I'm James Burd and you've just been made virtually wiser

I passed the test on 10/29/2011. Like everyone else has said, this test is nothing like VCP 3 and VCP 4. Those tests seemed to focus more on what a product was and the features. The VCP 5 exam seemed to focus more on features of vSphere 5 and how things work. Everything that is a part of the optional install for vCenter is fair game. This includes update manager, Auto Deploy, Web client, Dump collector, Syslog collector and Authentication Proxy. There were also some tough questions on the Virtual Storage Appliance (VSA) and the vCenter Appliance (VCSA). My exam seemed to have about 40% scenario questions with exhibits to go over, these can be multiple choice or multiple answer. Don't count on this though, your exam could be different. VMware certainly raised the difficulty level with this exam. I would expect to see this level of skill on the VCAP exam or somewhere in the middle. For me, it was a little difficult to study for the exam with all of the other major (non virtualization) projects I have going on. You really have to take time and focus on the material.

 - I like to call this next section of my post my "rant" section -

There seemed to be some question asked that had to do with some very small topics, or should I say feature sets that I have not seen any enterprise environment use (I'm not going to say specifics because I don't want to get close to breaking the NDA). If I found a company utilizing the vSphere client in this way, I would recommend against it. To me, the vSphere client is not meant to be a general access method for using VMs. But the feature is there and it is a testable subject. Just because something may be a bad design doesn't mean it will not be on the exam.

There was also a scenario question on a topic that mentioned X amount of hosts in a cluster, but the exhibit did not have the same amount. I am going to have to contact someone on this issue. 

I also found a handful of questions that seemed to be poorly written. Maybe it was because I got in a hurry halfway through, but the manner in which some questions were asked did not make sense. I just wanted to shout like Samuel Jackson "what the hell is this?"!

- The testing center -

I took my exam at the eConsortium location in Houston. In the past, they had an excellent testing room that was private in an almost cubical area. You didn't have distractions from other people and you couldn't hear other people. Now, the exam area is in a classroom area. They have multiple tables set out where everyone sits right next to one another. This is a big distraction, especially when you have someone sitting right in front of you who is facing you! In the future, I'm going to have to find another Vue testing center to attend.

- The only advice I can give is -

Take your time and breath, but be mindful of the time restrictions. If you are unsure of something, just go back to it later. But try to set a limit on how many questions you are willing to go back to later. Remember, the answers to the questions will be right in front of you. Just eliminate the ones that you know don't make sense. The answer is sometimes in the question, make sure you understand the small details or hints that the question is giving. For example, know the difference in a product update and an upgrade. 

- What to study -

The VMware exam blueprint is right on the money when it comes to topics. But no one can memorize all that stuff. I don't think this exam is for anyone who may come right out of a VMware class or who has only read all the PDF's. A working knowledge of all the core vSphere products is needed. You really need a lab to work on. If you can find one that has CNA cards or FCOE adapters it would be better. Knowing a wide variety of errors also helps. This test will exercise your troubleshooting skills.

I used a few test engines to help study for some of the vShpere 5 products. Remember, these engines do not contain actual questions or answers from the VCP 5 exam, they are meant to get you thinking about a topic.

Simon Long's practice exam.

Robert Stothard's pratice exam.

Elastic Sky practice exam.

VMware mock exam.

Some exam blogs also helped me to create some flash cards to carry around and study with:

Cosonok's IT Blog

vSphere documentation.

A VCP 5 site.

Andrea Mauro's VCP site.

One type of practice exam I would like to see is one that is scenario based and has exhibits. I think I'm going to look into creating one.

VMware provided discount codes at VMworld 2011 for the new VCP 5 exam. This 50% discount code "VMWORLDLV50" may end once VMworld 2011 ends. There is also a 25% discount towards to the VCAP exam, which is "VMWORLDLV25". All exams must be taken before Nov 30th, 2011.

Sign up for the exam at Vue.com.

F5 has a great product that can be used in place of the Citrix NetScaler, the BIG-IP LTM. Although, you should watch the webcast from Citrix on how the NetScaler outperforms the F5 LTM. Citrix claims that in most test scenarios they have 3X less latency that the F5 LTM. 

