Qlik Sense Server is a great product and my clients are starting to believe in the potential of open standards.
The most interesting opportunity for my clients is to start with Big Data processing and report visualization using directly the cloud.
Amazon Redshift, EC2, S3, Machine Learning and soon Athena, and the Glue will send all the other solutions in the middle age, and the cost of all this is 70-80% less than any on-premise infrastructure.
But, there is one big issue: the deep feeling of frustration installing Qlik Sense on a Windows Server.
For me, it's a flashback: it's probably 10 years that I don't use any Windows server anymore and I manage about hundred of servers in Enterprise environment.
The humiliation is using Windows Authenticator (a total failure on non-Microsoft Browsers: white page, no login, refresh to login) and creating the local users on the Windows machine(why?), dragging ugly windows on a remote desktop connection. Using a local Domain with Active Directory when the final users connect with Android, iOS, Chrome OS on large tablet with a pencil. Trying to download Qlik sense from qlik.com using Explorer with security patch! I wanna cry! (ops)
Let's try to imagine the other way:
yum install qliksense
http://localhost and then configure internal users/password as a simple solution, before any other authentication or access management solution.
mto citation: Qlik Sense is born to embrace open standard! Great!
So, why Windows? Not Open, not even anymore a standard for Servers on the Internet.
while you're waiting for the Linux distribution, want you aware, even now you are not forced to use the windows authentication and AD as user directory. Your user directory can sit on an Open LDAP installed on Linux and you can use authentication methods like header, ticket, session or SAML. (Ref. Authentication solutions ‒ Qlik Sense)
I passed through all those steps. I'm using now LDAP as user directory, and an Apache Reverse Proxy with LDAP authentication. Upgrading soon to SAML2 with ForgeRock OpenAM Access management.
I don't want to remark here why Linux is better than Windows Server. It's a closed challenge, just check the usage share of Windows Server. It's done.
Qlik Sense and QlikView are very similar in the development process but totally opposite in the reception of two different audiences. Probably the last servers running a Windows Server are those with Qlik products and Active directory (hard to die).
My clients (marketing, advertising, retail, distribution companies moved to Linux already ten years ago and to AWS in the five years) never thought about purchasing QlikView: a Windows past century application.
Servers, Desktop and Mobile Operating system: Microsoft, Mac, Google
Microsoft vs. Google Docs office applications
on-premises vs. cloud servers,
license vs. SaaS
working at the office and working from everywhere
do meetings and don't do meetings
presenting with a laptop or a big tablet
QlikView vs. Qlik Sense
split audiences worldwide. They do and always did.
Qlik has the chance to listen to both audiences and fork the distributions.
The QlikView legacy audience of Qlik is broad but is self-referring with its own standard of 20 years ago based on declining company like Microsoft, that stopped to innovate 10 years ago.
I'm here to help to define Qlik Sense modern audience working on open standards.
Going back to the point: the opportunity is to launch a server on AWS Linux EC2 and deploy a QlikServe server with:
yum install qliksense
(faster: launch a Qlik Sense pre-installed from AWS MarketPlace choosing the CPU and number of servers for the cluster)
QlikSense Server on Linux is an opportunity for a no-Microsoft new audience to get a better product quickly and for Qlik to open the market to new clients.
P.S. If you are a proud member of the Microsoft audience, no need to engage. I understand you.
If the black windows are the most exciting world you know, nobody prevents you from installing Qlik on a non GUI OS using the command line (Silent installation ‒ Qlik Sense), but honestly, I don't think the installation procedure have some impact on the business purposes
You are right, I totally removed from my mind the availability on MarketPlace. In fact, I evaluated that months ago, but I didn't understand why the instance was installed on R2012, instead of R2016. And I restarted from scratch with my own installation.
I also asked information about Qlik Sense on Spot Request Instances. Do you have any experience?
Hello, I totally understand your frustration, but due to the fact that Qlik is geared toward more to business users, small technical things like this usually being put aside until it becomes a business justification topic for Qlik to take action.
In order to resolve the authentication issue, you can either implement SAML federation or with OpenID external broker *ahem* Cognito from AWS. Bear in mind, production Qlik Sense site rarely uses the local user to administer. after all, it is a £1000 per token product.
So, how to effectively deploy QSE into a highly available, cost efficient and secure AWS environment?
To scale effectively, you will need the combination of PowerShell + Cloudformation to do so, this allows you to deploy a parallel QSE infrastructure to act as a slave node. OR Opswork will do the trick by implementing recipes in setup and configuration stage and QS repository API call to register your nodes as an extra slave node.
To utilise spot instance, you can call EC2 spot instance termination notice API and insert a pending terminate state to do an application snapshot and store it into S3, you then can repeat step 1 by bootstrapping all required dependencies to another spot instances.All of these be automated.
I've yet to test WindowsDocker with QSE. Maybe you can give a try share us the result too? Afaik the window 2016 core edition is prebuilt into the Docker, theoretically, once.net framework is installed, I reckon we could install QSE with share persistence into it.
Hope this helps. Let me know if you have any further questions.
EDIT: Honestly, from a technical perspective, the marketplace image is a waste of time for existing user like us. If a customer wants to try out Qlik Sense, we would point them to QS Cloud / Desktop. Lastly, with jenkins, codepipeline and cloudformation, we could build a better quality image of QSE. or ARM with Azure function.
