Earlier this year it was announced that vSAN had grown to over 7000 customers since launch, which is a pretty decent number given the product went GA just over 3 years ago and we’re on the 6th iteration! What’s even more impressive is how quickly VMware are turning these updates around (almost every 6 months we get an update of sorts), we only got vSAN 6.5 at VMworld last year and 6 months later we now have version 6.6 – what’s funny is half my customers haven’t even started implementing their 6.5 upgrade plan yet and now they will have to re-write that plan…. Lol… =)
In fact I see the number of customers growing quite significantly this year given the huge drive towards HCI – something that I’m seeing within my company’s customer-base (and in the market in general)!
Today sees vSAN 6.6 go GA, and it amazes me on how many new features VMware have packed into this release – features that make vSAN more faster, cost effective and much more secure! And to think that this is just a “minor” patch release! With vSAN 6.6, customers can now evolve their data centre without risk, control IT costs and scale to tomorrow’s business needs (sorry, that was a marketing blurb that I just had to fit in somewhere as it sounded good).
(Note: I know that slide says “Not for distribution”. However, the vSAN vExperts have been given permission to use the material in their blogs)
The biggest features in my opinion are vSAN Data-at-Rest Encryption, Unicast communication and Enhanced Stretched Clustering with Local Protection – these are the 3 features I’m going to concentrate on within this post, trying to expound on all the new features would involve me writing a lengthy technical whitepaper! =)
That said, other new features are as follows:
- ESXi Host Client (HTML-5) – management and monitoring functionality available on each host in the case where vCenter server is offline.
- Simpler installation/configuration – The ability to create a single node vSAN datastore by using the vCSA installer and then allowing you to deploy vCSA/PSC onto that vSAN datastore.
- Enhanced rebalancing – allowing large components to be split up during redistribution.
- Site Affinity in Stretched Clusters – a new Affinity policy rule allows users to request where a VM gets deployed to, although this is only applicable when the PFTT is set to 0. Although it’s worth noting that DRS/HA rules should be aligned to data locality!
- Always-On Protection – Enhanced repairs with Re-sync traffic throttling – allowing vSAN to respond to failed disks/nodes more quickly, intelligently and more efficiently. New Degraded Device Handling (DDH) intelligently monitors the health of drives and proactively evacuates data before failures can happen.
- Maintenance Pre-Check – enhanced checks to ensure there are enough resources for vSAN when entering maintenance mode (or decommissioning vSAN nodes).
- Stretched Cluster Witness Replacement UI – simpler method of changing the Witness host without having to disable the Stretched Cluster.
- vSAN Cloud Analytics – pro-active, real-time support notifications and recommendations with real-time custom alerts through the vSAN health Service.
- API enhancements – vSAN SDK updated to handle all new features, with additional enhanced PowerCLI support.
- vSAN Config Assist / Firmware Update – Enhanced health monitoring and HCL checks using health-check assistant to ensure the vSAN hardware has the latest firmware and drivers installed.
- Enhanced Performance – up to 50% higher all-flash IOPs performance per host and Health Monitoring
- New Hardware Support – Support for Intels new Optane technology, NVMe SSDs and larger 1.6TB SSDs for cache drives.
- Support for Photon Platform 1.1 as well as a Docker Volume Driver – great for customers (ie DevOps) who prefer working with micro-services/containers. This allows customers to use vSAN as storage for Docker VMs giving them the ability to apply storage based polices (such as FTT, QoS, access permissions, etc) to the VM, it also gives customers the ability to support persistent storage to allow stateful container apps to be built (such as DBs).
EMC love calling this by the acronym D@RE…. But this hasn’t quite filtered down to the VMware team…. =)
VMware vSAN 6.6 introduces the industry’s first native HCI security solution with software-defined data-at-rest encryption within the hypervisor. Data-at-rest encryption is built right into the vSAN kernel, and is enabled at the cluster allowing all vSAN objects to be encrypted (ie the entire vSAN datastore).
In my opinion this is one of the most important new feature in vSAN 6.6 – we all know that security within IT has become top priority, featuring very high on a company’s risk-register, but IT Admins have always been reluctant to either deploy encryption at the OS level or let application owners encrypt their apps and data. Data-at-rest encryption takes away that decision by encrypting when the data resides on your vSAN Datastore.
It’s hardware-agnostic which means you can deploy the storage hardware device of your own choice – it doesn’t require the use of expensive Self-Encrypting Drives (SEDs)!
vSAN Encryption is available for both All-Flash and Hybrid configurations and integrates with KMIP 1.1 compliant key management technologies. When vSAN Encryption is enabled, encryption is performed using an XTS AES 256 cipher and occurs both at the cache and capacity tier – wherever data is at rest, which means you can rest assured that if a cache or capacity drive is stolen the data is encrypted! Plus vSAN Encryption is fully compatible with vSANs all-flash space efficiency features, like dedupe, compression and Erasure Coding, delivering highly efficient and secure storage – as data comes into the cache tier it’s encrypted, then as it de-stages it’s decrypted and any relevant dedupe or compression occurs to the data (4k blocks) before it’s re-encrypted as it hits the capacity tier (512b or smaller blocks). As it’s data encryption at rest, I believe that vSAN traffic traversing the network maybe sent in the clear which means you will need to ensure vSAN traffic is protected accordingly.