The deployment guide document from F5 for XenDesktop needs a little clarification for first timers. On page 7, it seems like a simple process of copying a "create-citrix-monitor.tcl" file from your local windows station to the F5 LTM via WinSCP. The TCL file can be downloaded from here. But, what you need is a "create-citrix-monitor.sh" file to run from the command line on your F5 device. The link provided in the documentation contains a script that does not work. You may receive a line error "Failed to load TMSH script". I received a replacement script from the F5 support team that can be found here. This was run on a F5 LTM with 10.2.1 Build 297.0 Final.

Take note that this script calls against a "wpnbr.dll" which exsists on the DDC, not the web front ends. Another option on the CitrixWICredentials is to create a simple send and receive string to check against web services on the web front ends.

Send string: GET / HTTP/1.1\r\nConnection: close\r\nHost: CompanyVirtualName.company.com\r\n\r\n

Receive string: /Citrix/DesktopWeb

Please test this in a test environment first.

 If you were like me, you had to do a double take on the new vSphere 5 licensing model. Sure, there are a ton of new and great features in the latest version of vSphere 5, but it is really going to be hard to polish this vRAM turd of a license model that claims "it gives customers the greatest flexibility for vSphere configuration and usage". If anything, it has caused more confusion. The blogosphere has been on fire with posts from many blogs (including vExperts) and the VMware communities detailing this latest license model from VMware. I have no doubt that this will be a hot topic that everyone at VMworld 2011 will be grumbling about. I really can't see anyone saying "hooray! I'm so glad VMware came out with a license model that gives customers greater flexibility for vSphere configurations and usage, it's so much more cost effective than the old model". I'm sure like many of the blogs, everyone will be saying "dang, I'm gonna have to start deploying Hyper-v and some Xen".

This really makes me want to stop wearing my VMware shirts around the datacenter. I don't want to get questioned about this licensing model by some angry customer.

So what is this vRAM entitlement? It's essentially the RAM that is assigned to your virtual machines. Here is an example:

I have two dual core Dell R710's with 96GB of RAM and VMware enterprise licensing.

- Under vSphere 4.x licensing I would only have to buy 4 licenses based on the physical sockets. Wow, that was easy!

 - Under vSphere 5.x licensing, the limit on the physical CPU socket is gone and replaced with vRAM (kinda). Under the Enterprise licenses that I own, I'm entitled to 32GB of RAM per physical socket on my servers. This means I'm entitled to 128GB of RAM for my "powered on" virtual machines. So this "pool" of vRAM entitlement can be used by my virtual machines. But, the physical RAM on my hosts equals 192GB. As long as the vRAM assigned to my virtual machines stays under 128GB, I do not need to buy any additional licenses. But if I want to take advantage of that extra 64GB of RAM, I will need to upgrade to Enterprise Plus, which entitles me to 192GB of vRAM (48GB per proc). It is not clear yet if I would be able to just buy additional Enterprise licenses to to make up for that missing 64GB of RAM. My guess is no since the entitlement is by physical socket on the server. Remember, you still need to account for hypervisor overhead with your physical RAM (which does not count against your vRAM entitlement).

So there it is in a nut shell. Companies with heavy amounts of RAM in host servers are going to be the real ones that suffer in this scenario. All those companies that virtualized those Tier 1 RAM hungry apps are put in a really difficult position here. How do you now cost justify moving those to VMware? Or is this just a way to force those apps out to cloud providers? 

There are many different option to setup Lync 2010 in a VMware environment. This guide will not cover specific deployment or migration options for your install, but mainly the layout of roles that exist today. I will not be covering any SQL server specifics either. See the best practice guide from VMware on virtualizing SQL. There are many different guides on deploying Lync 2010 in a virtualized environment and I will give you many of references to all the major ones I have found. I developed this guide while I was migrating an existing non-voice OCS 2007 for 3,000 users. Migrating environments is pretty easy in Lync 2010, especially if you know the work load.

I have to say, one search term I see most often is "how to virtualize Lync 2010 voice". The Mediation server is a supported role in VMware. Pay attention to the sizing guide for your deployment and scale for the number of concurrent calls you will have. This role should be given the same attention as all the other roles in the deployment.

First we will cover the server side virtualization and then we will get into the client side options.

So let's take a look at the highlights of the server side.

