Host Access Methods:
VPLEX offers two Host Access Methods enabled by the AccessAnywhere Feature (active-active LUN ownership model between directors and sites)
These are called Uniform and Non-Uniform Host Access and are shown in following picture for a VMware Metro Storage Cluster. The picture is also correct if you swap ESX by Hyper-V or any physical Server with a Cluster Filesystem. Please note that this is only a logical picture to keep it simple.
Actually the only difference are the additional FC paths from the hosts to the remote VPLEX directors (“Path 2” and “Path 4”). At anytime you can switch between both models online by adding or removing the paths through zoning and LUN masking.
In VPLEX wording Uniform Host Access is called “cross connect”. The cross connect can share the same redundant fabrics and therefore the same ISLs used by the VPLEX interconnect.
Having an ISL with a non neglectable RTT you want the paths “Path 2” and “Path 4” only be used in a failure scenario to keep the IOs local to the host.
As long as your round trip time (RTT) between both sites is lower than 1ms, both access methods are supported and you have the choice. There are good reasons for both options but won’t be discussed here.
Up to 10ms RTT you have to use a Non-Uniform Host Access method.
The VMware ESXi build in multipathing software has two reasonable policies: “Round Robin” and “Fixed Path”.
A challenge with fixed path is, that only one path will be actively used by each LUN and when this path fails you can not control which path will be used next. In an Uniform Host Access environment it could be possible that a path to the remote site will be used, which introduces unnecessarily additional latency and load on the ISL. Furthermore it takes some effort or a good script to balance the primary path of each LUN on each host to all available paths and targets.
The next option, “Round Robin”, will send by default 1000 IOs to one path and then select randomly a new path for sending the next 1000 IOs and so on. Again this will introduce additional latency and ISL bandwidth in an Uniform Host Access environment for the times where the remote paths are used. In contrary to “Fixed Path” where this only would happen in a failure scenario does this (negative) behavior happen for “Round Robin” every time the algorithm selects a remote path.
To overcome the limitations of many native multipathing implementations does EMC offer a multipathing software available to ESXi, Windows, Linux and Unix. In case of VMware the product is called Powerpath/VE (Virtual Edition) and makes use of the VMware Pluggable Storage Architecture (PSA).
Please have in mind that Powerpath is not mandatory to connect a ESXi host to VPLEX.
Powerpath offers a great number of features. To put some of it down, it can leverage all available paths for a LUN concurrently. To distribute the IOs efficiently does Powerpath measure among others the target port queue length. Therefore if you have a larger host environment you don’t have to do any manual configuration to efficiently use the available resources. Especially for a virtualization host with its very variable workloads can Powerpath increase your overall performance. (See )
Powerpath also offers a central management virtual appliance, “Powerpath Viewer”, for your environment.
The last but most powerful feature for the course of this blog is “autostandby”. Powerpath is measuring the latency of all paths and will automatically detect the paths that are not local to the hosts. These remote paths will be set to standby and only be enabled when the last local path fails.
The following video demonstrates the autostandby feature of Powerpath/VE with a VPLEX Metro installation. The VPLEX installation consists of one engine (two directors) and three backend storages (VMAX, VNX, CLARiiON) in each site.
VPLEX is used by a VMware vSphere 5.5 Metro Storage Cluster demo installation. The ESXi hosts (one in each site) are using Powerpath/VE 5.9.
With Powerpath/VE and VPLEX access anywhere you will have a VMware environment which automatically configures, monitors and reacts on failures+changes in an smart manner.
Good to read:
 ESG Lab validation – Powerpath/VE
 Implementing vSphere Metro Storage Cluster (vMSC) using EMC VPLEX
 VMware vSphere 5.5 vMotion on EMC VPLEX Metro – Performance and Best Practices
 USING VMWARE VSPHERE WITH EMC VPLEX – Best Practices Planning