Chris Wolf
Research VP 4 years at Gartner 16 years IT industry
Chris Wolf is a research vice president for Gartner's IT Professionals service. He covers server and client virtualization, cloud computing, data protection, management and classification, disaster recovery, and business continuity. Read Full Bio
The first PCoIP vs. ICA/HDX WAN performance test in my experiment began at Newark airport. I’m not one to spend more time than absolutely necessary in an airport, but I still gave myself plenty of extra time to conduct and record my tests.
The AT&T 3G connection between my laptop and home office lab offered great bandwidth (1.08 Mbps), but not-so-great latency (122 ms). Both PCoIP and ICA/HDX responded well in this environment. I had no problems with Microsoft Word, navigating the Start menu, or drawing with a Flash-based application. Here are the videos:
After arriving at my hotel in Munich, I was ready to again compare PCoIP and ICA/HDX performance. When I connected to my lab, I was surprised that latency from Germany was actually less than with my AT&T 3G connection in the US. While I measured latency at 122 ms when connecting using my AT&T 3G card, latency from my hotel in Munich measured only 106 ms. As expected, bandwidth wasn’t as good as with my AT&T 3G connection. I measured 480 Kbps from the hotel to my office lab.
PCoIP and ICA/HDX performed very well in this environment. Text entry in Microsoft Word was smooth and had a natural feel. Navigating the Start menu, launching IE, and working with the Flash application was also easy to do. Here are the comparison videos:
Last Wednesday I flew from Copenhagen to Stockholm, and then drove with my Gartner colleagues to a hotel in Gavle, Sweden, which was close to where we would meet a client the next day. I didn’t have much time to work with my virtual desktops while at VMworld Europe; that was mainly due to the fact that Internet connectivity was not working at my hotel.
Last Tuesday I used my VMware View client while listening to the conference keynote. During the keynote, I used TweetDeck in my View virtual desktop to tweet my thoughts. The conference Internet connectivity was surprisingly solid; I only had 117 ms latency back to my office lab. VMware View and the PCoIP protocol performed extremely well. At one point during the keynote, I opened this video on my View desktop. The video played without missing a beat. Alessandro Perilli was seated next to me and he too was impressed by the results.
Fresh off a solid performance at VMworld, I was excited to compare VMware View and Citrix XenDesktop from my hotel in Gavle. The hotel’s internet connectivity was reliable, but slow. I measured 208 ms latency and 265 Kbps bandwidth to my office lab.
The VMware View desktop connected without a problem and performance when adding a comment to a Word doc was OK. There was a slight delay between my keystrokes and when the text appeared on screen, as you’ll see in the video. PCoIP struggled when I tried to launch Internet Explorer from the start menu. It took several seconds for the menu to load before I was able to click on the IE shortcut. Once the flash application loaded in IE, performance was pretty good.
View 4.5 PCoIP over 208 ms latency
XenDesktop fared slightly better with Microsoft Word. There was also a slight delay with text entry, but I found it better than PCoIP. You’ll notice a greater difference with the Start menu navigation. With ICA/HDX, the response was almost immediate. Opening the browser and working with the Flash application was pretty good.
This is the last post in my series “PCoIP vs. ICA/HDX: A Simple Experiment.” As a final test, I connected to my VMware View and Citrix XenDesktop environments from my hotel in Oslo, Norway.
I considered myself pretty lucky. For the last test, the conditions were pretty extreme. I consistently measured 405 ms of latency to my office lab, and bandwidth was low as well. My netperf tests yielded a point-to-point bandwidth of 131 Kbps.
Let’s start with VMware View. It took me three tries to complete a video of the test actions. On the first two occasions, my connection dropped. I’m not sure if the issue was due to the unreliability of the network connection or PCoIP’s ability to handle the very high latency. However, I did not experience similar problems while using ICA/HDX. As you can see from the video below, once I loaded an application, text entry was OK. However, user experience was far from ideal.
