Thin Client FAQ

Frequently Asked Thin Computing Questions (FAQ)

GoVirtual® is the industry leading provider of VMware training videos and VMware training courseware books and educational materials. Their VMware training products and services are some of the highest quality and most comprehensive available today. Thin client computing within virtualized environments is one of their training specialties and is something they’ve done recent consulting projects with as well.

Today I’ll be Gregory Perry, CEO of GoVirtual.

Why is thin client and remote desktop computing getting so much attention these days? What are some of the business factors that have changed within recent years to make these applications desirable?

There are several reasons why thin client computing has taken off recently, namely that large scale corporations and organizations have already invested heavily into their server virtualization infrastructure, with thin client and desktop computing being the next logical segue in their progression of green datacenter initiatives.

Desktop virtualization products such as Citrix XenDesktop and VMware View are built and predicated upon large scale virtualized server farm environments such as VMware vSphere; once a company has invested in the underpinnings required for server virtualization, thin clients and remote desktop computing requires very little additional investment to implement and deploy.

The primary impetus for thin client and remote desktop computing is reduced management complexity. Traditional desktop personal computer management requires dedicated IT personnel to image, deploy, manage, and maintain end user PCs.

From a management and maintenance perspective, thin clients replacing traditional PC technology on the desktop almost eliminates the requirement for dedicated IT personnel; at a large company’s next hardware refresh, instead of imaging and deploying new desktop PCs, thin clients can be provisioned and drop shipped to end user locations where they are simply plugged in and then automatically associate themselves with a virtual machine running in a virtualized datacenter server farm.

Thin clients have no moving parts, consume just a fraction of the power normally used by a desktop PC, and are virtually indestructible. Further, from a regulatory compliance standpoint, thin clients and remote desktops reduce the overhead associated with keeping information private – with no storage on the thin client, the desktop PC and information stored on its local hard drive can’t be lost or stolen, at which point regulatory compliance and records privacy then becomes a function of the datacenter instead of a company having to worry about safeguarding client records on each and every desktop PC used within the organization.

Lastly, thin clients offer a tremendous advantage from a patching and maintenance standpoint, as virtual machines can be centrally managed and patched using a variety of centralized patch management frameworks. In addition, if there is ever a hardware failure, instead of dispatching IT personnel to fix the desktop PC the end user can simply request a new thin client which is drop shipped to the end user location instead of the usual headache associated with traditional desktop PC maintenance and lifecycle management.

Who are some of the major players in the thin client market, and how are they differentiated?

Thin client computing has two primary components: the end user hardware, and the server that runs the virtual PC on the backend. For thin client computing hardware, probably the biggest player in the market today is Wyse, although almost every major manufacturer is getting into thin client computing hardware.

To get a feel for the many different vendors that are offering thin client hardware, visit the VMware Hardware Compatibility Guide for VMware View to see a list of everyone they are doing business with for thin computing hardware: http://tinyurl.com/27r6j96

The backend datacenter software for thin client computing is broken up into two distinct parts: server virtualization software (the hypervisors that run on the datacenter blade/server hardware), as well as the desktop broker component that actually brokers connections and makes associations between thin clients and the corresponding virtual machines that are running in the datacenter.

Without a doubt the leading provider for the server virtualization piece is VMware, with VMware Infrastructure 3.5 and VMware vSphere 4.0/4.1 running in most large scale datacenter environments, followed by Citrix Xen server.

Most large scale companies and organizations are just now finishing with their first revision of server virtualization, which means they are probably a year or two into production with their virtualized server environments.

The next step is deploying desktop broker software on top of their virtualized datacenter server environment (as well as procuring the thin client computing hardware).

The industry leader for the desktop broker piece is Citrix XenDesktop, followed by VMware View 4.5. Most companies are mixing both VMware and Citrix to get the best of both worlds, as VMware dominates the server virtualization piece while Citrix leads the desktop virtualization part of the equation.

In terms of presentation protocols (the actual presentation layer that connects the thin client to the virtual machine over the LAN), Citrix ICA leads the market, with Microsoft RDP being used for everything else.

