I put together this grid collage from the photos taken with my Blackberry Curve. The photography is not too good, but the memories are great!

Click the collage for a larger view.
Doug Hazelman of Veeam made a quick video of his trip to Orlando for PE09 and posted it on his The Virtual Traveller blog. At the end is some brief footage

of our Tweetup during the conference opening reception and then another tweetup during the conference party.

VMWare Partner Exchange 2009 Highlights

Here is a listing of Brian’s posts on PE09 so far, a few quotes, and some of my own comments.

• Partner Exchange Day 1: PTAB

“Overall, it was a great experience. It was a series of presentations that generated a lot of great discussions and introduced us to some of VMware’s future products and features. They even had a couple of presenters that would throw out ideas just to see if it was a feature we thought was worth developing. Meeting some of the elites within the partner community was definitely a highlight as well.”"Overall, it was a great experience. It was a series of presentations that generated a lot of great discussions and introduced us to some of VMware’s future products and features. They even had a couple of presenters that would throw out ideas just to see if it was a feature we thought was worth developing. Meeting some of the elites within the partner community was definitely a highlight as well.”

I echo Brian’s thoughts. PTAB was awesome. I am honored to have participated and extremely grateful to have been invited. 13 hours straight in a room that was cooler than most data centers sounds brutal, but the content, discussion, and value of the presentations kept me engaged for all 13 hours.

• Partner Exchange Day 2: The Conference Begins

“I now need to admit that it’s really hard practically starting my blog at a conference where most of what I hear is covered under NDA. What I will say is that, fortunately, most of it will be sharable soon.”

Patience Brian, patience.

• Partner Exchange Day 3: The vSphere Experience

“Yesterday’s keynote started with Carl Eschenbach, VMware’s VP for Global Sales, who presented a series of improvements to the Partner Program. Being a recent addition to the Partner community, I have felt some of the pains with the online access to Partner information, but have had little introduction to the program itself. I have to admit to being very impressed with the changes they are making to the program. Though most of the improvements will only be seen by the Partners, I think the agility and access to resources will be felt by the customers. Very cool stuff.”

The VMware Partner program changes and the tools available to help take advantage of the upcoming technology announcements are impressive. Considering next week’s event, I’d say all VMware partners now should know how to get everything they need.

Although I am at Partner Exchange 2009 (PE09) in Orlando, FL this week

In the meantime, here are some summary links and some quick thoughts about what has been discussed publicly so far at the conference.

Dave Lawrence has summarized the first 2 Keynote/General Sessions on his VMguy Blog:

• First day of keynotes at Partner Exchange
• Day 2 Keynotes at Partner Exchange

Seemingly related to the PE09 Keynotes but not confirmed (by VMware or myself),

• Mental Ward: vSphere Set for Debut
• Watch the vSphere Launch and Win a TrainSignal ESX Training Kit
• VMware vSphere is coming
• vSphere 4.0 may not support some other VMware products

Chad Sakac, VP VMware Technology Alliance at EMC, and Ed Bugnion, VP/CTO Server Access and Virtualization at Cisco, were on stage this morning to talk about their latest announcements, products, and partnership efforts with VMware. Of course, Chad and EMC also announced some big news this week in advance of VMware’s event next week.

“The new EMC® Virtual Matrix Architecture™ integrates industry-standard components with unique EMC Symmetrix® capabilities to deliver massive scalability—enabling systems that scale to hundreds of thousands of terabytes of storage and tens of millions of IOPS (input/output per second) supporting hundreds of thousands of VMware and other virtual machines in a single federated storage infrastructure. It is the first storage architecture that combines the performance and efficiency of a scale-up architecture and the cost-effective flexibility of a scale-out architecture. It was designed and built from the ground up to break the physical boundaries of data center storage, incorporates automation to simplify storage management, enables resources to be scaled on demand and uses less energy per terabyte of data stored than traditional high-end storage systems.”

Check out Chad’s blog post EMCs VMware Storage Strategy – The 3rd Shoe Drops for more on EMC’s V-Max.

What’s up with Chad using the PE09 General Session for trying to boost his Virtual Geek Blog standings?! He showed his Virtual Geek URL on a slide and he begged for attendee help breaking into the top 3 VMware blogs in his presentation!! All kidding aside, Chad’s blog is a must read for VMware admins and the new EMC V-Max and Cisco Nexus 1000v technologies are awesome advances for virtualization. V-Max really made me think when I learned it was possible to extend VI shared storage across multiple, separate physical data centers. Could geo-location VMotion finally be a reality?

On a personal note, the social networking has exponentially increased for me at PE09. I’ve really enjoyed meeting in person all the on line friends made through both VM /ETC and Twitter!

Check out my PE09 tweetgrid to virtually attend the conference the rest of the week – it works even if you don’t have a Twitter account yet. The conference party tonight should produce some entertaining tweets and twitpics.

