Zerto Virtual Replication 1.0 is a new software-based solution for Disaster Recovery/Business Continuity which got a lot of attention since its release in 2011. It won the Techtarget Best of Show award at VMworld 2011. It also won the Product of the Year 2011 Gold award in storage & backup from SearchServerVirtualization.com.
Zerto Virtual Replication competes with VMware Site Recovery Manager. Another DR solution for vSphere is VirtualSharp ReliableDR. For an overview of the three solutions see my posting here.
This posting will give an overview of the product, its features, architecture and pricing. In a next blog posting I will give more insight in the configuration and technical details and specifically the replication feature. As the name of the product shows, replication is what Zerto Virtual Replication is good at.
What it does:
The company name Zerto stands for Zero RTO and is an Israel based software company. Their product named Virtual Replication enables organizations to protect their application data and virtual machines running on VMware hosts against disasters and major incidents by near-synchronous, storage agnostic replication of data to another site.
Storage agnostic means Virtual Replication can be used to replicate from any storage platform to any storage platform supported by VMware. Since the software operates at the hypervisor level the brand of storage (HP, Dell, IBM, EMC, NetApp etc.) in the protected site can be different than the one in the recovery site. This enables Disaster Recovery as a Service as the service provider does not have to own the same brand of storage as the customer. Also a storage adapter (piece of software which acts as a gateway between the storage and DR software) is not needed.
A major advantage of hypervisor-based replication is that replication is notbased on LUN replication but on VM replication. For example in Site Recovery Manager when a virtual machine disk file is moved (storage vMotion, Storage DRS) to another datastore located on a non-replicated LUN you will lose the protection. As Storage DRS moves virtual disks automatically between datastores protection can easily be lost if LUN-based replication is used.
Not by using Zerto. Zerto replicates virtual machine disk files independent from its location on storage and type of storage (NFS, local storage, VMFS, etc)
How it works
Zerto virtual Replication does not use VMware virtual machine snapshots or disk array snapshot technology to have a point in time (frozen) state of the data which can be replicated. VMware snapshots can have a severe negative effect on the performance of virtual machine and can even result to the virtual machine being unavailable for some moments because of stunning. Those issues are explained in my blog posting titled Know the performance impact of snapshots used for backup !
Instead of creating snapshots at a regular, admin defined interval all SCSI data sent from the protected virtual machines to the storage layer at the protected site, are forked/split/copied in the ESX(i) hypervisor layer and sent to the Zerto Virtual Appliance. This is done using the VMware iSCSI filter.
How the vSCSI filter works is explained in this article titled VMware’s vSCSI filter: What it does and how it works . The data received by the Zerto Virtual Appliance is stored in a journal and then replicated to the recovery site. At the recovery site the data is received at the Zerto Virtual Appliance and stored in a journaling file. From the journaling file the data is transferred to the virtual disk files of the vm’s.
Zerto Virtual Replication supports ESX(i) 4.x hosts and vSphere ESXi 5 hosts . Hosts running ESX(i) 3.5 or lower are not supported. A vCenter Server 4 instance at the protected site and at the recovery site is required for installing Zerto Virtual Replication. Also required is a supported Windows Server operating system (Windows 2003, 2008, 2008 R2 and Windows XP SP3 and higher). All editions of vSphere are supported. Unlike Site Recovery Manager 5 which supports vSphere Standard Edition or higher, but not the Essentials and Essentials Plus editions.
Zerto Virtual Replication components
A Zerto Virtual Replication Appliance needs to be installed at each host with protected and replica VM’s. The appliance consumes around 6 GB of disk space. This is a Linux based virtual machine which takes care of several tasks.
At both the protected and recovery site the Zerto Virtual Manager needs to be installed. This is a small Windows component . This will take care of the replication and integrates with vCenter Server on both sites.
