Let’s call these new generation of projects V4DR (virtualization for disaster recovery) or V4BC (virtualization for business continuity).

The comparison

If I rewind 3 to 5 years ago in my career, capacity planning for server consolidation was a weekly project and topic of discussion. Customers were either in the process of converting physical servers to virtual machines or they were exploring the possibility to do so. In both cases, capacity planning scenario spreadsheets and reports were frequent “ground zero documents” to almost every project plan I was involved in.

Just like P2V projects, VM replication today also requires some of the same considerations for job scalability and times to complete – i.e. using multiple hosts as targets and making sure the network can support getting the job done as quick as possible. Not to mention ip addressing, VLAN assignments, and application connectivity after the fact. Thank goodness we no longer have to deal with hardware drivers and other unneeded software a second time. Hopefully, VM alignment is a thing of the past too!

I’m not seeing the same “ground zero documents” for replication projects, however.

Use the same capacity planning tools?

So, I’ll ask the question: 

are the existing capacity planning tools we have used previously for our P2V migrations good enough to help today with the V2V replications needed for our new V4DR and V4BC?

In theory they should be: monitor the virtual machines for several weeks and then determine the capacity and resources needed at the DR site. Leverage the capacity planning scenarios in these tools for reports like “What If I lose a Host?” or “What If I add X number of VMs?”.

Replication Capacity Planning

My point is, I think most look at today’s capacity planning tools as either no longer useful since the migration is over, or useful just for primary site capacity purposes. Looking at the output of these tools a little differently could also provide secondary site planning. A report like “What if I lose a host?” could also be viewed as “Can I consolidate my VMs on even less hosts for DR?”. “What if I add X number of VMs?” could also be viewed as “How much capacity do I need to purchase at my DR site?”.

Is what we have today good enough to be reused , or do we need a whole new version of capacity planning and virtualization assessment tools? Call the new tools Cloud / Hosting / DR/ BC / Replication Capacity Planners? Take your pick!

I’m curious. What tools that exist today do you think are already providing output that is useful for V4DR or V4BC planning? Leave a comment!

disclaimer: I work for Veeam Software. Veeam Backup and Replication is a VM replication solution. Veeam Reporter has Capacity Planning reporting capabilities.

” .. VKernel is now offering FREE perpetual licenses of our Capacity Modeler software until 12/31/09. VKernel Capacity Modeler is an alternative solution to the high-priced VMware vCenter CapacityIQ.

Download VKernel Capacity Modeler now and get the same function for FREE

Recognizing that VMware CapacityIQ’s capacity planning and ‘what if’ functionality is roughly equivalent to VKernel Modeler, Vkernel is offering anyone that downloads Modeler between now and December 31, 2009 free perpetual licensing for unlimited CPU sockets.

Download the FREE VKernel Capacity Modeler and get the ability to quickly plan, simulate, and validate hundreds of real–world “what if” VMware resource allocation and VM deployment scenarios. With VKernel Capacity Modeler, user can perform:

• Capacity Planning
• Interactive “What If” Modeling Scenarios
• Proactive Problem Prevention
• Predictive Capacity Impact Analysis

If you want to quickly solve more serious VMware issues, download FREE trials of our Capacity Analyzer and Optimization Pack software.”

I haven’t had the chance to use either product, but I thought VM /ETC readers could benefit from this offer.

Note the “roughly equivalent” statement above. I cannot speak to a comparison of the two products, but I still feel that any capacity analysis and modeling tool for free is worth investigating. If any readers have any experience with this solution please comment for the benefit of others.

Anyone who has already “harvested” their own “low hanging fruit” knows there is so much to consider. The p2v tool and process are the tip of the iceberg, change control is just below the surface, and there are many more technical challenges hidden in the depths. I’ve blogged in the past about treating the migration to VI the same as you would changing physical data centers. It’s not just server builds and operating system installs.

These same challenges experienced during the initial consolidation are still there for the rest of the bunch, but most likely on a much more public and political scale. In fact, since more times than not these same servers were left out of the first consolidation scenario as “bad virtualization candidates, it’s likely time to

To help visualize what is really needed, Rick Vanover created a great flow chart that, although it illustrates a complex process, may be the simplest starting point for understanding how to get the rest of the applications/servers virtualized. Vanover’s Virtualization Review post titled Advanced Virtual Conversion Flowchart provides for download a Visio diagram he presented recently at the 2009 TechMentor Conference. Vanover says it best:

“This is more detail than you would need for the typical conversion, but is helpful when you have difficult systems that you haven’t converted for any reason yet”

What impressed me is that the flowchart takes you through a logical process to consider all factors and their impact. For example, storage and networking each have their own decision points. Using the flowchart makes the administrator realize that just because the consolidation tool says there is room in the current VI and the hosts have enough RAM, CPU, network ports, and disk doesn’t mean the p2v will be a snap and the planning is over.

