This self paced workshop demonstrated 3 key components of PowerPath: Virtual Multipath Management, Encryption, and Migration Enabler. The presenters for the session were Rob Pellowski, Brian Stultz, and Eddie Thigpen. This H.O.W (Hands on Workshop) was one I was looking forward to. My primary focus was on PowerPath/VE, I spent little time allocated to the other components and their demos due to the lack of time.

As we all know PP/VE will be offered as an add on to both ESX 4.0 and 2008 Hyper-V. Here are some of the notable features:

• Dynamic Load Balancing
• Dynamic path failover and auto recovery
• Device prioritization
• Monitoring and reporting of I/O stats
• Auto restore of paths
• Auto path testing
• ESX cluster support
• Non-EMC array support
• Remote management using vSphere CLI

PP/VE taps in to VMware’s Pluggable Storage Architecture  (PSA). This allows PP to operate within the storage I/O path via VMkernel API’s. PP/VE can coexist with NMP (Native Multi Path) which is used on arrays not supported by PP.

Installation takes place remotely either from a Windows host or a RHEL 5 host using the vihostupdate util (as part of vSphere utility kit?). You do not initiate the installation from the ESX directly. Post installation of PP does require a reboot of the ESX host.

Once the installation is complete and the host is reboot, PP is managed through remote CLI using the rpowermt (powermt equivalent) command set. This allows one to manage and configure PP/VE as well as any paths under its control. Multiple hosts can be managed from a single host simply by specifying the IP or FQDN of that particular host. Chatter between the ESX host and the remote CLI host is via CIM or Common Information Mode). An RSA lockbox stores all creds and other pertinent Info on the RC host. The following are some common command sets with rpowermt tool.

• rpowermt check registration-displays licensing info
• rpowermt config-updates Clariion nice names
• rpowermt display-shows HBA, storage arrays, optimization settings, I/O to name a few for each device
• rpowermt display latency-monitors latency
• rpowermt display options-shows management and auto restoration of paths of all array classes

Other commands without descriptions:

• rpowermt display paths
• rpowermt display ports
• rpowermt register
• rpowermt restore
• rpowermt path_latency_monitor
• rpowermt path_latency_threshold
• rpowermt periodic_autorestore
• rpowermt set policy/set priority/set mode/update lun_names/unregister/version

----Points of Interest----

• Errors represented with the rpowermt display are indicative of the total number of times the path has transitioned from “alive” to “dead”.
• PP will load balance across to paths to a Clariion on the same SP. Therefore it is recommended to have at least 4 paths to the Clariion whether ALUA or PNR mode is in use.
• Only devices managed by PP will have a pseudo name. This name will not be presented in the OS.
• Up to 32 paths are supported for a single device. Load balancing and failover algorithm are optimized according to the array in use.  Available policies are SymmOpt, ClarOpt and Adaptive for third party arrays.
• Generally PP/VE is monitored and managed using the rpowermt command with the remote CLI. vCenter can be used, however, to identify whether a device is owned by PP or NMP. Future, more granular integration is expected.
• PP supports FC, iSCSI and FCoE. Additionally full interop with DRS/VMotion and other layered feature sets.
• PP is VM independent, the guest OS has no knowledge that PP is managing the devices that it lives on.
• PP has a unique position in the I/O stack, right above the HBA but below the filesystems, applications and guest OS’s.

 

 

This session focuses on new advancements with EMC’s PowerPath particularly in the areas of virtual workloads. Note: This post documents some of my notes during the session, so please be tolerant of the flow.

PowerPath, as you know,  is one of the most successful licensed products from EMC with over 800,000 licenses and across 20,000 sites world wide. To further advance this momentum and the virtual revolution, EMC has developed PowerPath VE (or Virtual Edition). This in essence brings path management to virtual environments, with support for both VMware ESX and  Hyper-V.

Focus was spent on the following areas primarily…

• Configuration Checker-This essentially is a ~19 point check against your host for any potential road blocks with new and upgrade installs of Power Path. CC was released the week of May 11th 2009 and is available from PowerLink.
• Path Latency Monitoring or PLM- Common use cases for this, and its a big one, is fabric bottleneck detection. The ability to detect anomalies within your SAN traffic based upon predefined thresholds (I/O response times essentially)
• Migration Enabler Host Copy-Gives you the ability to migrate, non-disruptively, from source to destination LUNs   without the need for underlying replication. Support for Symmetrix and Clariions, non-EMC arrays by RPQ only.
  • TimeFinder/clone-use case, movement of data between tiers in a Symmetrix
  • AIX and Linux support as of late, Windows and Solaris initially
  • Support for migrations from thick to virtual and virtual to virtual provisioned LUNs
  • Remote event monitoring via SNMP