PCoIP over 405 ms latency
I wouldn’t call my experience with ICA/HDX outstanding, but as you can see from the video below, performance was substantially better. Text entry was responsive and menu navigation wasn’t perfect, but still performed really well considering the high latency. Here’s the video of my XenDesktop experience.
ICA/HDX over 405 ms latency
What do you think? Have you experienced similar or better results under similar conditions?
Today the Wall Street Journal reported “VMware in Talks to Buy Novell Unit.” The rumor likely comes as no surprise to those who have followed the recent VMware/Novell OEM agreement. The agreement resulted in VMware including a SUSE Linux Enterprise Server (SLES) subscription with vSphere licenses. VMware and Novell first started talking about an extended partnership in June. At the time, VMware noted that it would include SUSE Linux with its vSphere hypervisor as well as train its support organization to offer SUSE Linux support. The fact that VMware was making an investment in its support organization hinted at the potential of a larger deal.
In addition, VMware stated that its virtual appliance authoring tool, VMware Studio, would eventually offer SUSE Linux as the default VMware appliance OS. This marked a significant departure from VMware’s own Just Enough OS (JeOS) operating system.
If the rumors are true, there are a few reasons why I think acquiring a mainstream Linux distro makes sense for VMware:
JeOS wasn’t seeing wide acceptance in the enterprise. Many of our clients with strict OS certification and support policies were only comfortable supporting specific Linux distros (e.g., SUSE and RHEL) within their organization. For some special purpose VM appliances, JeOS had success. However, for many use cases JeOS was not acceptable due to factors such as integration with operational software (e.g., security and backup).
Three of VMware’s major competitors – Microsoft, Oracle, and Red Hat – are offering a hypervisor as part of a vertical stack that also includes their own supported OS. VMware is at the mercy of OS vendors and it’s understandable that its competitors would optimize their own OSs to work best with their hypervisor offerings.
VMware’s Open PaaS strategy is designed to give organizations the option to deploy Java applications to a variety of cloud providers as well as to internal infrastructure. When you deploy an application internally, it has to be on an OS accepted and supported by the IT organization. JeOS wasn’t a sure bet, so VMware needed more. The logical answer is SUSE.
Acquiring SUSE Linux is not about what VMware is doing over the next 2-5 years. They’ll continue to do very well running Windows OSs as VMs. However, over the next 5-10 years, the stakes will change. Microsoft will work to more closely integrate the Windows OS and its Hyper-V hypervisor. Paul Maritz knows that as a platform for Windows apps in a market where Microsoft is a direct competitor, VMware will eventually succumb to Microsoft. VMware has to work to chip away at their customers’ reliance on Microsoft OSs and applications. Propping up partners such as Google and Salesforce is a necessary step. Building a competitive application platform (via the SpringSource acquisition) is another. Offering an accepted OS for the application platform completes the puzzle.
If VMware acquires Novell SUSE Linux, it will have a vertical offering much like Microsoft, Oracle, and Red Hat. Many of VMware’s predecessors have proven that the best technology often doesn’t win in the long run. You can ask Novell that. This strategy will not guarantee long term success for VMware (Maritz is intimately familiar with Microsoft’s ability to compete and win), but offering an OS and a vertical stack is necessary to fend off similar offerings from competitors.
Once upon a time an IT user named Goldie Locks tried to access her company’s applications…
I’ll spare you the rest of the analogy. Some devices are too big, others too small, and some are just right. Of course, “just right” often depends on the application being consumed. I talk a lot with clients about user-centric computing and the fact that many in today’s mobile workforce do not want to be tethered to a single physical device. “Give me my apps on my terms” may represent a Utopian user viewpoint, but it’s still one that’s a priority for many IT organizations. Of course, we’d love to meet the user somewhere in the middle, and ensure that security is not compromised for the sake of device-agnostic application access.