Here is a screen shot of one of the tabs in the .vsd.

Read Vanover’s post and get a copy of the Visio. On a server by server basis, combine your favorite capacity planning tool’s results with the flowchart’s process before picking the rest of the “fruit in the canopy”.

Read the entire article for several great points, but here is one quote as an example argument:

“The first thing to determine is whether installing virtualization on older hardware actually saves money. Today’s servers are equipped with quad- and six-core CPUs that can deliver much greater consolidation ratios that single- or dual-core models, he said. “VI3 [VMware Infrastructure 3] is licensed per socket, so you’re paying the same cost for older CPUs but you’re getting less bang for your buck.”

But if the budget for new hardware simply isn’t there, consider the age of a server and how much life it has left in it. When identifying a candidate for a virtualization host, identify a box that is three years or younger and has “had a relatively low runtime,” the systems administrator said.”

Why? Because I am not aware of another virtualization capacity planning tool that can monitor a company’s existing desktops, physical and virtual servers and then provide a comparison of consolidation scenarios across Citrix XenServer, Microsoft Hyper-V, Parallels Virtuozzo Containers, Virtual Iron, and VMware ESX.

Lanamark now appears to fill a current capacity planning gap for the following scenarios:

• Companies looking at implementing / migrating to virtualization hosts other than VMware ESX or Microsoft Hyper-V (each with their own capacity planning and TCO analysis tools).
• Companies undecided on which virtualization platform to use and want to explore scenarios with an independent partner of multiple vendors.
• Companies needing customized capacity planning for virtual desktop infrastructure (beyond  adapting server virtualization capacity planning tools to vdi scenarios).

The remainder of this post contains details about Lanamark’s benefits both for server capacity planning first and then desktops.

More info from the Lanamark February 17 press release:

“The Server Virtualization Design Module offers advanced optimization capabilities for server configuration and workload placement, determines the number of virtualization software licenses required and facilitates planning of physical-to-virtual (P2V) and virtual-to-virtual (V2V) conversions. It allows what-if scenarios to be compared side-by-side with comprehensive TCO calculations and enables channel partners to build end-to-end solutions with server, storage, software and service components. The module supports servers, blade enclosures, storage arrays and HBAs from Brocade, Dell, EMC, Emulex, Hitachi, HP, IBM, NetApp, QLogic, Sun Microsystems and Xiotech.”

Lanamark’s web site previews high level details of what the Server Module provides in it’s consolidation comparisons:

• Design end-to-end server virtualization solutions across platforms from Citrix, Microsft, Parallels, Virtual Iron and VMware.
• Determine the number of virtualization software licenses required for each platform.
• Explore what-if scenarios for servers, storage arrays and HBAs from leading vendors.
• Compare scenarios side-by-side using TCO, carbon footprint and power metrics.
• Share and reuse solutions, scenarios and reports across projects.
• Optimize server configuration and workload placement.

It seems very possible that Lanamark’s capacity planning and TCO results could be the cause of a future (heated – aren’t they always!) debate between hypervisor vendors. If my interpretation of what Lanamark can do is correct, then I will be surprised if sooner or later someone doesn’t use it in a public competitive comparison!

Lanamark Desktop Capacity Planning

More information from today’s press release:

“The Desktop Analysis Pack distinguishes desktops from laptops to identify mobile users and determines associations between workstations and external devices such as monitors, printers and scanners. It leverages the Lanamark Software Intelligence (LSI) technology to recognize software applications installed versus ones actually used across the enterprise. The combined information about desktop hardware and software enables channel partners to identify desktops most suitable for virtualization, determine what type of thin client devices would be most appropriate for end users and recommend how software applications should be allocated to minimize licensing costs.”

Mark Angelo, Lanamark CEO, further described VDI specific capacity planning benefits to me in an email:

The Desktop Analysis Pack is designed to accelerate desktop virtualization projects by providing comprehensive analytics around desktops and laptops:

• Identify candidate desktops for virtualization
• Distinguish laptops from desktops to identify mobile users
• Determine associations between workstations and external devices such as monitors, printers and scanners
• Recognize software applications installed versus ones actually used to determine which applications should be made available to end users and reduce software licensing costs and lower the risk of non-compliance.