VMware has recently announced their own presentation protocol, VMware PCoIP, but few companies are actually in production with PCoIP yet. The highest performer is Citrix ICA based upon their Speedscreen technology, with Microsoft RDP 6 in a close second place.

When is thin client computing advisable for businesses, and when is it not recommended?

Thin client computing is advisable whenever there are already robust server virtualization (or Microsoft Terminal Services) and Storage Area Networking (SAN) environments in use, with qualified and competent staff on hand to deploy and manage the server virtualization and/or Microsoft Terminal Services pieces.

It should be noted that the cost benefit of using thin client computing in conjunction with Microsoft Terminal Services has gone down significantly with the advent of Windows 7 and Server 2008, based upon the increased per annum CAL licensing prices and the more restrictive license roles that are associated with Microsoft Terminal Services used in production Server 2008 environments.

Thin client computing would not be advised within organizations that don’t have server virtualization environments already in production, or are using applications that are not suitable for thin client computing (such as heavy media rich applications, CAD/CAM software, graphic intensive desktop publishing environments etc).

Anything multimedia or graphics rich (which requires graphics acceleration on the desktop PC) is not an application suitable for thin client and/or remote desktop computing.

Other barriers to entry would be a heavily decentralized organization with low speed/high latency links between geographically disparate sites; as the thin client is basically a “dumb terminal” that talks over the network to a virtual machine running in the datacenter, any network obstacles between the client and remote VM can impede the user experience and potentially disrupt company business if their WAN environment is of substandard performance or improperly maintained.

Can you provide a list of questions that a company should ask when evaluating thin client solutions?

Support is key, ie which desktop presentation brokers are certified as supported for use with the thin client hardware, and which desktop presentation protocols are supported for the presentation interface (Citrix ICA, RDP, PCoIP). Obviously the thin client that supports the widest range of desktop broker and presentation protocols is the better choice, although cost goes up with functionality.

Other questions to ask would be related to security, ie what type of link layer encryption if offered by the thin client vendor to protect in transit thin client to virtual machine traffic, and what are the enterprise capabilities associated with the initial provisioning and day to day maintenance of the thin clients.

The company has to build and support the thin client provisioning process, where the thin client is taken out of the box, inventoried, provisioned, and then drop shipped to the end user location. Periodically new firmware updates are deployed to remote thin clients to address stability and security issues, so the thin client vendor has to support a streamlined and efficient initial provisioning process, as well as an efficient and secure method of managing the firmware images that are pushed to the remote thin clients that have been deployed in production environments.

Secondary missives would be seamless integration with existing network management platforms so that health statistics can be polled at periodic intervals to proactively detect hardware and/or firmware issues with deployed thin client hardware, with Simple Network Management Protocol (SNMP) being used for that piece.

With thin clients, is it better for organizations to manage their own servers or outsource this work to a trained professional? What is the learning curve like?

The theory is that thin client computing should drastically simplify the initial deployment and day to day management of end user PCs and workstations. The reality is that there is always a balancing act done between these various missives, and with any technology you definitely get what you pay for in the thin computing arena.

If a company has selected a thin client vendor based upon enterprise management capabilities, then the learning curve should be fairly simple; if on the other hand a company chooses a thin client computing vendor on a cost basis alone, then more skilled IT personal would have to be leveraged to build a corporate workflow for initial deployment as well as the day to day management of deployed thin computing devices. For the backend datacenter piece, highly skilled engineers have to be used, especially if you are dealing with several thousand thin clients in a production environment.

This goes back to the “advisable for business” question; if a company does not have a robust server virtualization capability in place including skilled Storage Area Network (SAN) management personnel, chances are they are ill equipped to deal with a large scale thin client deployment based upon the larger consolidation ratio numbers associated with desktop virtualization compared and contrasted with server virtualization consolidation ratios.

Consolidation ratios are defined as the number of virtual machines that can be shoehorned onto a single blade or server; more recent generation datacenter computing hardware can support upwards of 100:1 virtual machines to blade/server ratios FYI.

What are some of the biggest mistakes that companies make when implementing and managing their own remote desktop servers?