Accomplishments-

• Automated path management, simplified channel management for VDC
• Same PP CLI commands as in the physical world
• Notable improvements in IOPs, response times and CPU utilization
• Predictable service levels, as channels go offline, as there are periods of I/O bursts,  as VM’s are moved around, PP adapts and continues to balance across the available channels.
• Built into hypervisor as part of the storage API’s for multipathing. Essentially plugs into PSA framework.
• Supports ESX 4.0 only and essentially whatever guest OS VMware supports
• Arrays-Symmetrix, Clariion ,HP, HDS, IBM via RPQ-FC and iSCSI
• Interop with VMware layered software (HA, DRS, etc)
• All channels are active and tunable with QoS

Note: Mention of PowerPath release 5.3 due out in June ‘09

Packaging and Licensing

• As of July 2009 there will be a couple different flavors of PP Standard  (2 x86 models/2 non x86 models). The goal here is to simplify the complexities around pricing today with this product.
• Controlled via ELM or Electronic Licensing Management same mechanism that VMware is using.
• This will give you the ability to to download software and licensing from one site called within PowerLink.
  • Token based, allows user to retrieve a key based upon a LAC letter sent to the customer. LAC letter encompasses instructions on how to obtain license and software binaries.
• There are 2 separate license types
  • Served licenses-Simply put, licenses that are tracked and licensed via a license server, licenses can float. Use case for 25 ESX servers or more.
  • Unserved licenses-What typically is accomplished today, license is locked to host. Use case is for environments that are less than 25 ESX servers.
• VE is a separately charged license, current licensing for PP for physical does not entitle you to free licensing for VE.
• Licensing is per socket not per VM, same licensing model as VMware.

Customer experience with PP/V (ComputerShare w/David McCue)

• David started with a quick rundown of what CS does: transfer agency core/maintain list of share holders for companies
• Snapshot of Environment-over 200TB of storage, 150 blades, 1000 VM’s. Using SRDF/S for backend replication. <1ms SAN fabric latency, crossite zoning in place.
• Problem before-NMP “round robin” not distributing load across all paths. Common problem as NMP only utilizes a single path.
• Problem after-VM creation 24% faster using Symm Opt Policy rather than Basic

Additional benefits-

• Easy ID of Symm device IDs
• PP/V reduces administrative costs, easily “learned” technology for non-storage staff
• Auto Restore-facility that scans dead paths and determines whether or not paths are valid again
• ComputerShare gives a thumbs up for PP/V, will continue geo rollout of this product going forth.

Forward looking-Centralized console for PP across your fabric is currently being discussed internally and developed (unconfirmed)

 

 

Hey, me to, fancy that. My first actually. Until now, I have dedicated my once-a-year conf stint to a little get together called VMworld. This year, however, I will break the mold and focus more on what made me whole, EMC product centricity. What possibly could EMC announce this year, is my question? Here’s what I speculate..

• Cisco buys EMC, monopoly would you say? Highly unlikely, purely from a regulatory stance with the new administration in office and all. Can you say unstoppable force of nature if it comes to fruition?
• More acquisition news? more partnerships formed? etc, etc..Anyone’s guess. Who is the next Kashya?
• Tucci and Maritz shake every conference goer’s hand presents them a voucher for a free NS-120 and a copy of ESX 4. Just their way of giving back to us customers/partners.

Anyways all kidding aside, I hope to see there. I will be the guy with kaki pants, a golf shirt, a book bag and a thingy around my neck…

Schedule as follows:

Monday

8:30 AM-9:30 AM
Symmetrix V-Max Series with Enginuity Overview

3:00 PM-4:00 PM
Test Drive EMC Avamar – Hands-on Workshop

4:30 PM-5:30 PM
Fibre Channel over Ethernet (FCoE), iSCSI, and the Converged Data Center

Tuesday

8:00 AM-9:00 AM
Virtual Datacenter OS Simplifies Computing

9:30 AM-10:30 AM
EMC PowerPath and PowerPath/VE: What's New in 2009

11:00 AM-12:00 PM
Celerra Performance Analysis and Troubleshooting - Hands-on Workshop

1:30 PM-2:30 PM
Managing the Virtualized Data Center

2:45 PM-3:45 PM
FCoE - Considerations for Deploying Fibre Channel Over Ethernet

4:15 PM-5:15 PM
New Storage Features in VMware Infrastructure

Wednesday (Daughters B-Day! Morning Sessions only for me)

9:30 AM-10:30 AM
CLARiiON: Using Navipshere Analyzer to Optimize Performance for Serial ATA, Fibre Channel and Enterprise Flash Drives

11:00 AM-12:00 PM
CLARiiON: Performance and Design Implications of Mixed Workloads