When Veeam Backup & Replication v11 went Generally Available on the 24th of February I posted the What’s in it for Service Providers blog. In that post I briefly outlined all the new features and enhancements in v11 as it related to our Veeam Cloud and Service Provider Partners. As mentioned each new major feature and enhancement listed below deserves its own seperate post. While these posts are targeted at Service Providers, the majority of these features can be levered by all types of organizations. In this post I am focusing on the long awaited Continuous Data Protection feature and how VCSPs can still take advantage of it in v11.

As a reminder here are the top new features and enhancements in Backup & Replication v11 for VCSPs (with links as created)

What is the CDP

Continuous Data Protection (CDP) is built on-top of the vSphere APIs for I/Os and allows VMs to be replicated at sub minute RPOs with minimal impact to the workloads that are being run on the VMs. Essentially we are intercepting I/Os as they happen at a pre-defined policy internal and sending them from a source to a target vSphere Host as a VM replica. The VAIO mechanism allows those lower RPOs than what you get in traditional snapshot based replication. For snapshot based replication the RPO is generally set to 15 minutes… but can be as low as 5 minutes when on vSAN.

Continuous data protection (CDP) is a technology that helps you protect mission-critical VMware virtual machines when data loss for seconds or minutes is unacceptable. CDP also provides minimum recovery time objective (RTO) in case a disaster strikes because VM replicas are in a ready-to-start state. CDP constantly replicates I/O operations performed on VMs. To read and process I/O operations in transit between the protected VMs and their underlining datastore, CDP uses vSphere APIs for I/O filtering (VAIO) that gives an option not to create snapshots. Because CDP is always on and does not create snapshots, it allows reaching a lower recovery point objective (RPO) compared to the snapshot-based replication — near-zero RPO which means almost no data loss.

With CDP, there is now the ability to have a short-term retention window. Within this window, backup admins can set the tolerated level of data loss via recovery point objectives starting from as low as two seconds. Within this short-term retention window, data is replicated based on the configured RPO, and recovery is as simple as sliding back to a point in time within this retention window. For longer-term retention of CDP-protected workloads, backup admins can create additional restore points more aligned to traditional points in time, based on an hour value and then stored at the target site for a set number of days. These can also be configured for application-aware processing and will always have a crash consistent point to fail over to.

v11 CDP Service Provider

I have given a number of sessions on CDP over the last couple of years and have included one from Tech Field 20 at the bottom of the post:

Simplicity, Scalability and Performance

Scalability-wise, with Veeam CDP, you can start small and grow as CDP requirements change over time. Depending on the number of virtual machines, the RPO policy and the amount of data being created on the virtual machines, VMware CDP proxies can be deployed as you grow without the need to deploy one per host across the cluster. This is significant as other CDP solutions generally require a blanket approach to data movers. Scaleable CDP Proxies significantly reduce and optimize the resources required and allows targeted configurations against vSphere clusters where only the Tier One workloads are being protected.

Installation is handled at the cluster level with VAIO Filter Drivers installed on each host in the source and target cluster.

To help with right-sizing, the policy wizard has a built-in CDP Infrastructure Assessment tool that looks at what resources are required at a CPU, RAM and network bandwidth-level, as well as shows the required bandwidth of virtual machines configured in the policy. The source proxy aggregates changes and stores them in RAM and on disk if required. All data is compressed and encrypted in flight. for CDP replication, networking is critical and bandwidth can become a constraint, however between clusters, 1Gbit/s is required with a max latency of 80ms recommended.

You can failover to the latest state or go back to a point in time that is crash consistent or leverage long term retention points which can be app aware. It must be noted that failbacks are done with existing replication functionality but has enhanced features in v11 around automatic, schedules or manual modes.

Benefit to Service Providers

While this feature isn’t specifically built for Service Providers in that there is no integration with Cloud Connect Replication or support for VMware Cloud Director to Cloud Director Replication, the fact is that a lot of tenant VMs still operate on a vCenter which means that CDP can still be utilized for Managed Services Clients with remote vCenter deployments, or even can be used within Service Provider IaaS platforms where VMware Cloud Director isn’t deployed. Through automation it should be possible to productize the feature and have it part of a DRaaS service offering. Remembering that CDP is all about protecting the Tier 1 workloads which we roughly estimate to be about 3-5% of all current workloads. This means that with v11, Service Providers can begin to offer a targeted premium (but still cheaper than some incumbents) service that works together with traditional replication, point in time backups and offsite/archival backups… this is end to end now!

CDP Content and Materials