One of Zerto’s strongest assets is its replication feature. First of all it has WAN resiliency. This means if the WAN connection between the protected and recovery site is broken, the replication data will be queued. When the connection is restored the replication will continue without any data lost. Also the replication data is compressed. Replication traffic can be capped so it will not consume all available bandwidth. Replication traffic can be prioritized so that TIER1 applications get their replication data first at the protected site.
When replication of data to the recovery site is performed for the very first time, all the data of protected virtual machines needs to be transferred. When using a WAN connection this could lead to very long transfer times and congestion on the WAN. Seeding is the ability to transfer the data to removable media, ship that media to the recovery site (FedEx, UPS etc.) and import the data into the replica VM. From that point only incremental data needs to be transferred over the WAN.
Zerto Virtual Replication latest version 1.0u5 supports seeding using the user interface.
Disaster Recovery and Business Continuity is not about restoring virtual machines. It is about restoring applications with consistent data such that they become operational within the RTO and RPO limits. To be able to recover an application, and this is especially true for databases, to a consistent state the snapshot taken of the database needs to be consistent. This means that outstanding transactions are written to the database, the database is quiesced and only than a snapshot should be created.
To enable application consistent snapshots Zerto delivers the ZertoVSSagent. This software, available for 32-bit and 64-bit Windows servers is installed on each server which needs consistency.
The VSSagent is delivered as a command line tool and needs to be executed by a scheduled task at the target server. Unfortunately the ability to create application consistent snapshots cannot be managed and monitored using vCenter Server.
Replication continues during testing.
In case a DR testing is performed in the recovery site, replication continues. So the data is always up to date. Even when during a recovery test a failover needs to be done this can be done with the RPO specified.
Point in time recovery
Zerto Virtual Replication does not use a schedule to save data to a replica. Data is replicated all the time to the recovery site. This means data can be recovered to any point in time. One minute ago or 5 hours ago.
At some time after the failover virtual machines needs to be moved back to the protected site. Zerto Virtual Replication supports an automated failback. By pressing a button the data is replicated from the recovery site to the protected site and virtual machines are restarted in the protected site.
There are situations in which you know in advance that it is wise to move virtual machines to the recovery site. A flooding or hurricanes does not arrive all of a sudden. The need for a migrations of datacenter (disaster avoidance) are known in advance as well. Zerto Virtual Replication has a Move feature. When this option is selected, virtual machines are shutdown in a nice way after a snapshot has been taken. In the recovery site the VM’s are then booted.
Zerto Virtual Replication 1.0 does not have an automated runbook like Site Recovery Manager has. A runbook will start virtual machines in a predefined order and using dependencies so a vm will only start if a required vm has been started first. For testing and disaster recovery virtual machines will need to be started from vCenter. Zerto utilizes vCenter boot order capabilities rather than its own internal runbook in order to enable configuration of boot order for production and DR in one place.
Replication to the cloud
Zerto expects to release a new version of the product in Q2 2012 which will support replication to service providers. By using Virtual Replication multiple customers can use a shared infrastructure in the cloud serviced by a Service Provider (Disaster Recovery as a Service or DRaaS)
Zerto Virtual Replication is licensed per protected virtual machine.
Links to blog and other sites
Zerto gave a presentation at Virtualization Tech Field Day #2 in February 2012. See a good overview of Zerto at Musings of Rodos.
VMware SRM and Zerto Data Replication: The Art of Disaster Recovery good article describing the use case of Zerto Virtual Replication.
Zerto Replication and Disaster Recovery the easy way by Gabrie van Zanten. Very comprehensive report on the features and technique of Zerto Virtual Replication.
Q&A with Ziv Kedem of Zerto at Virtual Strategy.com
Read more: Zerto: Or What I Learned at Tech Field Day #6! by Ed Czerwin
The Secret is out and it’s called Zerto. by Tom Howarth
Zerto by Christopher Wells at vSamurai.com
Hypervisor-Level Replication with Zerto by Virtualization Review
Steven Johnston has a nice blog describing the re-IP process using PowerShell