We may never find Utopia, but accessing corporate apps from an approved set of devices isn’t unrealistic. We’re already doing that in many cases today (e.g., email). And that brings me to the problem. Tools like Citrix Receiver and Wyse PocketCloud can allow users to access their virtual desktops (or also published applications with Citrix Receiver) on their iPhones and other smart phones. These tools are useful, but they have their limits.
If you would like to see my point, try accessing Microsoft Word (either as a published app or as part of a virtual desktop) using Citrix Receiver or Wyse PocketCloud on an iPhone. It is a lesson in frustration. I don’t have a similar experience checking email on my iPhone because I’m not navigating a full blown version of Outlook using screen scraping technology. I get just enough email app, and it works.
Don’t get me wrong. Tools like Citrix Receiver are extremely useful, and are even better on devices like iPads. However, they’re only going to take us so far. Goldie Locks will never be satisfied if she continues to look for the perfect device. The real problem lies in the application platform. We no longer consume our productivity apps on PCs alone. Ideally, forthcoming improvements to today’s application platforms will dynamically adjust the presentation layer based on the endpoint device. That is a far more scalable means to support user-centric computing than either having to recode apps for specific devices or try to shoehorn an app onto a device not suited to present it well. Granted, this is a pretty tall order and we’re not going to get there overnight (probably closer to 5-10 years). In the interim, tools like Citrix Receiver will remain an excellent option for remotely presenting applications to users on a variety of endpoint devices. Choice shouldn’t have to be inconvenient (too much app on a small device) or expensive (too many application packages and platforms to support a variety of devices).
Are tools like Citrix Receiver on the iPhone the best we can do? Or does cloud give us the opportunity to expect more out of our application platforms. What do you think?
Today at VMworld North America, VMware announced the availability of VMware View 4.5. The View 4.5 Premier release represents a significant milestone as it places VMware in very select company with Citrix as the only two vendors that offer enterprise-ready server-hosted virtual desktop (SHVD) solutions.
The evaluation criteria was developed over a five month period. During that time, we worked with numerous early server-hosted virtual desktop (SHVD) adopters, as well as the key vendors in the space. In the end, vendors were supportive of the criteria in spite of the fact that not one met all of our requirements. The reason for the support was simple – customers were telling vendors they needed the same elements that we identified in the criteria.
We evaluate and score SHVD platforms across three stratifications:
Required: absolute necessities
Preferred: important features that result in better experience, operational management, and improved TCO
Optional: use-case driven features needed in select deployment scenarios
The assessment is broken down across major areas of focus such as user experience, management, and security. I blogged about our criteria when it was first released in May. At that time, no platform had all 52 features we consider requirements for the typical enterprise.
Note that I’ll be presenting the criteria along with scorecards for VMware, Citrix, Quest Software, and Microsoft at VMworld North America this week. You can view my presentation schedule here. The View 4.5 Premier scorecard is shown below.
When we first assessed VMware View 4.0, there were several significant shortcomings that were deal-breakers for many large enterprises:
Role-based access controls (RBACs) for delegation of administrative duties
Administrative change logging capabilities to provide an audit trail for all administrative actions
Official support for Windows 7 operating systems
Enterprise management software integration
View 4.5 addressed all four of the above shortcomings, and the breadth of their feature improvements were deeply scrutinized with hands-on assessments in our lab. To VMware’s credit, they didn’t try to address customer management requirements with band aids. Instead, they literally scrapped their previous management console and replaced it with a far improved Adobe Flex-based console. In addition, they unveiled a Microsoft System Center Operations Manager (SCOM) management pack for View 4.5 management. That was another common request I’ve heard from early VMware View adopters. On the scalability side, View 4.5 is now capable of scaling to 10,000 managed desktops per management domain, which is currently double the maximum scalability supported by Citrix.