I will tag all posts from this event as citrixsummit2008 for future reference.

The conference is being held at the Walt Disney World Swan and Dolphin Hotel. It’s a location in right in the middle of the Disney properties and parks. This morning after breakfast I was able to enjoy a great walk to the Disney Boardwalk, and then I followed a paved trail around the hotel’s central lake which took me past the neighboring Beach Club and Yacht Club Hotels as I worked my way back to the Dolphin building again.

I’m not sure of the cost of a family vacation at the Swan and Dolphin or one of the properties I walked past, but if you get the opportunity I would highly recommend staying here. I’ve seen a lot of families here so it must be reasonable. There are free water taxis to Epcot Center, and free buses to everywhere else. Click on the resort map image to the right for a larger view of the location.

As I am writing this I can hear fireworks, so I assume it must be Epcot’s nightly show. I’ll be sure to catch that tomorrow!

Tim Myers – Senior Consultant

This is my third 2 hr sesion today. I had to get coffee and refuel. I’ve sat in the same seat in the same room since 10:15 …. except for lunch.

Dan Anderson is the drill sargeant of VCB. If you’ve attended one of his sessions you know what I mean. He should keep me awake.

VMConverter 3.01 can now be used to restore from VCB images – that is huge!! You can restore to ESX, VMServer, workstation. This is the first “ton of bricks” concept that has hit me this week. Makes the restore process for full VM backups much more feasible.

Notes and questions from session:

• Separate physical proxy – W2K3 sp1 or R2
• network conn to Vc2
• FC HBA
• backup software ( on supported list)
• ESX 3.02 for VCB 1.03
• VMFS 3 on SAN (fibre) or iSCSI (it works, but not supported ??? according to Dan)
• Can not reside on same system as VC2 – vlun driver of VCB will screw up customized template deployment. It will work up until you try that it works … best for VCB to be it’s own server.

Advantages

• Reduced resource contention
• lan-free backup
• optional backup agents
• performance
  • traditionally backup agent uses 5-7% cpu and will increase exponentially with multiple vms being backed up in a VI. VCB eliminates this potential issue.
• timing – VC2 receives request and creates snapshot and allows VM to keep running in production mode and allows admin to backup whenever.
• cold backup – can be backed up while VM is down (must be registered in VC2 – must have been powered on at least once so it was registered in Vc2)

VCB File level backup notes:

• quiesce (sp?) script freezes application – snap id .vmdk created
• thraw script restarts application and moves frozen snap id to VCB proxy server – syncs and flushes
• mount snap id .vmdk in specific area on the VCB proxy server
• then use backup software to backup mounted .vmdk – backup software archive bit feature must be turned off to use – check compatibility list
• unmount .vmdk and gets rid of snap shots – cleans up.

Full VM Backup notes:

• export VM in COW format – copy on write – does not take whole VM but a collection of 2 GB files without white space to VCB Proxy server
• Backup software can not access until export is finished

Integration Modules

• Intention was to help kickstart partners to enable VCB use – baseline scripts only. Never intended to be “total solution”. Very little changes will be made to current scripts. Some vendors are way ahead of others:
  • Netbackup 6.5 requires “0″ scripting – go see Veritas booth this week
  • Commvault Galaxy
  • Vizioncore ESXRanger
  • CA Brightstor ArcServe

Not supported:

• If the VCB Proxy can not see the disk images it can not back up it – SAN zoning for ESX VMFS
• RDMs in physical compatibility mode – therefore MSCS clusters unsupported
• VMS on NAS/NFS not supported at this time

Backups on ESX Console OS – uses ESX resources

• only full VM backups
• backup any VM on any ESX on a datastore
• can not backup to/from non SAN storage (local storage / alternate datastores)

How Do I size & design VCB ???

It depends on several factors !! VCB is 5 – 10% of total backup solution

• VCB sizing - running any backup software such as Media Server, NetBackup, etc. on same server? How many expansion slots for multiple HBA’s for added throughput??
• VCB Bottleneck – usually I/O bound, not CPU bound
  • FC throughput to proxy
  • HBA speed
  • FC Switch speed
  • HBA qty – more hbas more speed
  • 3rd party backup software – # of i/o(s) they can push, parallell or serial jobs?

VCB Best Practices

• High I/O VMs are not good canidates for snap shots / VCB – if you wouldn’t snap shot a VM during the production day you probably should not use VCB to back it up at night.
• HBA setup
  • example: customer had 20 tb, 900+ vms wanted backed up every night in 6 hrs
    • backup server, vcb, and media server all in one physical server
    • one set of hbas for ntfs, another set of hbas for vmfs, tape arrays on third set of hbas
    • tested with 1.5 TB was able to move 140.8 MB ps and took 3.0 hrs – customer needed 40+ hrs
• consult backup software for sizing and tuning
• OS must be 2003 Sp1 or R2 on physical server
• disable auto driver letter assignment in windows – failure causes data corruption – disabled by default in enterprise edition only. if this happens STOP and open SEV 1 support with VM support.
• if backup software can , do multiple parallel jobs – not streaming one job, but multiple jobs.
  • low number of parallel jobs should be OK – 8 is rule of thumb max amount jobs in parallel.

