MetroCluster was the last major feature of 7-Mode to be ported over to Clustered Data ONTAP (it is included in the recently announced 8.3 version). Metro cluster solutions enable zero RPO and near zero RTO and they are typically a requirement for building a VMware Stretched Cluster (to enable vMotion, HA, DRS and FT over distance). Let’s take a look at how MetroCluster compares with EMC’s flagship continuous availability solution – VPLEX Metro:
MetroCluster is standard feature of ONTAP, rather than a separate product, and requires:
- A 2-node cluster at each of the two sites – all nodes in a MetroCluster need to be identical (FAS2500 series not supported)
- 2 x FC switches per site and 4 x FC ports per controller – dedicated to MetroCluster (for FlexArray the same switches are used for both MetroCluster and storage attachment)
- 1 x dual port FCVI card per controller – provides remote NVRAM mirroring
- 1-4 dedicated ISLs per switch – up to a maximum of 200 km (dark fibre or xWDM)
- 2 x FibreBridges for each disk stack – connects the SAS disk stack to the FC switches (not required for FlexArray)
- 2 x Ethernet ports per controller – used to replicate the cluster configuration between the sites
- Tiebreaker software (optional) – automatically triggers a switchover in the event of a disaster by monitoring the environment from a 3rd location
VPLEX Metro is a storage virtualisation appliance, that can simultaneously read/write to the same data across two data centres, consisting of:
- GeoSynchrony software – enables N+1 clustering, non-disruptive hardware and software upgrades, and the ability to virtualise storage
- 1, 2 or 4 Engines per site – each consisting of two high-availability directors
- 2 x FC switches per site – for connectivity of hosts and storage
- Inter-cluster connectivity – FC (dark fibre or DWDM) or IP with up to a maximum 10 ms RTT
- Host and storage connectivity – FC only
- Witness software – automatically makes storage available on the surviving site in the event of a disaster by monitoring the environment from a 3rd location
The core capability of both solutions is to provide continuous availability (zero downtime) – hosts are not impacted by the loss of local storage as the remote copy seamlessly takes over data-serving operations, but let’s see how they compare in other areas:
Ease of Use
Easy win for NetApp as MetroCluster is a standard feature, essentially it is a “set it and forget it” solution – any changes to the primary storage are automatically mirrored to the secondary.
Easy win for NetApp as MetroCluster is a standard feature therefore there is no additional charge for the software (additional connectivity hardware is required), whereas VPLEX requires a licence for all the storage managed as well as additional hardware appliances.
Advanced Storage Features
Easy win for NetApp as MetroCluster supports nearly all of the features of Clustered Data ONTAP (i.e. de-duplication, compression, snapshots, integrated data protection and NAS), whereas VPLEX only provides storage virtualisation and non-disruptive operations (i.e. LUN/array migration).
The only features not supported by a MetroCluster are Infinite Volumes, NSE drive encryption, disk partitioning on the root aggregate and SSD partitioning for Flash Pool.
Connectivity and Scalability
- Inter-site connectivity – easy win for EMC as VPLEX can replicate over both FC and IP, MetroCluster is limited to FC
- Host connectivity – easy win for NetApp as MetroCluster supports FC, FCoE, iSCSI and NFS, VPLEX is limited to FC
- Scalability – easy win for EMC as VPLEX supports eight directors (controllers) per site, MetroCluster is limited to two
Planned and Un-planned Site Failure
With MetroCluster a volume or LUN is online in only one cluster at a time, client/host access is not possible on the remote cluster unless a switchover is performed. Switchover operates at the site level – all aggregates, volumes, LUNs, and SVMs will switchover to the other site. The configuration is active-active, so that each cluster can serve its own separate workloads while providing DR protection for the other.
VPLEX is far more flexible as it is able to “stretch” a LUN across sites and allow hosts at each site to have read/write access to the local version of the LUN. With VPLEX there is no concept of a switchover as LUNs are simultaneously active on both sites.
Both solutions support disaster avoidance (i.e. planned site failure) by manually taking down a cluster at one site such that all storage is active on the remaining site – in the event of an un-planned site failure the disaster recovery process can be automated by the Tiebreaker/Witness software.
External Array Virtualisation
Win for VPLEX as it supports more external array platforms and can utilise data on an existing LUN – NetApp FlexArray can virtualise external arrays, but the data on the LUNs must be destroyed before they can be used.
It is important to note that MetroCluster can be deployed with internal disks only, with external disks only or a combination of the two – VPLEX does not support internal disks.
VMware Metro Storage Cluster (vMSC) support
Both MetroCluster and VPLEX have full support for vMSC and therefore will enable vMotion, HA, FT and DRS between data centres.
One significant advantage of VPLEX is that it does not require the hosts to be configured to access the storage at both sites (cross-cluster connect), VPLEX does support cross-cluster connect and there are some availability advantages to doing it, but it is not mandated. This makes it possible to move Virtual Machines from one site to another and have both the compute and storage resources delivered locally – with MetroCluster the storage is only ever active on one site.
MetroCluster does have more flexible protocol support – FC, FCoE, iSCSI and NFS, VPLEX is limited to FC.
So which is the best?
MetroCluster wins with its simplicity, advanced storage features and lower cost, and VPLEX wins with its ability to replicate over IP and to simultaneously access the storage at both sites. Therefore there is no clear winner, it comes down to which matches your requirements and budget the best, but there could be if:
- NetApp were to add in some key missing features (i.e. replication over IP, volume move across sites, multi-node scalability and simultaneous LUN access)
- EMC were to integrate VPLEX into their storage platforms and support block and NAS protocols
So there you have it, hopefully a balanced view of the two solutions, and as always comments would be appreciated.
- Enabling 24×7 Continuous Availability with a vSphere Metro Storage Cluster
- Cost-effective Metro Storage Cluster solution now available from NetApp
- Is NetApp Clustered Data ONTAP finally ready for primetime?
- VPLEX Metro now more affordable for VNX customers