Of course, while View 4.5 is a major release, there is still room for improvement. For example, if you want to optimize PCoIP traffic through WAN accelerators such as Riverbed appliances, you’re out of luck. Out of the box, VMware will not support highly constrained WAN environments (i.e., 50 Kbps bandwidth and over 150 ms of latency). VMware’s current recommended solution is to combine the View 4.5 “local desktop” feature with an HTTP proxy server at the remote site.
While there is always room for improvement, View 4.5 meets all of the core requirements of the typical larger enterprise. Enterprise customers understand that any next generation desktop solution requires a long term (typically 5 year) commitment to yield significant ROI and TCO savings. So today many are looking for a platform that they can standardize on and grow with. With the release of View 4.5, VMware is making the case that it should be the enterprise desktop and application delivery platform of the future. Citrix and other VMware competitors will not take this news lying down. To see a full comparison of how View 4.5 stacks up against its competitors, I encourage you to attend my VMworld session on the topic. Also, next week we’ll publish our full 25 page assessment of View 4.5. The assessment, available to Burton Group research clients, takes a much deeper look at View 4.5 Premier, and exposes key strengths and shortcomings at a very low level. Once the report is available, I’ll update this post with a link.
I have three sessions at this year’s VMworld North America conference and thought I would post them here in the event you’re interested in attending one of them. Each session will run two times, and the session descriptions, times, and locations are listed below. I hope to see you at the conference. The in-depth discussions I have with fellow attendees has always been one of my favorite parts of the conference.
Server-Hosted Virtual Desktops: What the Vendors Aren’t Telling You (DV8044) Many organizations are beginning to implement or plan server-hosted virtual desktop solutions. Vendor platform assessments in the emerging client virtualization market are often difficult due to a lack of defined acceptable standards. In this session, Gartner Research VP Chris Wolf shares Gartner’s benchmark for evaluating server-hosted virtual desktop solutions, including criteria for evaluating a solution’s deployment, management, performance, integration, and user experience capabilities. The session concludes with a breakdown and scorecards of popular vendor solutions, including the current VMware, Citrix, Microsoft, and Quest Software products.
Session times:
Tuesday August 31st, 2:00, Moscone South 302
Thursday September 2nd, 10:30, Moscone West 2003
Private Cloud Security: Vendor Secrets and Hypervisor Competitive Differences (SE8098) When it comes to building a private cloud or sizing up public cloud providers, hypervisor choice heavily influences security. In this session, attendees will see side-by-side differences in the security architectures of VMware vSphere, Microsoft Hyper-V, and Citrix XenServer hypervisors. Emphasis will be placed on the capabilities of each solution to monitor and enforce organizational security policy within the virtual infrastructure, along with the ability of each solution to securely host multi-tenant environments. Attendees will leave this session with pointed questions to ask vendors regarding architectural deficiencies in their platforms, along with a detailed list of industry best practices to securely build private cloud infrastructure.
Session times:
Wednesday September 1st, 1:30, Moscone South 301
Wednesday September 1st, 4:30, Moscone South 301
Cloud Futures: The Infrastructure Authority (MA8092) To realize the potential of private cloud, infrastructure must be capable of not just dynamically provisioning and optimizing systems, but also not violating any security, regulatory, or organizational policy constraints in the process. In many enterprise environments, dynamic IT consists of several disjointed solutions and oftentimes blind faith that policy, security, or regulatory constraints will not be broken. The bottom line – someone has to be in charge. The infrastructure authority (IA) is the future nerve center of cloud infrastructure as a service (IaaS) operations. Among the many roles the IA possesses are:
Provides a central metadata store
Leverages common data models to request or offer services
Maintains physical, virtual, and policy dependency maps
Ensures security and regulatory compliance
Ensures that service level requirements are met
Stores and enforces organizational policy
Ensures accurate capacity forecasts
Integrates with third party management and orchestration tools to authorize IT operations such as provisioning or relocation before they proceed
Typical questions answered by the IA include:
Are security zoning rules checked before live migrating a VM?
