In Part 1, I wrote about some of the new developments in the world of protocols, both in the technologies (VMware Blast Extreme, Citrix HDX, and Teradici PCoIP), as well as how the dynamics of Cloud/SaaS and VDI are changing the marketplace. Today, we’ll talk about how new use cases are driving technologies, as vendors expand beyond EUC and cloud into IoT, VR/AR, and other new industries.
New use cases and considerations
So, what is driving innovation in protocols? Often, it’s the niche cases that eventually find their way into the mainstream.
Virtual reality/augmented reality
VR/AR requires frame rates of 90 to 120 frames per second (fps), way above the levels of commodity VDI (commonly 16 to 24 fps), high-end graphics VDI (30 to 60 fps), or helpdesk (8 to 12fps—those budget VNC-based protocols). While most of the VR/AR industry focuses on headsets and localized deployment, there are strong use cases for treating the headset as a thin client, with the bulk of the processing in the data center, like VDI. (Gabe wrote a bit about Citrix’s efforts last year.) The higher frame rates and progressive display features that are emerging for VR/AR will indirectly benefit your average VDI user.
Recently I reviewed a cloud product from VFX (Visual effects software) vendor Foundry, which indicated its intention to utilize Teradici; interestingly, there were strong hints that this was linked to Foundry’s intention to offer HDMI with 10-bit (or more) per channel (40-bit VDI). Very few protocols support 10-bit per channel, with the major VDI vendors such as VMWare and Citrix offering 32-bit VDI by default.
If you’re not familiar with these terms, a 32-bit VM uses 4 channels to transmit each pixel, with 8 bits per channel, and the channels being RGBA (or Red, Green, Blue, and Alpha, which is a measure of transparency). Like 24-bit color, 32-bit color supports 16,777,215 colors, but with the additional alpha channel it can create more convincing gradients, shadows, and transparencies. So, 32-bit color actually supports 4,294,967,296 color combinations. Most protocols do 32-bit color; more bits per channel means more bandwidth.
I wouldn’t be surprised if we see HDMI support (i.e. even deeper color depths) from Teradici fairly soon, opening up markets beyond DaaS, such as software vendors in high-end graphics sectors like VFX and medical. Mechdyne TGX is another protocol to watch in the high-end visualization space; they indicated their intention to support stereo 3-D.
For your average Citrix/VMware user, 25% extra bandwidth for a practically indiscernible improvement in image quality won’t be of interest, but it will enable a class of industries previously slow to move to the cloud (medical, film, visual effects) to start joining the EUC party.
8-bit and 16-bit VDI
One of the smartest innovations in recent years has been Citrix’s offering of 8-bit and 16-bit VDI (that’s 2-bit or 4-bit per channel in the protocol). The focus on marketing of protocols is often on high-end graphics CAD/design, whereas for an awful lot of remote solutions, bandwidth is now the top consideration—so why use unnecessary bandwidth to branch locations if you don’t need to?
I spoke to a user delivering terminals for fruit packers who was terribly excited about this. In this case, the screens are simple primary-colored buttons (often dusted with soil), saying 1 kg potatoes, etc., and the terminals are often in remote rural areas with high constraints on bandwidth. 8-bit and 16-bit options make Citrix/HDX technologies suitable for a large genre of simpler commodity applications.
GPU integrations and adoption
With growing adoption of GPUs and hardware that protocols can leverage for encode/decode, there are increasing numbers of protocol integrations with GPU, CPU, and SoC vendors, as well as increasing choice and competition. For example, over the last few years AMD and Intel have joined NVIDIA in the GPU virtualization and sharing space, and increasing numbers of protocols can now leverage AMD and Intel hardware.
With HDX and Teradici following the hybrid codec approach, and most GPU vendors only supported for full screen H.264 encode/decode, it’ll be interesting to see if hardware partners bring forth integrations to support hybrid codecs.
All the bells and whistles
It’s easy to focus on the graphics in protocols, but as long as they’re good enough, the decider is often around a load of fiddly bits protocols need to throw, in such as:
- Peripheral support; such as smart cards, USB, 3-D mice, and Wacom drawing tablets. If a customer uses these, their absence is a showstopper regardless of frame rates or pixel perfection.
- Application integration; for example, protocols optimised for unified communications (Cisco Jabber, Google hangouts, or Skype for Business).
- Client OS availability; with end-points becoming more diverse, a large, well-QA’d portfolio of software for Android, iOS, Linux, and Windows is a must for the large players.
So, who is winning?
