Awingu: Simpler Citrix alternative
awingu.com
awingu.com
The whole concept doesn’t work. I have spent years and years dealing with miserable staff who have to sit on the end of a mucked up, crawling semi paralysed Citrix deployment because someone ticked a corporate IT box and got a pat on the back of it. They literally have their potential productivity halved on a good day. Not to mention the pain of trying to get some desktop software I’ve pushed out to standard windows machines fine working with some fucked up implementation of partial profile redirection that some MSP thought was a fantastic idea.
Really 99% of Citrix and terminal installs I see these days are box ticking and a false promise of cost savings through density and centralisation.
The end story is an end user trying to use a LOB app in chrome with about 200Mb of RAM at their disposal with the CPU and IO rammed at 100% while the MSP who sold them this pile sells them the lie that it’s the LOB vendor who’s at fault.
Edit: AWS Workspaces is just as bad. It’s like working with oven gloves on after a heavy night drinking.
As for data exfiltration I’ve worked in the defence sector in IT which has pretty stringent security controls in place. At the same time it’s very easy to do without much effort.
The key thing with this technology is box ticking to say you have a policy in place and to avoid liability, not to solve the problem effectively. This is only done through picking ethical, rigorous and professional staff. But that doesn’t scale because they are the 1% which is the real problem. So back to liability box ticking.
Also, it keeps getting improved-today’s RDP has very little to do with what we had five years ago.
RDP works well for me personally in a variety of use cases (albeit none of them are graphics intensive) - however, user feedback I have received especially from engineers and designers has shown that they are unhappy with it.
For example you cannot (anymore) use discrete graphics in an RDP session (e.g. in certain CAD or 3D software). I think this was under development but got removed, iirc. Other protocols like the one used by Splashtop do support this but in our tests RDP on W2k19 still provided higher fps and a smoother experience than Splashtop (with the downside that some software wouldn't start over RDP because it couldn't directly access the graphics card).
Are there any particular optimisations you do to RDP beyond what can be done in the "Experience" and "Display" tabs of mstsc.exe?
One successful example is a support call-center with 2000 users. The specific role used about 30 apps and they were all packaged and maintained by my team. Using a ci/cd fashion to build os image and apps we could roll out updates to base image and sandboxed applications completely seamless to these users.
(Side note: We built an automation that synced the two latest versions of all applications to a locally attached SSD drive on all hosts - app startup was near instant for all users which was fantastic at this point in time)
Running the clients inside the datacenter brought the latency to a minimum which greatly improved the performance of specific business critical apps.
Managing all parts pertaining to logon properly had the users ready to take calls in 10-15 seconds - a great improvement compared to the previous distributed client/server setup.
- Reducing latency between application and backend was a major win. Greatly appreciated by users as we saw massive reduction in latency for specific operations (crappy apps).
- Being able to integrate with the ip-telephony call-center software required a windows workstation environment hence the choice of “VDI”
- No noticeable os and app updates EVER was greatly appreciated by the users. Even when moving from XP to 7 at one point.
- Our ability to roll out massive OS and app updates at will to pilot groups who can roll back with a logoff/login was greatly appreciated by the team
Granted, this was 10 years ago and my opinion was always that these business applications should be rewritten - BUT I have yet to see a better managed and performing windows client & app delivery.
Citix provisioning services (PVS) coupled with app-v/app streaming was an incredibly progressive way of delivering windows workspaces.
During my time at citrix I often raised the fact that PVS for linux coupled with docker was an amazing setup - never got any traction on this though… everyone was way too focused on windows delivery.
I hear the same success stories from the MSPs while simultaneously their users are crying into their coffee on a daily basis. In fact “sorry our systems are slow today” is a normal catchphrase these days.
Only high performance call centre outcome I’ve seen in recent years is run by Indian scammers.
These integrations was built at HQ, ”works on my laptop style”, but at a branch office 1000km away… no fun at all!
An unholy mess that would take many years and a couple of $100.000.000 to untangle.
I had weekly “continual improvement meetings” with the power/super users, one or two from each branch office. The whole team spent a lot of time at the local branch offices dog fooding the service and talking to the users.
I used the service as my only workplace for three years in a development & ops role.
The success factor here was that we built a solution for a specific set of users that we continuously improved. I’m talking “make something slightly better today than it was yesterday - every day” type of mindset.
We started out with a PoC by brokering sessions to physical machines located in the datacenters. First users was the assigned "super-users". They were assigned due to being the most demanding and loud user base.
After a few week they _demanded_ that every user should get the same setup - we interpreted this as a successful PoC and went on to explore options.
At this time neither Citrix or vmware had a concept called VDI and we started out using pre-release code from vmware.
First prod version was using vmware but we only gained performance, which ofc was a win, but we still needed to manage all the vm's through standard automations (os patches, asset and inventory, application repackaging and roll-outs)
When Citrix bought ardence to get PvS it was a no-brainer - ICA protocol, PvS and app-v streaming was a great combo to rapidly deliver a high performing workplace. BUT - you need automations and reliable network links, both of which we could provide.