Do any policy restrictions prevent VMs from migrating to different data centers or to public cloud infrastructure?
This session takes a practical look at the emerging role of the IA, and details how existing management frameworks such as VMware vCenter and industry standards such as OVF can be used in this capacity moving forward.
Today I was working with a client on their next generation data center architecture. They are building a highly virtualized data center with the goal of offering cloud IaaS to other departments within the organization. While talking about VM templates we discussed a favorite topic of mine – virtual hard disk structure.
For several years, I have recommended to clients that they use at least two virtual hard disk files per VM. One virtual disk file is used for the OS and application files, and a second virtual hard disk is used for paging, swap, and temp files. Alternatively, a third virtual disk may be created for data files.
The result for VMware environments would be at least two .vmdk files per VM. For Hyper-V, that would mean two .vhd files. On the surface, this may seem like an academic exercise, but it’s important.
Supposed that sometime down the road you wanted to begin leveraging asynchronous replication to replicate VM data to another site. If you have your transient data (e.g., pagefile or swap file) on a separate virtual disk, it’s much easier to filter the data so that it’s not replicated. You may say “Well if I need that functionality I can just configure it later” and that’s true. However, remapping a pagefile to a new disk, for example, requires a reboot. The result is downtime. If you create a separate virtual hard disk for the pagefile and include it in your default template, all newly created VMs would be able to more elegantly and efficiently take advantage of storage features such as asynchronous replication. Asynchronous replication is just one reason. The amount of storage intelligence and flexibility creeping into virtual infrastructure is rapidly expanding. VMware’s new vStorage APIs for Array Integration (VAAI) is a good example. In the end, what’s the harm of having a second virtual disk in your default VM template? The result is one more file in a VM’s folder. If you don’t separate the pagefile and wish to more intelligently manage it at a later time, the result may be downtime. What do you think?
The evaluation criteria was developed over a five month period. During that time, we worked with numerous early server-hosted virtual desktop (SHVD) adopters, as well as the key vendors in the space. In the end, vendors were supportive of the criteria in spite of the fact that not one met all of our requirements. The reason for the support was simple – customers were telling vendors they needed the same elements that we identified in the criteria.
We evaluate and score SHVD platforms across three stratifications:
Required: absolute necessities
Preferred: important features that result in better experience, operational management, and improved TCO
Optional: use-case driven features needed in select deployment scenarios
The assessment is broken down across major areas of focus such as user experience, management, and security. I last blogged about our criteria when it was first released in May. At that time, no platform had all 52 features we consider requirements for the typical enterprise.
The initial XenDesktop 4 release lacked:
Role-based access controls (RBACs) for delegation of administrative duties
Administrative change logging capabilities to provide an audit trail for all administrative actions
Enterprise-class support (3 year minimum) for all products in the XenDesktop 4 Platinum portfolio
Among other improvements, the above three shortcomings were addressed in XenDesktop 4 SP1 release. You can read the updated support policy here. Burton Group extensively tested the above features in our lab to ensure that they met our requirements, and the result was a platform that we can recommend for large-scale enterprise environments.
The XenDesktop 4 SP1 scorecard is shown below.
As you can see, while meeting our essential requirements, XenDesktop 4 SP1 meets 76% of our preferred features and there is still room for improvement. Areas we would like to see improved include:
Management consolidation: Citrix must reduce the number of consoles required to manage the XenDesktop environment
Management complexity for very large environments: In Citrix’s reference architecture, each XenDesktop 4 Desktop Delivery Controller (DDC) runs 5,000 desktops. Citrix recommends that organizations build separate management domains to horizontally scale XenDesktop management, placing greater challenges on areas such as configuration management.
Like any product, there’s always room for improvement. Regardless, I’d like to congratulate Citrix for listening to the needs of the enterprise customer. Today Citrix delivered a product that customers can confidently deploy in large-scale environments while satisfying all of their key SHVD platform requirements.
Chris Wolf