Also, the Desktop Analysis Pack can be used in conjunction with the Server Virtualization Design Module (SVDM) to design virtual desktop infrastructure solutions across leading virtualization and server hardware platforms. We have also added thin client devices from leading vendors such as Wyse to our repository so that they can be included in the end-to-end solutions designed with the SVDM.

Check out Lanamark.com for more about professional services offered to assist companies planning migration and consolidation to virtual infrastructure.

First things first. Yes, now VMware is providing free licenses for Capacity Planner to qualified Partners. Virtualization.info reported this announcement back in June, and as Alessandro points out.

“Partners will not be required to buy any license anymore (even if they will still have to attend a classroom course that doesn’t come cheap). They will just have to login on the online portal, create a new profile and start monitoring the customer’s infrastructure. The data will stay online for six months and then will be archived.”

Depending on the relationship between VMware, the partner, and the customer, free licenses were not uncommon before either. Either way, I want to emphasize there is an initial investment by the Partner both financially and intellectually. To get the free licenses Partners must join the Partner program, certify VCPs, and attend the Capacity Planner classes and the Virtualization Assessment boot camp.

What does this mean to VMware’s customers new and old?

The rest of this post lists some common Capacity Planner service engagements for VMware customers to consider, and I will briefly explain the differences. VMware provides data sheets for some of these projects on their Services page, but Partners are free to either deliver the packaged offering developed by VMware or construct their own.

Special note to Sales Reps and VMware customers: When considering Capacity Planner projects, be sure to understand what level of assistance is needed. There is a difference between the deliverables of these offerings, and, yes, there is cost associated with the services of the more involved analysis. I feel the confusion about assessments stem in part from the fact that all of offerings use Capacity Planner. That is, the data collector is required to be installed and configured and the best practices and methodology for using the tool (previously covered in my FAQ post) is the same. It is the objectives of the different projects that vary considerably and thus require different levels of professional services. Complete capacity analysis tends to get lost in the licensing and hardware quote’s shadow. Unfortunately, this important function is too often viewed as either in the way of the sales process or unrealistically expected to take place as a pre sales validation. That may be a fact of business, but taking the time to complete some level of a virtual assessment will make the difference in smooth migration to VI, the performance of the VMs and the ESX hosts, and achieving and maintaining the expected capacity.

Consolidation Estimates (CE / BCE)

The pre sales CE, formerly known as the BCE, is a free offering to help determine the number of VMware licenses (VI3.5 or ESX) you will need and the server hardware to run it on . Basic consolidation scenarios are automatically calculated and are delivered to the customer in a generic presentation.

CE deliverables are usually

• A VMware software quote
• A server hardware quote
• A proposal for future VMware professional services
• A quick, high level Power Point presentation about the stats collected by
  Capacity Planner.

Customer’s who are best candidates for the CE pre sales study already have VMware ESX experience and do not need planning, design, or implementation help, or they are customers in the budgeting phase who need numbers for planning. The key aspect of this project to understand is that there is not any architectural design or planning services delivered.

Capacity Analysis (CA) and / or Virtualization Assessment (VA)

It’s debatable whether to group these together or not, but both a CA or VA project are paid professional services projects. Where the CE is just an automated monitoring and analysis, these projects get more intimate with the customer’s data center. Therefore, the results are a detailed and tailored roadmap and design for virtual infrastructure. For example the VCP delivering this project interviews the client’s IT team for current issues and future expectations. Time is taken to understand all aspects of the virtual migration such as networking, storage, and services infrastructure (Active Directory). The strategy and methodology of a physical data center move is followed in that all aspects of the implementation are considered .

CA / VA deliverables are usually

• Corporate level report containing detailed stats and analysis gathered from Capacity Planner along with multiple proposed consolidation scenarios.
• Virtual Infrastructure Design Visio
• Recommended P2V candidates and ESX host assignments.
• Recommended SAN configurations and VMFS design.
• Recommended network configurations and subnet / VLAN design
• Recommended directory services design
• TCO and ROI calculations
• all of the CE deliverables

Target customers for these projects are brand new to virtualization and want help understanding how to plan for not only virtual infrastructure, but also how VI changes their entire data center.