   – Virtualizing Lync supports roughly half the amount of load that the physcial counterparts do. Why? Well, this was based on virtualizing Lync 2010 in Hyper V and was due to the limitation of 4 CPU's per VM in Hyper V. VMware vSphere 4.1 supports up to 8 CPU's per VM (rumored to be 32 CPU's per VM in vSphere 5). So does that mean that Microsoft will acknowledge a greater workload capacity for Lync 2010 in vSphere? There is no solid documentation on this. And why should they document it? I'm sure they do not want to promote another hypervisors features. At first glance, I would say that you should be able to achieve the same work loads that the physical counterparts do since Microsoft has identified CPU as the limiting factor for work loads. Check out page 25 of the Microsoft white paper "Server Virtualization in Microsoft Lync Server 2010". Also pay attention to the sizing guidelines. I have not found any documentation from Microsoft so far that states they will not support Lync 2010 VM's in VMware with 8CPU's.

   – More about the CPU. Some documentation calls for a dedicated processor. But the consensus seems to lean towards just having a higher priority set in VMware for the CPU cycles on Lync 2010 servers. Because of this factor, I decided to place my enterprise deployment in a dedicated cluster. If you mix the Lync servers with to many non-Lync production servers, you may start seeing CPU ready values increase for other production servers since Lync servers are given a priority.

   – Networking considerations. I have seen some books that call for NICs to be presented to VM's via VMdirect path and other documentation that stats the Lync 2010 servers can share a 1GB or 10GB connection from the host server. I think the best scenario for this would be to use a 10GB connection for your Lync servers on each host. It is a safer bet over using 1GB because the A/V conferencing servers can use 500MB or more in peak times. You can also assigned multiple 1GB nics on a single VSS or DvS. If you plan on mixing your Lync servers with other production systems, consider creating a seperate VSS or DvS with dedicated NICS. This is also the recommended solution from Microsoft. Remember, there is no method to assign traffic priority to VM's (as of vSphere 4.1) like there is with CPU, disk and RAM. Also, do not share your host management network with the same network your virtual machine traffic is on. This is a standard practice.

   There is also a lot of recommendations around using the "Virtual Machine Queue" (VMQ) feature on the physical NIC. This is something that is enabled from the Windows Server (HyperV) side. Enabling the VMQ filter offloading increases receive and transmit performance, as the adapter hardware is able to perform these tasks faster than the OS. Offloading also frees up CPU resources. So what options do you have for this on the vSphere side? Check with your NIC vendor to see if the drivers loaded on your host have a similar feature for vSphere.

   – No vMotion allowed on an active Lync server. You can however migrate the workload off of an active server and move the VM to any other host server. For this, you will need to create some DRS groups and rules to keep the VM's on separate hosts.

   – Storage. Don't assume that just because your Lync servers are connected to your high performance SAN that there is not any I/O going on that you don't have to worry about. Always evaluate your SAN I/O. There is no specific recommendations for storage on any Lync roles. Always use best practice when laying out any virtual storage. Pay attention to the performance of your I/O.

    – Monitor your infrastructure with tools like HP SiteScope, Quest vFogLight, Quest Message Stats Lync plugin, Nagios and of course Microsoft SCOM. Myself, I use Quest Message stats, this provides great visibility into the usage of Lync and has great reports for your messaging infrastructure. I also use and highly recommend Quest vFoglight to monitor the virtual infrastructure. Microsoft SCOM also provides another layer or monitoring for all of the Microsoft products.

I would have liked to have called this post a "best practice guide", but there are to many variables that are involved with a Lync 2010 deployment. This post is mainly a guide on how Lync 2010 VM's should be placed and configured in vSphere.

   – All servers in the pool must be either physical or VM. The only exception is the back end SQL server, which can be physical. You can however have one pool of a virtual servers and another pool of all physical servers. Just as long as all of the servers in the same pool are either physical or virtual.

Client side virtualization

There is an article published from Microsoft "client virtualization in Microsoft Lync 2010" that states the supported platforms to use the Lync 2010 client with. Currently this document only mentions Microsoft and Citrix as the supported platforms for the client side app. No love for thinapp just yet! This is mainly due to the audio/video drivers required in the client.

Check out the Dell documentation on page 7 of the Lync 2010 solution architecture white paper. It has a guide to virtualizing 1,000 / 5,000 and 10,000 users. Kudos to Dell for publishing this doc.

Microsoft Lync Server 2010, Planning Tool

Lync 2010 Visio Stencils

Lync 2010 Protocol Workloads Poster

I hope that this guide helps in your future deployments.