Some one ask my why I haven't blogged in so long today, they had missed the note I posted that I was now blogging on the the Virtualization Team Blog. I just did a post this morning in fact with some details on a very cool new product we have RCU 2.0. Be sure to check it out. I al have been busy with a Ask the Experts VDI event. Be sure to check that out to as it just started today and already we have some very good questions and hopefully what you think are some great answers. Be sure to sign up and ask us anything you have relating to VDI.

Keith

Most people I talk to have never heard about our Rapid Cloning Utility for VMware (RCU). This tool however is the magic behind the rather famous youtube videos that show how NetApp can very quickly create and deploy a very large number of VMs while using very little disk space on the array. With VDI being VERY hot right now (I have 5 projects on the go!) this is a great tool to get to know and use. The tool is now available on the NOW tool chest and is free. I had an opportunity to really test this tool out and thought I would share a few things and also look a little beyond the initial deployment to how would you perform routine tasks using this tool.

The RCU set up very fast and was incredibly easy to use. Hats off to the team that built this.  There is a little fine print that you need to read and be sure you do before hand prior to using the script. Like most IT guys, I rushed in and then had to start over properly to make sure everything was right. First, in your golden volume, you need a batch of VMDK files in a folder. Essentially build up your master image that you want cloned. If your are like me and using a Windows XP image, be sure to install the LSI SCSI drivers in the VM BEFORE cloning it. You need those drivers after the mass deployment. Once you have the image just as you want it, clone the VMDK as many times as you want VMs in a volume. For me, I was testing a block level install so I did 25 VMDKs in the volume. I used another tool for that which was the rfwm tool which also worked like a charm. Another tip here though, you want ONLY VMDKs in that folder. I deployed a VM to the folder and then cloned the VMDK files giving me 25 VMDKs and the other files associated with the original VM. This gives you some phantom VMs when you use the RCU (I think it tries to build VMs and attach them to vmx files instead of VMDK files. Easily corrected tough.  Once this was done, I deduplicated the volume, took a snapshot of it and started cloning.

I had many, many VMs within minutes and was very impressed how the VMs were properly named, organized and customized. Awesome! So what now? How would you use this tool going forward?

More VM’s Please

Dropped my command line into a simple .bat script sitting on my desktop. This way everything I needed another 25 VMs I could just double click the icon and poof, there is another 25 VMs. Very easy.

New Folder

I have to ask the developer but I learned that the VMs are created in whatever folder the VM Template lives in. Good to know. Since the VM templates have no disk file they are very small so it would not hurt to create multiples of them. Perhaps you have VMs that share that same disk image but you need some VMs with more memory than others Very easy to create multiple templates and then just choose which template to use for the cloning. All the clones will share the templates memory, cpu and resource settings.

Patching

Here is one I get questions about all the time. How do I patch? Well you have 3 options here.

1. You can simply patch all the existing VMs as you would physical desktops. You need enough space to hold all the patches but you would then run dedupe against each volume and reclaim that space back. For your base image you will either have to patch all the cloned VMDK files or destroy them, patch the original and reclone, the second option being the easiest. This is the preferred method for customers having persistent desktops.

2. If you have non persistent desktops, that is, users are not assigned one particular desktop, you can patch the original VM, create more clones from the patched image and then start to migrate users from the old image to the new. Most connection brokers will allow you to stop connecting users to one folder and start connecting them to another. By doing this users will simply get a newly patched VM the next time they log in. Once everyone has disconnected off a folder of the old VMs. You can unregister the VMs and delete the volume off the NetApp controller.

3. The 3^rd option is to simply destroy all the clones during an outage and reclone them back out. The tool runs so fast that is a very viable option if your VDI farm isn’t used 24X7

I know the team is hard at work on version 2 of the tool so I can’t wait to see that! Let me know if you have questions or want to know how you would perform a particular task using the RCU.