Plan and Design

The Plan and Design project starts with the CA or VA project as it’s initial phase, but then takes moves into a test implementation and validation phase. A full assessment as described in the previous section is conducted with Capacity Planner. After presenting the results of the collected performance data and the associated VI design and migration roadmap, the VMware Partner then begins a prototype implementation with the customer. Using either the proposed new hardware or similar existing and available hardware, implementation of the design begins for a subset of the complete VI solution. This subset can be a “dress rehearsal” or put directly into production, but the purpose is to be an exercise to validate, document, and provide knowledge transfer to the customer so that they can complete the design on their own.

Plan and Design deliverables are usually:

• Capacity and Financial Analysis
• • Includes TCO and ROI calculations
  • Includes all the deliverables from the CE pre sales project
  • Includes all the deliverables from the CA / VA project
• Detailed documentation including
• • VI Architecture Design
  • VI Assembly and Configuration Guide
  • VI Standard Procedures
  • VI Test Plan
  • VI Roll out Plan
• Design enablement and knowledge transfer using
• • Side by side implementation following VMware best practices
  • Utilization of a prototype proof of concept infrastructure

Target customers for the Plan and Design project mirror the CA/VA project as well as include customer’s with specialized application needs in virtual environments. For example, large database or messaging applications, usage of unique hardware devices shared to VMs, or non traditional deployment scenarios are all good reasons to engage in a Plan and Design services project.

Whether you are already running virtual machines or just now ready to migrate to virtual infrastructure, a VMware Partner using Capacity Planner can help you tailor and optimize a virtual infrastructure specific to your needs. If you are interested in more details for any of the projects discussed in this post you can contact me and I’ll be glad to talk to you about our offerings at Optimus Solutions (Optimus Solutions is now a Softchoice company).

The FAQs I used to create this post seem to be only available to VAC partners who’s accredited VCPs have completed the necessary Capacity Planner services training. Upon completing the training a VCP is assigned a log in to the VMware Data Warehouse Portal where Capacity Planner data collector’s uploads are processed into consolidation scenarios. Here, from this portal, is where these FAQ documents can be downloaded. I happen to be an accredited VCP, so I have access to the FAQs. All of the FAQs clearly state they are customer documents so I am now providing this information here. I’m not sure why these documents aren’t publicly available because I am sure they would help generate more interest in Virtualization Assessment service projects.

The remainder of this post combines VMware’s information from 3 different documents – the Capacity Planner Technical, Security, and Sales FAQs. The individual documents are provided on the VM /ETC Files page and are linked at the bottom of this post. I have trimmed down or omitted most of the FAQs in this post (and it’s still a lot of info to read), but check out the full .pdfs for more details and other FAQs I left out.

I recommend you at least skim through some of these very popular Q & A’s:

• How is Capacity Planner sold?
• Why should we add Capacity Planner when we already have so many other tools installed?
• What ports need to be open?
• What operating systems does it discover?
• What impact does collection have on my network?
• What impact does collection have on my servers?
• How are the utilization figures determined?
• What are the security features within the software?

Go get a fresh cup of coffee or your favorite beverage and then enjoy!

Capacity Planner Sales, Security and Technical FAQ

Licensing

How is Capacity Planner sold?

Capacity Planner is sold as a service to customers by one of VMware’s selected partners or as a service from VMware PSO. Customers are unable to purchase licenses for their sole use; a partner always needs to be involved.

How is license usage monitored; for example, if a customer buys 500 server licenses, what prevents the customer from deploying it on, for example, 600 servers?

Capacity Planner support monitors actual versus expected server numbers analyzed and works with partner and Capacity Planner sales on discrepancies.

What if a customer discovers a need for more server licenses?

The customer should order more. In the future, customers should be aware that they have the option of using the Collector for free to discover servers. Discovery provides a server count, the operating system on each server, and typically identifies the type of server.

Competitors

What are the key competitive products?

The primary competitive products are IBM CDAT tool and PlateSpin PowerRecon.
Other enterprise capacity planning tools, usually agent based, include BMC Perform/Predict, Hyperformix, TeamQuest, and Metron Athene. There are overlaps with inventory, asset management, and performance management solutions. Therefore, some people might consider IBM Director, HP Openview, BMC Patrol, and Mercury SiteScope as competitive products.

How does Capacity Planner compete with competitor’s products?