Wish List : – not currently in development but should be!

• Plans to utilize MS VSS
• back up software understands VMFS

What is VLA? Questions and notes from session

Shared VM library

• stores vm configurations in state as a unit
• uses a fraction of the storage normally required – 16 mb files called differencing disks
• runs identical copies of vms simultaneously on the same network
• leverages central pool of virtualization resources

Does VLA need VC2? – not today, but next version will use VC2

Rapid provisioning

• deploy as needed
• web based

transient configured infrastructure

• software dev and test
• tech support & demos
• patch testing

Stage Manager is a new product that will take a dev machine in VLA to production

VLA uses separate ESX servers today then the production ESX servers. In the next release VLA will manage resource pools and integrate with production ESX servers.

VLA layout:

deveolpers use their workstations

VLA runs on it own MS windows server

Image Storage Library can be shared storage on FC, iSCSI, or NAS

VLA does not currently use DRS. They do distribute the creation of new VMs across ESX hosts but do not move VMs if the resource usage changes.

VLA does not use round robin. VLA queries the ESX hosts for available memory and cpu at time of deployment and intelligently places new VM differencing disks on hosts.

New feature – Storage Lease

• default lease is 7 days and is configurable via drop down.
• user notified when lease will expire and vm released back to library

You can RDP to fenced VMs – RDP to external IP assigned – NAT is autoconfigured via Lab Manager and a predefined pool of ips. NAT is running as a VM on the ESX server

VLA uses SOAP API and can be customized as needed

VLA Adoption Patterns

Basic

developer requests to provision server for testing on ESX

Intermediate

developer opens VLA and checkout environment

Advanced

completely automated dev test & qa process using VLA

VLA Features

Linked Clones – copy VM in seconds – whole vm in 16 mb initially

Saved State – never wait for vm to boot

Fencing – users can use benefits of linked clones and saved state simultaneously

• fencing uses virtual networking (switch) and bi-directional NAT-ing
• A VM router appliance is created for each fenced deployment – very small footprint and small memory usage

Replicate using EMC SRDF

manually change ip addresses of replicated machines?

Use F5, dynamic dns to self update or auto update ip addresses.

Host based replication

Replistor

Double take

Must have vm or host at both the primary and the secondary site for HBR. Must also pre-setup the networing, etc.

VMware is doing more with the SAN based relication

Site Recovery Manager will be able to interface with the replication product’s manager server 

How much bandwidth do I need?

How many hops? How’s the latency? How much data is changing? Synchronous vs Asynchronous?

Impact to RTO / RPO

RTO – time to recover 24 hrs

RPO – 1 hr with lower latency

Clone / or replicting data only

most cases are true cloning and working out re-ip of vms

SQL / Exchange

SQL2005 may lose certain transactions, but mechanism in application can roll back

Virtualization is not the solution, but each application must be handled according to the reccomended process fro each application

transactional databases will have to roll back to previous times, or restarting the VM will be like a crash recovery for the OS and the application

UUID changes on SAN

Site Recovery Manager will handle this

Doesn’t have to have different UUID?

break mirror, exposes luns, esx servers already know about the luns, rescan

Planned Integration with MS CCR?

Not in first release

Advantages to booting ESX from SAN

doesn’t gain you too much. Doesn’t hurt, but no real advantages

Multiple VC2 servers

Site recovery Manager can let you have multiple VC2 servers

Business Continuity Plan

strategic apps required for the business to survive

I.d areas of customer’s concerns

match core competencies of customer to develop solutions

Show ROI and TCO – typically high cost but value of solution will be worth it

Solutions to develop around

• HA
• DRS – resource pools to manage VMs so critical vms take priority in DR event
• VCB – resource loads on ESX host
• storage replication -
• OS/app clustering – make sure not set to suto start in HA failover
• fault tolerant (hardware & software) -

Site Recovery Manager - announced today

workflow for DR failover and testing- automated and manual components

call manager / email team / prompt to get Change Control approval

using fencing technology like Lab Manager.

Large services engagement possible with this product ??!! 6 – 12 weeks.

create recovery plans

Integrate with replication

Map recovery resources

specify recovery process – custom scripts, alerts, etc.

HA – ensure high priority machines have priority when HA goes into effect

VCB – Integration with 3rd party backup software may require extra steps

full vm and file level recovery

script cutomization may be required

Clustering

Kickstart for quickly deploying ESX hosts

RTFM site – UDA appliance – Ultimate Deployment Appliance - many customers are using this tool as the current automated method of fail over.