Keith

I assume that most people reading this are either NetApp customers or Partners. As such you have access to the NetApp simulator. This is an awesome tool for practicing your NetApp skills on and for putting together portable demos. Case in point is Tomas’ “How to Guide” for setting up VMware SRM using a pair of NetApp simulators, available here http://tendam.wordpress.com/2008/11/18/srm-in-a-box-final-release-the-complete-setup/  I recently set this up myself and it really works great. The newest Simulator runs 7.3 and now even has Dedupe allowing you to place a couple of VMs in it and see for yourself the savings potential of Deduplication.

One limitation of the Simulator was the fact that you can only place 28 “virtual” disks into it and the biggest disk you can use is 1GB. This gives you enough space to play but for something like SRM it makes things a little tight. I did discover though that if edit the Max disk setting in the setup.sh file in the install directory then you can add additional disks. This is handy as it buys you some extra space to have a couple VMs loaded up on SRM for testing.

Some words of caution though. I can’t say how well or how long this will work. I just tried it and found that it worked. I can’t say if anyone “official” has ever done this. I don’t know how many disks you can actually add, I only needed another 10 so that I could fit another VM on. I would bet an additional 28 would work but no more (that way you could cluster 2 simulators each with 28 disks and they could fail over, which means going beyond 28 on one would prevent your cluster from working) Also a reminder that the simulator is for learning and demos. Not to run anything off of. It is not supported so please don’t put it into production. Happy testing!

Happy New Year! I hope everyone had some fantastic time off. Can you believe that we are in 2009 already! 2009 is certainly going to be a very interesting year. I know there are all sorts of new years  predictions out there but here are my guesses for the virtualization industry;

1.     Hypervisor battle will heat up – I think VMware will continue to lead the market but I suspect the economy will force some customers to look at other less expensive options. Ultimately competition and choice is a good thing so this will certainly make things interesting.

2.     Data Center OS will begin to take shape – VMware’s next version of hypervisor includes APIs to manage several aspects of the data center, including storage. This increased range of management will allow VMware administrators to manage a greater number of items in the data center.

3.     Deduplication will become part of most virtual designs – Deduplication has had such a huge effect of virtualization costs that it will become part of most designs. It may however be in different forms, either storage block level as in the case with NetApp, hypervisor level or 3rd party. Which form is selected will depend on the customer environment but I think NetApps block level storage level will be hard to beat since it is transparent to the hypervisor and can enhance performance through intelligent caching.

4.     Virtualization Optimized Storage- As Virtualization continues to become mainstream I expect more features to be built into storage to enhance virtualization. These will include deduplication, enhanced and tuned intelligent cache and provisioning which can be controlled via the hypervisor.

5.     An increase in the customers who have gone “All Virtual”. I think that more and larger customers will completely adopt a virtual strategy. This challenge virtualization architects to design environments for those traditionally difficult to virtualize workloads.

6.     NFS will become an even more popular choice for virtualization storage connection. NFS already is very popular but with the cost of 10G Ethernet coming down and more customers sharing their experience with NFS I think it will challenge FC as the most popular choice. FCOE also may shake things up late in the year.

Ok, I know these are pretty safe predictions but I still think 2009 is going to be a very interesting year in virtualization and I NetApp is ideally positioned for the year ahead. Once again…Happy New Year!

Keith

When I ask customers about their pain points in a virtualized environment, I consistently get the same answer. Backup. I won’t spend a lot of time on the issues as I expect you have already read about them, seen a presentation or are living the pain yourself. In a nutshell, traditional file level agents which have been the standard for years just don’t cut it in a virtual world. They place tons of load on the virtual layer, they tend to only back up data (not the OS which can be very difficult to recover) and are slow to restore.

The fact that virtualization encapsulates the server into a small number of files should make this easier. Unfortunately when you move to the VM layer you often end up dealing with way too much data or lose your granularity of restores. This tends to leave customers stunned about what to do. Most I think default to traditional agents since they are then no worse of than before.