Capacity Planner is faster because it is doesn’t use an agent and does provide automated analytics and decision support. Capacity Planner is more accurate because it can discover and enumerate the entire (heterogeneous) IT infrastructure within an enterprise through a variety of means. Capacity Planner correlates inventory and performance data to provide more accurate and meaningful analysis than most other tools on the market. It also simulates various planning scenarios, including virtualization and procurement, to help test for accuracy before implementing the best solution. Capacity Planner’s Information Warehouse is a unique differentiator that houses a growing set of industry reference data that capacity planners can leverage to drive intelligent, benchmarked IT capacity decisions for the enterprise. Furthermore, Capacity Planner can provide anomaly detection and ongoing recommendations based on these industry benchmarks to ensure capacity optimization for the enterprise.

Is Capacity Planner a performance management tool?

Although Capacity Planner collects performance statistics and makes them available for review, the tool is not a typical event management or real-time performance management tool. The main purpose of the product and service is to identify candidates for consolidation using virtualization. The tool can also be used to provide advice around tuning servers and applications prior to consolidation or virtualization.

Potential Difficulties

Why should we add Capacity Planner when we already have so many other tools installed?

Many organizations collect similar data using various different tools (event management, asset management, patch management, and so on); however, the data from all these tools usually resides in different databases or datastores. With Capacity Planner, all the relevant data can be collected quickly with little overhead on the collected systems. In addition, many applications sit on the shelf because of the amount of time and expertise required to deploy them. Our agent-less technology alleviates this issue. Most clients install our software within minutes and are reviewing inventory and performance data within a couple of hours. In most environments, our solution is fully deployed within a couple of days.

Security Methods

How do we know our data is safe?

Our client base, which includes Global 2000 banks, insurance companies, and retailers, researched this issue and concluded that the risk of transferring data that is not sensitive through Secure Socket Layer (SSL) encryption over HTTP was low to nonexistent. Sensitive information like IP addresses reside only at the customer site. Customers access their information through a secure, Web-based dashboard and manage user IDs and passwords so that they can implement the same password restrictions as required by their internal security policy. An example of transferred data is available.

What are the security features within the software?

To keep sensitive data secure, Capacity Planner collects the data securely and stores it using highly secure methods. Our collection methods use standard OS APIs. These APIs are the same APIs that you use to copy files from one system to the next and to logon securely. These APIs have been accepted as C-2 level secure and are updated with hot-fixes every time a problem is found. The data is stored in a local database that is protected by file system security. Passwords are encrypted before being stored. Our products have the ability to send information through the Internet to our Web site for analysis and comparison. All data sent and received from our Web site uses HTTPS. This is a secure transmission protocol that is used by all institutions.

The basic security methods are the following:

· Local administrator accounts are required on all target servers.

· Accounts are stored and encrypted on the Collector server in an Access database.

· Capacity Planner uses a RC4 password encryption with a private key.

· Passwords are not shown in clear text on the Manager.

· CSV files sent from Collector to Capacity Planner data center do not contain usernames, passwords, IP addresses, or share information. CSV files do contain domain names and server names.

Requirements

What is the necessary inventory information that you need to conduct an assessment?

For server consolidation and other capacity planning activities, project teams need to know detailed hardware information around four core hardware components: processors, memory, disk, and network interface cards. Detailed intelligence for applications, services, and shares is equally valuable.

What administrative rights are needed to access servers?

On a Windows environment, users enter either global, domain, or individual administrative rights to access remote servers into the Manager. UNIX and Linux systems require root access to access the appropriate data from the software.

What ports need to be open?

We need the following ports open: 135, 137 through 139, and 445. If there are servers behind firewalls that can’t be opened or will be challenging to open, it is often easier to install a Collector inside the firewall.

Installation

Do I need to install agents?

With Capacity Planner, you do not need to install any agents on the target servers. Capacity Planner is agent-less software that collects across server infrastructure, leveraging existing data sources already on the systems. As a result, Data Collectors only need to be installed to pull the data from the target systems.

Do I need to purchase other software or hardware to run this tool?

Unlike other competing products, Capacity Planner was designed to leverage basic hardware and software requirements. Capacity Planner requires a Windows 2000 or newer operating system on a laptop, desktop, or server with at least 1GHz CPU, 512 MB of RAM, and 250 MB of free disk space.

What is the appropriate operating system for the Collector and Manager?

Capacity Planner requires Windows 2000 or 2003 with an ASCII language operating system. VMware mandates US English Windows 200x operating system as best practice.