In traditional large-scale VDI, Citrix ICA/HDX is probably still the technology leader. (Disclosure: Until 2015 I was a product manager for HDX at Citrix, though now I work as an independent consultant.) This is not just in their hybrid protocol and codec strategy, but also in their integrations with networking products such as NetScaler and SD-WAN (previously known as Branch Repeater and CloudBridge). These integrations enable caching and de-duplication (even between sessions) across large deployments.
Citrix’s weakness, though, has sometimes been its complexity and quality of implementation, as well as constant change. With numerous restructurings, product changes, and team shuffles, there’s a risk that the overall experience of using Citrix is defined by other factors, regardless of the technological design of the protocol.
VMware has done a good job on their H.264 implementation, but as Citrix found some years ago when they made H.264 their default, H.264 alone just doesn’t cut it in enterprise scale remoting. (Citrix backtracked and enabled their “Thinwire+” alternative as the default for most users.) Personally, I’d like to see VMware embrace Teradici’s newer PCoIP variants (now found in Teradici Cloud Access Software) and allow the large numbers of their user base with PCoIP hardware to leverage it.
Back in 2015, Brian Madden wrote about Shawn Bass’s predictions that H.264 would dominate, but I’m still not convinced. As long as VMware Blast Extreme announcements remain focused on full-screen H.264, I’m struggling to get excited about its protocol technology, as I don’t see what has changed to make full-screen H.264 fundamentally more appealing. Yes H.264 has a very important role for protocols, but not alone.
Winning the protocol war may not win the sale!
It’s very easy to get caught up in technological advantages, but at the end of the day in mainstream EUC, price and value is king; far more people drive a Ford Mondeo or Fiesta than a Tesla or Aston Martin. So if the protocol is good enough and reliable enough, while you might not win the protocol war you may win the VDI sale!
Interestingly, neutral broker vendor Leostream dropped support for Citrix products recently. Could this be a sign that the big Citrix bundle isn’t taking hold in the cloud world of DaaS and SaaS? While HDX may be very good, you have to buy into the whole Citrix caboodle and it seems many aren’t.
“Why did they drop support for Citrix? As Karen explained it, there are a few reasons. Some customers are indeed moving away from Citrix because it’s complicated and expensive, and they very rarely saw customers using XenServer. Net-new customers are looking at other options, including hosting on public clouds, OpenStack, and even platforms like HP Moonshot Systems, which they’ve been noticing recently. And when it comes to graphics, RDP is fine for many tasks these days, and in other cases, customers are looking for support for specialized protocols like Mechdyne TGX and HP RGS.”
Citrix’s VDI baggage may not place it well for many in the cloud, and VMware is likely to face the same challenges.
Teradici is now well placed for the cloud/SaaS markets but their products aren’t cheap. Additionally, confusion over the PCoIP brand and some hostile benchmarking may mean Teradici often isn’t at the table when, actually technically and business model-wise, they are a perfect fit. VMware and Citrix don’t license their protocols to third-parties, so the market for Teradici as a standalone is wide open.
Where are the masterminds?
AWS acquired the NICE DCV team collectively and this is becoming a trend in protocol land. Building protocols is not easy. Many going the DIY route will find a v1 H.264 product is a fairly easy first step, but going beyond takes investment and experience.
While Citrix has some very strong technologies, the Citrix churn and re-organizations have led to large numbers of teams key to HDX’s development moving en masse to other vendors. A large number of the U.K. HDX performance team are now with NVIDIA, and a large chunk of the famed Sydney Citrix Labs team joined VMware. With 100s, if not 1,000s, of years of ready-made protocol experience out there, I wonder if something interesting might pop up from a few ghosts! I’m particularly hoping we may see something exciting for VMware Blast Extreme.
With so many developments, there has also been a glut of very good articles on benchmarking protocols and measuring bandwidth consumption, methodologies, and results. A few I’d personally recommend checking out include:
- Citrix’s Daniel Feller is a prolific blogger and tester of protocols in all sorts of scenario,s with real results and a pretty vendor neutral tone.
- There are a number of interesting articles from Teradici regarding PCoIP vs Blast Extreme. These are high on technical details, but have a hint of where the relationship with VMware may be heading.
- Brams Wolf and Barry Schiffer’s RDAnalyzer project has built up tools and a wealth of community knowledge sharing around protocol analysis.
- Marius Sandbu is another EUC community guru who’s put out a lot of info on protocol comparison. He has some interesting observations on Teradici CAS PCoIP and VMWare PCoIP comparisons, both in this blog and the resultant comments.
So are the protocol wars over? Of course not—don’t be silly!