It’s worth mentioning that whist the cryptographic mechanics are similar to VM encryption that was introduced in vSphere 6.5 (ie it requires a KMS and uses the same encryption modules), there is a vast difference in the way they’re implemented – VM encryption is per-VM (via vSphere API for IO filtering – VAIO), whilst with vSAN encryption it is the entire datastore. Also you get space-saving benefits from vSAN encryption as previously mentioned. The other major difference is that vSAN encryption can carry on functioning if vCenter Server is lost or powered off because the encryption keys are transferred to each vSAN host and via KMIP each host talks directly to the KMS, whereas VM encryption requires you to go through vCenter Server to communicate to the KMS. Not to mention VM-encryption does have some performance impacts and requires Ent Plus licenses.
Turning on vSAN encryption is as simple as clicking a checkbox within the settings of the vSAN cluster and choosing your KMS (which does need to be setup prior to enabling encryption). However, it’s worth noting that a rolling disk reformat is required when encryption is enable which can take a considerable amount of time – especially if large amounts of data residing on the disks must be migrated during the reformatting.
With the enhanced API support, customers who like to automate their infrastructure will be able to setup an encrypted vSAN cluster with all the relevant KMS configuration via scripting – great for automating large scale deployments!
Removal of Multicast
Another big announcements with vSAN 6.6 is that VMware are switching from multicast to unicast for their communication mechanism. This obviously makes networking a lot simpler to manage and setup as customers won’t need to enable multicast on their network switches, or IGMP snooping, or even PIM for routing. It may even mean that customers could use cheaper switches (which may not handle Multicasting very well).
Bit of background:
Typically IP Multicast is used to efficiently send communications to many recipients. The communication can be in the form of one source to many recipients (one-to-many) or many sources to many recipients (many-to-many).
vSAN used multicast to deliver metadata traffic among cluster nodes for efficiency and to optimise network bandwidth consumption for the metadata updates. This eliminates the computing resource and network bandwidth penalties that unicast imposes in order to send identical data to multiple recipients. vSAN depended on multicast for host discovery – the process of joining and leaving cluster groups, as well as other intra-cluster communication services.
While Layer 3 is supported, Layer 2 is recommended to reduce complexity. All VMkernel ports on the vSAN network subscribe to a multicast group using IGMP. IGMP snooping configured with an IGMP querier can be used to limit the multicast traffic to only the switch ports where the vSAN uplinks are connected to – this avoids unnecessary IP multicast floods within the Layer 2 segments.
Although one of the issues that could occur was when multiple vSAN clusters reside on the same layer 2 network – the default multicast address should be changed within the additional vSAN clusters to prevent multiple clusters from receiving all multicast streams.
I believe vSAN now relies on vCenter Server to determine cluster membership, however I haven’t yet read about how the vSAN team have managed to implement unicast communication as that information is still in limited supply. It’ll be interesting to understand how they have done it considering multicast was an efficient and easy way of replicating instructions to multiple nodes within the vSAN cluster when a node needed to perform an action. Although one thing worth noting is that unicast communication probably lends itself to cloud platforms a lot easier than trying to implement a multicast solution!
Local Protection for Stretched Clusters
Stretched vSAN Clusters were introduced back with vSAN 6.1 and built on the foundations of Fault Domains, it was basically a RAID-1 configuration of a vSAN object across two sites – which basically means a copy of the data in each site with a witness site for cluster quorum type services during failure events. The problem was if 1 site failed you would only have a single copy left and an additional failure could lead to data loss. It also meant that if a single host failed in any of the sites then the data on that host would need to be resynced again from the other site (to rebuild the RAID-1).
This new enhancement to Stretched Clusters now gives users more flexibility with regards to local and site protection. For example, you can now configure the local clusters at each site to tolerate two failures whilst also configuring the stretched cluster to tolerate the failure of a site! Brilliant news!
When enabling Stretched Clusters, there are now two protection policies – a “Primary FTT” and a “Secondary FTT”. Primary FTT defines the cross-site protection and is implemented as a RAID-1. It can be set to 0 or 1 in a stretched cluster – 0 means the VM is not stretched whilst 1 means the VM is stretched. Secondary FTT defines how it is protected within a site, and this can be RAID-1, RAID-5 or RAID-6.
One thing to note is that the witness must still be available in order to protect against the loss of a data site!
This new feature doesn’t increase the amount of traffic being replicated between sites as a “Proxy Owner” has been implemented per site, which means instead of writing to all replicas in the second site, a single write is done to the Proxy Owner and it’s then the responsibility of this Proxy Owner to write to all the replicas on that local site.
So that’s about it for now…. if you require more information then pop along to the following sites:
Duncan Epping (Chief Technologist in the Office of CTO for the Storage & Availabiliy BU at VMware) has created some great demos of vSAN 6.6 which can be found on his blog site: http://www.yellow-bricks.com
Things to Note
The underlying release for vSAN 6.6 is vSphere 6.5.0d which is a patch release for vSphere 6.5. For existing vSAN users upgrading to vSAN 6.6, please consult VMware Product Interoperability Matrices to ensure upgrading from your current vSAN version is supported.
Please note that for vSAN users currently on vSphere 6.0 Update 3 – upgrade to vSAN 6.6 is NOT yet supported.
The parent release of vSAN 6.6 is vSphere 6.5 and as shown by VMware Product Interoperability Matrices, an upgrade from 6.0 U3 to vSphere 6.5 (and hence vSAN 6.5) is NOT supported. Please refer to this KB Supported Upgrade Paths for vSAN 6.6 for further details.
p/s: I’ve always liked Rawlinson Rivera‘s Captain vSAN cartoon!! =)