Discovery and Collection

What does a discovery collect?

Discovery using Capacity Planner is simply a system count of the environment. There is relatively little information beyond a physical count that is provided. The discovery enumerates the list using the Active Directory, IP Scanning, DNS queries, and NETBIOS options. It does not verify that the machine is online or accessible.

What operating systems does it discover?

Capacity Planner collects from Windows NT 3.51, NT 4.0, 2000, and 2003; Red Hat Linux 8 and 9 and Enterprise Linux (ES/AS/WS) 3 and 4; SUSE Linux 8, 9, 10, and Enterprise Server 9; HP-UX 10.x, 11.0, 11.11, 11.22 (PA-RISC) and 11.23 (Itanium); and Solaris 7, 8, 9, and 10 (Sparc) and 9 and 10 (x86) Operating Environments.

How are discovery and inventory different?

Discovery is a free count of the number of systems in the customer’s environment. Inventory is a paid service that collects all the hardware, software, and services inventory data.

What inventory data is collected?

Capacity Planner collects hardware, software, and services inventory data.

What performance metrics are collected by Capacity Planner?

Capacity Planner collects from targeted servers 300+ core performance statistics and additional relevant statistics for specific applications. This low-overhead query collects performance metrics from the four main data groups of processor, memory, disk drive, and network utilization. For memory, for example, Capacity Planner collects not only paging data or what is available in bytes but also specific cache information that affects the overall project decision strategy. This data is then correlated with the previously collected inventory data.

What impact does collection have on my network?

Capacity Planner typically only uses 20,000 bytes during data collection across a network.

What impact does collection have on my servers?

The impact on the target server varies, depending on what is collected. Windows systems traditionally are affected less than 1% utilization. UNIX systems might peak at 5 to 10% utilization during inventory collection cycles.

How long does Capacity Planner take to collect inventory?

This varies depending on a number of factors, such as the Collector system, location of target systems, and network speed. Traditionally, Capacity Planner collects inventory on one system every 20 seconds. Multiple systems can be collected at the same time as a configurable parameter.

How long do we need to collect performance data?

VMware recommends a minimum of three weeks of data collection to start to compile a server profile with at least 1,000 performance samples. In addition, performance collection over three to four weeks provides server trending on a daily, weekly, and monthly basis that could be critical to understanding when servers are peaking. In addition, the longer you collect data, the more valuable it is.

Does Capacity Planner collect application statistics?

Yes, it collects limited application statistics that are written to Microsoft Performance Monitor (PerfMon).

Utilization and Performance Counters

How are the utilization figures determined?

Utilization and performance counters are determined by collecting multiple samples each hour of each day for each week. The statistics for each hour over a week are correlated together to determine the average, hourly, prime time, and non-prime time utilization for each hour. Average utilization is typically the average utilization during the prime time hours of 7 AM to 6 PM, for example, or for the entire 24-hour period. Capacity Planner also maintains weekly summary statistics that track maximum observed, minimum observed, average, hourly, prime time, non-prime time, and weekend loads. Capacity Planner also maintains a summary for the most recent four weeks of performance statistics on these same criteria. The summary is used to determine peak load for consolidation recommendations. The peak load is determined by evaluating each metric over the most recent four weeks of collection and locating the hour of the day that has the highest average value. It is not the maximum observed value as any statistical analysis eliminates the high and low values from consideration.

What is the difference between peak hour performances versus average utilization?

Peak hour is the one hour in the twenty-four hour period that has the highest average utilization. Average utilization is typically the average utilization during the prime time hours of 7 AM to 7 PM, for example, or for the entire twenty-four hour period. The difference between these two metrics is significant.

Figure: Peak vs. Average Utilization

Assume that servers A, B, C, and D in Figure 1 were Microsoft Exchange servers that were to be consolidated based on average utilization. If this were the case and all peaked to 40 percent utilization in the morning and at the lunch hour, the server would run out of capacity at those critical times.

As another example, assume that the servers in Figure 1 were Citrix servers. On average, most Citrix machines that Capacity Planner has monitored run low CPU utilization, low memory utilization, and extremely low disk utilization. However, the largest variable is network utilization. During peak loads, usually in the morning during sign-on, the network traffic may go from an average of 100,000 bytes per second to millions of bytes per second. If one used average utilization or even prime-time utilization metrics for consolidating a Citrix server, problems would likely arise.