They don’t adequately test everything up front, with a detailed and specific early adoption plan in place. Anticipated desktop OS and application workload analysis is key for thin client computing environments, which means the company must absolutely understand from an empirical perspective the exact behavior model that is associated with their standard desktop PC workload.

If you are deploying 2000+ thin clients in a production environment, then you must unequivocally have the hard numbers associated with the anticipated workload for all deployed thin clients because that directly corresponds to 2000+ virtual machines running in the datacenter environment.

Do you have enough CPU resources (blades/servers) in the datacenter? Do you have enough SAN storage for the virtual machines, and does your desktop broker software offer any type of desktop cloning to reduce the amount of SAN storage space needed?

Does your SAN vendor support deduplication to reduce the amount of SAN storage needed for all 2000+ VMs? What is the anticipated workload of the virtual machine, ie does it have a higher write to read ratio and/or random write bias which would then require RAID10 storage (more costly, less storage) as opposed to RAID5 (less expensive, less write performance)?

What is the anticipated network bandwidth required for each deployed thin client, and do you have enough aggregate bandwidth, throughput, and low latency at each site to support the anticipated consolidation ratio? These components are typically broken down into four (4) core resources: CPU, memory, disk, and network.

The most common workload analysis program used these days is typically Microsoft Perfmon, which can be used to monitor and provide analytics for the average workload that would be exhibited by current production desktop PCs, which would then translate to the anticipated workload the company would expect for the four core resources in a virtual machine environment.

Desktop virtualization is mind bendingly complex for large scale organizations, simply by virtue of the much higher anticipated consolidation ratios that are associated with desktop virtualization.

Taking 200 physical servers virtual is simple; virtualizing 2000+ desktops can be disastrous if the company has not asked and answered all of the foregoing questions by way of early adoption trials and detailed workload analysis or workload simulation of the current desktop OS and application environment.

How much should an organization budget when planning a move over to a thin computing model? Can you provide any advice for the transition process?

It depends.

What are the numbers involved?

How many virtual machines in the datacenter are anticipated, and which IT personnel are going to manage those VMs?

How much SAN storage space will be needed, what type of disk groups are required (RAID10 vs. RAID5 question above), and which SAN personnel are going to manage the storage piece?

How much additional network bandwidth is needed at each site for the increased network utilization of the thin client presentation protocol?

These are not easy questions to answer and require detailed up front planning to deploy properly. Based on this complexity, the two mainstream desktop virtualization vendors out there (Citrix and VMware) are now starting to offer “pay as you go” per seat flexible licensing, so that companies can deploy the infrastructure (hypervisor and desktop brokers) up front for free and then just pay a per-seat licensing cost for each thin client they deploy in production.

If a company has not performed detailed workload analysis of their existing desktop PC environment, chances are they will end up stopping their deployment a quarter of the way through to go back to the proverbial drawing board based upon unrealistic expectations that were assumed at the initial launch.

What are some of the biggest security threats that come from switching to thin clients?

Most security issues go away with thin clients, as the data is stored in the datacenter/SAN. However, the desktop presentation protocol then becomes the weak link, ie can thin client presentation traffic be viewed, does my thin client vendor support encryption of the transport layer, does it require a PKI, does the link layer encryption require additional CPU resources on the datacenter side based upon the computationally expensive nature of link layer encryption, etc.

In theory deploying thin clients should greatly enhance the security posture of any organization by virtue of the assumption that all end user data is stored in a heavily fortified datacenter and SAN environment.

At that point, the only security concerns should be the link layer between the thin client and the desktop broker that sits in between the server farm and thin clients, as well as the process that is used to digitally sign and deploy new thin client firmware images as they are released from the hardware manufacturer.

What are some of the biggest trends that you see in the future for thin client computing?

Probably touchpad integration (my iPad becomes my desktop PC/thin client) as well as mobile thin client computing (thin client laptops). With the advent of high speed 4G wireless connectivity, there is no reason why employees should even be tethered to a conventional thin client on the desktop.

If Internet connectivity is ubiquitous, then in theory the employee should be able to access their desktop from any remote desktop modality including handheld touchpads and netbooks with 4G connectivity.