The first step in designing a backup solution in a virtual environment is honestly think about how you want to restore data, not how you restore it today. For example I recently heard a backup software vendor state that 90% of the restores they see are file level based.  I wondered though is that because that is the only option most people have which is why the number is so high. If every customer could restore they machine at any level what would the preferred choice be?

In one server environment where I designed an image level based backup solution to compliment the file level backup the numbers are quite different. The image level backup was based on NetApp snapshots and we would take them hourly throughout the day.  In the following months after the setup we got the support team to think about the restore and the support requests that came in. Restore requests could often be satisfied in either method but which we found interesting was that many of the support requests could be satisfied with a simple image restore. Rather than spend time trying to figure out why a particular app server went down and how to fix it, they often could default to restoring the complete VM it to when it worked. Obviously this would not be the case for any type of data loss but is a very powerful tool to have at your disposal.

Ultimately I expect to see converged server backups. That is, backup the server once (my guess is that will leverage storage based snapshots but that is just my guess), from that one backup though I want to have the ability to Archive the server as required, replicate for DR as required and provide image level, file level and object level restores. Don’t duplicate the backup duties and only save the data once not multiple times.  I know this sounds pretty far out there but I don’t think it truly is and I like NetApp’s chances of being one of the first to offer this. 

Hello world! My name is Keith Aasen and I am a Virtualization Specialist with NetApp covering Canada. I know there are plenty of blogs about virtualization out there these days so what am I going to write about that hasn't already been written? Good question. Since I am based in Canada the customer set I tend to work with is Small and Medium businesses, heck even our large companies are often considered mid sized as compared to the American market,  I think the customers I work with often look to virtualization for slightly different reasons. Most don't have sprawling data centers with massive power problems. Instead they have a smaller number of servers trying to service a massive geographic region.

Of course they have common challenges, customers I speak to are trying to do more with less, grow in a more flexible manner and the environment is becoming more of an issue all the time.

Recently I was asked to present at a VMware user group about VMware's Site Recovery Manager. I have been involved with a couple of SRM project and have it running in my lab between a pair of NetApp 2020 controllers. I often make the joke "If one side of my basement floods..."  I started the session by asking who here currently has a DR plan. A couple folks raised their hands. I then followed by asking who here is currently looking at building a DR plan or has been asked my management to investigate DR strategies. Not unexpectedly a large portion of the room raised their hands. I am excited for this, DR is truly an area that virtualization is a game changer and I think every sized company can design a DR plan that will fit their environment and budget. The smallest DR plan I ever built was a customer who had 6 servers (in total!)  and replicated 4 of them to a field office on a nightly basis. There was no way that that would have happened for that customer if not for virtualization.

After my SRM demo we discussed what the topic should be for the next meeting. Overwhelmingly Backup of a VMware environment was the popular choice. At first this struck me as odd. Here was a group that just told me they want a DR plan and they haven't yet figured out backup?

The more I thought about this though, the more it made sense and actually it was a good thing. Backup of a VMware environment has historically been one of the thorniest bits of virtualization and is honestly still a developing practice. I believe however that it will logically merge with DR so that they will no longer be 2 different solutions but instead just be a backup and DR solution. Therefore, having the opportunity to design the 2 solutions in parallel is actually a very good thing. Yo can them maximize the efficiently of the solution and have the 2 components compliment each other rather than duplicate each others work.   I plan to talk more about backup in my next post so perhaps I will leave it at that.

I am presenting a keynote presentation at a VMware Forum in Vancouver next week, December 10. My presentation is called "A Day in the Life" and will be about all the crazy things that are asked of the typical VMware/NetApp Admin. These things include, being asked to deploy dozens of servers in a silly short period of time, having business analysts forget to budget for storage in a project, having to backup TB of VMs with no outage (and bring them back!).  All things that 4 or 5 years ago would have been impossible. I guess the average VMware admin seems to have super powers in the eyes of most businesses.

If you are a VMware Admin and have had a crazy request made of you please let me know. I would love to have more examples. Also if you are in Vancouver on the 10th, stop in and say "Hi!"

Keith