Identifying the peak-hour average provides capacity planners with an upper threshold for consolidating servers. How important is the difference between prime time averages and the peak hour average? Analysis of data in the information warehouse reveals that the peak-hour average is two times higher than the prime-time average for 68 percent of all servers. For 32 percent of all servers, the peak-hour average is five times higher than the prime-time average.

The standard installation samples each server once an hour. Are all servers sampled simultaneously or divided across the time period? Can this be configured?

Servers are sampled in groups of ten servers. They are not all done simultaneously; instead, they are sampled in order until all are sampled. Performance collection is a configurable option within the Manager.

How does Capacity Planner calculate peak hour and other calculations?

Capacity Planner collects data every hour and calculates peak-hour utilization for each one-hour increment in the 24 hour day. After several weeks, it identifies utilization for the busiest hour in the week. Capacity Planner also maintains weekly summary statistics that track maximum observed, minimum observed, average, hourly, prime time, non prime time, and weekend loads. Capacity Planner also maintains a summary for the most recent four weeks of performance statistics on these same criteria. The summary is what is used to determine peak load for consolidation recommendations. Peak load is determined by finding the hour of the day with the highest sustained load. It is not a measure of maximum observed values.

Hyperthreading

How does Capacity Planner handle x86 processors that have the hyperthreading feature enabled?

Capacity Planner detects hyperthreading based on the model for the server sent to Capacity Planner. If the chassis model is sent, Capacity Planner checks to see if the hyperthread flag has been turned on for that model in the information warehouse. If the hyperthread flag is on, Capacity Planner ignores half of their CPUs. If they send four, Capacity Planner only enters two into the system, assuming that they did not go through the process to turn hyperthreading off.

Capacity Planner flags a chassis as being hyperthreaded when it is either manually entered or detected by comparing the maximum number of CPUs allowed in the chassis to the number of CPUs being reported. If a chassis can only have four processors, but inventory is sent showing eight processors for that model, Capacity Planner turns on the hyperthread flag for that model.

Proposed New Hardware

Does proposed new hardware have VMware automatically included on it?

Proposed new hardware does not automatically have VMware included on it.

How can I help make VMware recommendations?

When leveraging Capacity Planner to run consolidation or virtualization scenarios, users can create a new server with specific hardware specifications and either VMware ESX Server, VMware Server, or standard Microsoft, Linux, or UNIX operating systems. These systems comprise a group of servers to run consolidation scenarios against. The scenarios will allow users to pick their various future platforms and operating system combinations, and then run scenarios to determine how many proposed systems are needed.

How does Capacity Planner determine the native MHz for proposed new hardware?

Capacity Planner reduces the amount of native capacity when multi-processors are involved. The rule of thumb in the industry is to reduce the amount of native capacity available to 80% of the original rating for the second processor. Subsequent processors are multiplied by .8 raised to the power of the slot number of the CPU. So the third CPU contributes .8 raised to the second power multiplied times the rated speed (the third CPU is in slot 2 because slot numbers begin with 0). If this is a 64-bit processor, the percentage is raised to 90% due to improvements in the technology. Therefore, the variables are the number of processors, the rated speed, and the slot number.

Platforms

Is VMware comparing like with like when comparing Intel and AMD based platforms? If people believe the AMD CPU is more powerful, is their a way to apply a correction factor for AMD platforms?

AMD Opteron is not more powerful. It is a different technology that provides better throughput under certain circumstances. In an Intel platform, all the processors make memory requests through a common physical interface. AMD Opteron has a separate physical interface for each processor. The bottleneck to memory has been removed, and the speed to the memory is faster as well. This is the big difference. It is also a 64-bit platform, but that alone does not offer an advantage over an Intel 64-bit platform. There are two things to consider. First, are you comparing a 32-bit to a 64-bit platform? Second, are you comparing Intel to AMD Opteron?

The only way you see significant improvements in throughput is in a memory intensive application. Most applications are not total random memory access intensive, only about 2%. If a system is paging due to limitations on the file system cache, 64 bit may solve the problem by allowing a much larger file system cache. If a system is paging because every reference to memory is for new data that has not been read into memory before, this is a disk access and memory speed issue, primarily disk access. If you have a system that is constantly pulling data from memory but not requiring disk access, this favors the AMD Opteron. This typically means large database servers. When we combine a number of systems in a virtualized environment, this creates a memory intensive situation on the ESX Server overall and favors AMD Opteron. There is no way to apply an algorithm to the statistics that would show lower processor utilization. The processor will not run at a lower utilization, but the throughput and response times will be better. Because it cannot be represented statistically, the technology and the ramifications are explained to clients. Essentially, the statistics point out that running an AMD Opteron at 65% utilization and Intel at 50% utilization provides greater throughput with AMD Opteron.