We had dedicated hypervisor hosts with the fastest local storage available. You should NEVER user SAN/NAS or other centralized storage systems for these workloads IMO.
Also, monitor the performance closely! Caching all application to host local SSD really took care of a lot of the potential perf issues we saw early on.
Depending on user role they could get dedicated hardware as well, but this was only for a select few users.
It was set up as a specific service delivery and my team was dedicated to improving and supporting the users.
One of the most rewarding projects I've been involved in, mainly due to the really close collaboration with end-users.
First batch of users used a windows client on their pc. When we after some time commited to the solution we built a really tight linux os image with only the citrix client that we pxe booted the PC’s with.
This allowed us a fallback in case of emegergency by simply rebooting a user.
This was a fun project in and of itself!
As we grew confident together with the users we started to phase in thin client instead of buying new PC’s - all in all we ended up with a really cost efficent and high performing setup.
We started to move the central infra from “hyperconverged” to commodity 2u servers with local solid state disks which really made it easy and cheap for us to scale.
It also made the whole service more reliable compared the integrated blade and storage solutions we were ushered in to at first.
Our technology demonstration operated beautifully but in the end we decided not to go with it because (1) we needed to rearchitect our SMB storage to centralize it, so there were no round trips or long distance SMB (2) Our network performance could be inconsistent at times due to jitter, which was not a problem with the solution, but the underlying network performance as specified, vs as-built, for some recent WAN enhancements.
A colleague at another location built a solution that I utilized as a desktop for sometime. Despite being ~1400 miles away and over our Center to Center network, performance was comparable to my engineering workstation, was always patched, excellent graphics performance, and was highly accessible provided strong authN.
There were some substantial differences in both of these:
(1) No oversubscription
(2) Fully capable of handling bootstorms
(3) Dedicated hardware / high performance VDIs
(4) Not viewed as a cost avoidance but as an access enhancer
(5) Optimized for expensive engineers / scientists where labor rates are the overwhelming cost, so 5 minutes of extra productivity more than pays for a solution.
Microsoft offers free 180-day versions of Windows Enterprise foe download at the click of a button that you can test and destroy or even use for a very realistic PoC. Why can't other companies offer that?
Unfortunately, once a company has something like this, they use it like a crutch to solve problems that are better solved the right way.
I saw a setup where people were forced to develop software over Citrix because of some mumble/security/mumble reason. Meaning everything via Citrix: IDE, browsers, test tools, terminal session, and so on. The productivity was atrocious.
I am SUPER sensitive to latency. When things are slow, it just feels broken to me.
But, I worked at a really successful company where everything happened in an Amazon Workspace and despite the annoyances, the place was productive and in particular, transitioned seamlessly to WFH with Covid. Seen a lot of other places like that.
So yeah it's annoying but you can't just look at the cons without the pros.
It is quite a different story on my 5K iMac however. The latency of that setup is too much for me.
I do have a fiber connection, but the entire setup is good enough to have Teams calls over redirected devices (video is laggy, but voice works OK).
Then again, I do have a fast connection, and have long learned to tweak some things that make a massive difference (you don’t need 32bit color to code, for instance).
It’s a tool. Like all tools, it can be misused if you just use it to beat down nails without some forethought…
- the 14-day trial (PoC'ing something like this in anything less than 8 weeks is just not something that happens in my experience, especially if upper management is involved and you still need to build the infrastructure around it)
- lack of lab licenses for free or cheap (I'd want to put that through its paces myself for a significant amount of time and dogfood it before recommending this to clients or moving over a large org)
- seeing 20 or 50 licenses min. purchase requirement is particularly disappointing when positioning yourself as a Citrix alternative. There is a ton of small outfits in niche industries out there with a requirement of 5 to 15 seats who might jump at this if it performs well.
- lack of attractive education and non-profit pricing which, again, might be a particularly thankful and interested target audience.
VDI is awesome, flexible and efficient if it performs well and is done right. Unfortunately in practice too often neither is the case. However the concept's potential still has a future if it sees further improvement and innovation. Cloud gaming, among others, shows that algorithms and tech stacks exist that can make this happen.
They are a pretty big company with a wide portfolio.
Or is Workspace all they are known for anymore?
I seem to remember the company was called a more unixy “Citus” early on.
Spent a few years working there with customers using it on solaris and eventually linux as well.
I remembered almost correctly then - something about it having unix roots. :)
Here's an interesting article on installing v1 of their product (on the same site there's follow-up articles on v2 of the Multiuser) - which builds on very early MS version of OS/2: https://virtuallyfun.com/wordpress/2017/11/17/installing-usi...
I helped manage many solaris servers running citrix metaframe at one point, many years ago.
NX never really took off and is now somewhat obscure. I rarely see it mentioned anywhere!