In today’s Business environment, providing High Availability is a must have capability, and a business necessity in order achieving Business Continuity. Can you imagine a financial firm loosing connectivity to its Business Critical Database in the middle of the day?
One of the essential components in ensuring continuous data access is Multipathing. Multipathing not only eliminates single points of failure between the server and the storage, but also helps in achieving better performance by leveraging algorithmic policies that balance I/O load to a device (LUN) in an intelligent and sophisticated manner.
At NetApp, we have a two pronged strategy as far as Multipathing is concerned. Support Native multipathing solutions for all OS platforms and support a 3rd party multipathing solution, that is open to all vendors so we can peacefully co-exist in today’s diverse and complex datacenters.
For the most, we’ve been successful, in accomplishing that goal, with the exception of one platform, VMware’s vSphere. With vSphere up until recently, we’ve only supported vSphere’s Native Multipathing solution (NMP) leveraging as a default the Round-Robin Path Selection Policy (PSP) for block protocols. While the solution works very well for us, over the years, we’ve gotten several requests to provide an alternative. I can’t we didn’t consider it, in fact we did, and to be more specific, we did have plans to write a NetApp specific PSP complementing the vSphere default (Round-Robin).
All that changed of course, when we were approached by Symantec who informed us, that they were porting their Dynamic Multipathing code (DMP) to vSphere. Given that NetApp fully supports Symantec’s DMP across all OSes, as an alternative to Native multipathing, and supports our multipathing strategy, we immediately jumped on board.
For those of you that have been in the IT industry a decade or so, I’m quite sure you know DMP and you’ve probably come in “contact” with it as some point. Especially if you were managing Solaris systems during the early part of the decade and later Windows, HP-UX, AIX and Linux. For the rest of you, DMP is Symantec’s Dynamic Multipathing solution. It’s been around for at least 10 years, it was developed by Veritas, which was eventually, acquired by Symantec and DMP itself, was part of the Veritas Storage Foundation Suite. In 2010 the solution was unbundled from the Foundation Suite and was sold as a standalone one.
DMP for vSphere is based on the 6.0 version of the Foundation Suite code, so in that respect, it’s not a v1.0 release. It’s is just an additional platform, and brings Symantec/Veritas’ decade long experience in device discovery and path management to the ESX(i) environments. Frankly, we’re happy not only to support it but also be able to sell it.
Symantec has worked closely with VMware to run their certification tests, to ensure that DMP works within the ESX framework and is on VMware’s HCL in conjunction with some of our arrays. We’re in process of qualifying every one of our platforms and as quals to get completed the HCL will reflect that.
DMP for vSphere consists of three main components:
Provides visibility to DMP within the vCenter UI interface
ESX(i) DMP Components
Installation of DMP packages supported through Update Manager or VMware CLI tools(vihostupdate).
Installed packages include DMP, ASL/APM packages, and the DMP CIM (VMware Common Information Model) Provider
Remote Command Line Interface (rCLI)
rCLI supported on Windows or Linux
Provides DMP for VMwareadm and vxddladm command line interfaces
Authentication through vCenter or individual ESX system credentials
The vCenter DMP Plug-in provides a centralized graphical management and like all plugins shows up as a separate. It can also be used to highlight the availability of graphical performance metrics at all points in the product, evidenced by the bar graphs near the bottom center of the graphic.
I/O policy changes can be done on the fly, as well as enable/disable of specific storage paths. It uses enclosure based naming so you can quickly identify the device and the array it sits on. Additionally,the array serial numbers, the array failover type, and the storage layout are also depicted.
Storage Performance & Statistics
From the vCenter tab, you have access to very granular statistics on a per device basis, enabling you to quickly identify bottlnecks, something which today, with the native NMP is not easily achievable. The stats auto-refresh every 15 secs.
DMP also has a Datacenter view, which enables the administrator to to have a single view of the storage deployed across the vCenter datacenter, with detailed statistics down at the LUN level and all the way to and virtual machine’s disk file.
The multipathing policies provided are the same tried and true policies being used for Unix and Linux environments furthermore, switching policies is an online event. Furthermore, DMP’s proactive path probing functionality, ensures that can take a path out of service proactively before I/O triggers failures. Should the path comes back, the path will reinstated as quickly as possible due to DMP placing priorities on re-enabling paths. The policy we will be recommending is MinQ, meaning I/O is scheduled according to the length of the I/O queue on each path. The path with the shortest queue is assigned the highest priority and gets the next I/O. With the highly random and variable I/O sizes hypervisors are producing, this policy ensures the highest performance and evenly balanced paths.
In closing,we support and resell DMP today, and we working thru qualifying the entire product line, with vSphere 4.1 and above. For those of you that have dealt with DMP before you know it’s a great solution. For those of you that haven’t, give it a shot should the opportunity present itself.