Transferring Data

Are there other ways to transfer data besides HTTPS?

HTTPS is the default transfer mechanism, but you have the ability to transfer data through FTP, email, or burning the data to a CD.

Can the data be exported?

Users can export data from either the Manager or Dashboard to a CSV file.

Reports

Can you create custom reports?

Dashboard includes Dynamic Reports that permits users to create customizable reports and export them to CSV files for further analysis.

Downloads

1. Capacity Planner Sales FAQ (1543)
2. Capacity Planner Security FAQ (2084)
3. Capacity Planner Technical FAQ (4053)

]]> http://vmetc.com/2008/07/11/everything-you-ever-wanted-to-know-about-vmware-capacity-planner/feed/ 10 http://vmetc.com/2008/04/09/treat-your-virtualization-project-like-a-data-center-move/ http://vmetc.com/2008/04/09/treat-your-virtualization-project-like-a-data-center-move/#comments Wed, 09 Apr 2008 11:59:11 +0000 Rich Brambley http://vmetc.com/2008/04/09/treat-your-virtualization-project-like-a-data-center-move/ Why is it that migrating to virtual infrastructure (VI) is most often considered to be the responsibility of the server administrators? Anyone who has already done it can tell you it involves much more than servers and hardware. Even for small companies, virtualizing servers potentially (and usually) involves networking, storage, security, and infrastructure services changes. In fact, it is often as involved and complex as moving your physical servers from one data center to another.

So why is there a perception that implementing VI is only a server team responsibility? Obviously, one reason is because on the surface it’s about installing and consolidating operating systems on server hardware. Another reason is because VMware, the source for all the interest in migrating to VI to date, has done such a great job marketing their products as just server installs.

The deliverables of the BCE are a 7 to 10 page .ppt presentation that tells you little more than the number of VI3 Enterprise licenses you’ll need to purchase to achieve an amazing server consolidation ratio. It’s an effective pre-sales tool to shock customers into realizing the savings in hard and soft costs that will be realized by a migration to virtualization. Sure, you can use the BCE results to quote server hardware needed and even guesstimate the appropriate storage requirements, but the BCE does not tell you everything you need to know to migrate your servers.

Don’t get me wrong, I am for BCEs and I think VMware’s marketing and product positioning is brilliant. VMware would not be where it is in today if ESX was not adaptable enough to be a implemented as just a server install and configuration. The VMFS file system, despite all the negative press and opinions lately, enables the server admin to request large chunks of storage from the SAN and then use that storage without any further interaction from the storage team. Virtual switches and port groups allow server admins to request only basic trunking from the networking team. Truly, that’s all that is needed for a “jumpstart” implementation.

What’s missing is the planning and design details for a high performance, tailored virtual infrastructure. That means requiring your engineers across all of your technologies have a responsibility in architecting the virtual solution. Therefore, companies need to treat the virtualization project like a data center move. Make sure the networking, storage, and operations engineers understand virtualization and collaborate. After all, you are moving your servers to a new location. It’s just not a brick and mortar facility.

For more reading about how others are expressing the same opinions (and my motivation for writing this post) check out:

Challenges integrating VMware into Cisco networks – Colin McNamara

“Last April forty of us, all senior engineers attended VMware Certified Professional training at the same time. The class was mixed up so there was an even distribution of CCIE’s, Systems Experts, and Storage Experts. Needless to say this presented our instructors with some extremely challenging questions, but more importantly it set the stage and created a venue for collaboration between these different practices within our own company.”

How virtualization impacts IT staff, security and change management – Scott Lowe

“Now the server administrators who manage the virtualization servers are talking to the network engineers about VLANs (virtual LANs), 802.1Q trunks, and the native VLAN. The network has been extended inside the virtualization hosts, and now the network group and the server group needs to work much more closely together than in the past.

The storage team is now much more heavily involved in server deployments, as every new virtual server that gets turned up is stored on the storage area network (SAN). And the storage team might even have a more in-depth involvement with networking as the use of IP-based storage –iSCSI, NFS, or FCoE — becomes more prevalent.

As server virtualization invades the data center, teams within the IT organization have to be prepared to work more closely together than they may have in the past.”