashArray |
Yes (estimated ~5-10%) |
Yes (estimated ~1%) |
|
ESXi CPU and Memory Overhead due to I/O Latency |
Yes (estimated ~5-10%) |
Yes (estimated ~1%) |
Software License Required for Storage |
No (VCF bundle) |
No |
Replication Software License Required |
Yes (vSphere Replication |
No |
Yes (estimated ~5-10%) |
No |
|
ESXi CPU and Memory Overhead for Data Encryption |
Yes (estimated ~5-10%) |
No |
Networking Details
Feature |
VSAN |
FlashArray |
Support FC (4/8/16 Gb) |
No |
Yes |
Support 10GbE |
Yes |
Yes |
Support 40GbE |
Yes |
Yes |
Support iSCSI LUNs |
Yes (In-Guest initiator) |
|
I/O Amplification (Increased Network & Disk Load) |
Yes (4.5X for FTT=2 with Erasure Coding) |
No |
Support Jumbo Frames |
Yes |
Yes |
Recommend or Require Jumbo Frames |
No |
No |
Security Certification Details
Feature |
VSAN |
FlashArray |
Support Data at Rest Encryption AES-256 |
Yes (via vSAN encryption) |
Yes |
FIPS 140-2 certification |
Yes |
Yes |
AES validation (FIPS 197 and SP 800-38A) |
No |
Yes |
DRBG validation (SP 800-90A) |
No |
Yes |
HMAC validation (FIPS 198-1) |
No |
Yes |
SHA validation (180-4) |
No |
Yes |
NIAP Certification
|
Yes |
Yes |
Common Criteria Compliant |
Yes |
Yes |