Business opportunity for Qlik (good point): If I was Qlik, spectator of a declining OS architecture (Windows Server), and willing to conquer new markets where often the sysadmins drive the procurement in the software choice. I will consider Linux as an opportunity more than a burden. Qlik Sense has some minors original sins in the quality of the solutions, because of Windows: the faulty default Windows authentication, slow restarting process, a conceptual legacy architecture based on "domains" that is fading out. Young competitors like AWS QuickSight, ElasticSearch (cloud native, snappy, price aggressive with flexible billing, Linux or multi-platform) are moving fast: even still too young and too far from Qlik Sense. After, all I'm a proud customer of Qlik Sense.
Authentication: I'm already using SAML2 with our corporate ForgeRock OpenAM for the production server. OpenAM is very versatile: users can log in with any options, we control password-less sessions for our mobile apps; huge, rocky, not easy to climb.
For Qlik, there was just a traumatic approach for the first installations using default Windows authentication:
on a non-Windows computer (browser form) is extremely faulty (white pages force the user to log in many times; the user press enter on the browser and even worse; 1-2 minutes of user panicking, and in my case an automatic call to IT Helpdesk).
Login with domain\user. The black slash is an offense for non-Windows users and is a complication for smartphone users ("\" is not in the first layout or even the second layout of the touchscreen onscreen keyboard)
"Domain" is an old concept that doesn't work anymore in the modern real world, and not even in the IT. My largest client is a conglomerate multinational with about 10,000 users who span on multiple companies, multiple countries with dozens email alias per user and user/password credentials; users moving from one entity to another continuously and with a turnover of 20% by month (including also clients and suppliers) in a huge federated ID manager with more than 50,000 unique entries. In the modern world, nobody work for 1 company in 1 domain. This is an American-corporation-style from the '80s. Are you working for only 1 entity and in only 1 domain?
Keep it simple: I believe that Qlik should offer an easy solution for default (out-of-the-box) authentication: an internal simple directory service (just a table: user, email, password, groups, attributes) managed with QMC not related to the Windows users. Also, we know that Windows security hassles are in the everyday chronicles.
Self-Preservation and opposition to change: I understand the thousand of sysadmin and tech experts tried to block the progress to self-preserve their jobs and lifespan investment in outdating skills. I don't think Qlik should recycle, or crush that (also with Qlik View users that are paying the bill). Some companies did it and failed. (note: the title of the thread is defiant on purpose to attract more posts 🙂 ).
Options, not alternatives. More options will attract more potential clients and will give the chance to legacy clients to migrate if the want (or not). In the 2-3 years when QSE Linux will be at 90% and QSE at 10%, and all the big clients already migrated, then Qlik could consider to clean-up and remove Microsoft 20th century trail.
Gut Feeling: My job in the last 20 years is entirely based on migrations of enterprises from past to new techs, from popular to best. QSE Win to Linux is just matter of time.
Sorry, if I didn't comment on the other specific solutions that you proposed, but for some, I'm not an expert.
My teammate is evaluating Cognito (not yet in my AWS Singapore) that looks interesting, we are just worried about versatility to handle extraordinary cases that don't fit in the common organization schema. Did you already try it?
QSE cluster: I will be there soon, now I'm still enjoying the vertical scaling of AWS (just pick a more powerful instance type). We are often using Direct Query to a Redshift cluster on the same VPC; QSE Repository is not reaching yet 40-GB of data that fit in memory easily. But it won't last long, soon QS cluster will be the solution.
I forgot to add a detail that is probably the MOST relevant about Linux vs. Windows Server for Qlik Sense.
I'm using on AWS m4.16xlarge instance: 188 cores, 256GB RAM, 20Gbit network.
Linux: 4$ per hour.
Windows: 7$ per hour !!!
OMG: A Windows server cost 75% than a Linux server!
Wow. I can almost buy a second server. The question is: is really Qlik Sense utilizing Windows Server layers for his architecture? We really to 75% more expensive cost just a Microsoft heritage? or is just a tax to be in the Microsoft group, so much popular 20 years ago; just to feel cool.
I don't know so much about Windows Server on premise, but I think should cost less than 1000$ per server, right?
In a multi-node cluster, I guess you should multiply the number of the server, right? Is really necessary this 'Windows' past century fee? What could Qlik Sense really needs of the magic environment of a Windows Server?
Windows Server is the Non-Sense of the Qlik Sense Server!
From a business point of view, choosing Windows Server as the only distribution platform for Qlik Sense is a suicide, a net loss. The development, test, and distribution costs for Qlik Sense on Linux are probably less than the cost of 1 year Windows Licence of qlikcloud.com: break even in months.
If some Qlik Sense partner or Qlik Sense Product manager is reading, please let us understand. Or as I hope, you are working already on this migration, share that with us.
I don't want to bother, perhaps Qlik Sense mentor mto could enlight us.
Writing this post, I'm aware that I probably represent the sentiment of a small segment of Qlik clients: Qlik Sense + AWS, but on the other side, I have a strong feeling that this small segment is going to be the mainstream quite soon.