cancel
Showing results for 
Search instead for 
Did you mean: 
Create Post

Signal and noise of the cloud

Level 13

The cloud is no longer a new thing. Now, we’re rapidly moving to an “AI-first” world. Even Satya Nadella updated the Microsoft corporate vision recently to say “Our strategic vision is to compete and grow by building best-in-class platforms and productivity services for an intelligent cloud and an intelligent edge infused with AI.” Bye bye cloud first, mobile first.

In reality, some organizations still haven't taken the plunge into cloud solutions, even if they want to. Maybe they’ve had to consolidate systems or remove legacy dependencies first. The cloud is still new to them. So, what advice would you give to someone looking at cloud for the first time? Have we learned some lessons along the way? Has cloud matured from its initial hype, or have we just moved on to new cloud-related hype subjects (see AI)? What are we now being told (and sold) that we are wary of until it has had some time to mature?

Turn off your servers
Even in the SMB market, cloud hasn’t resulted in a mass graveyard of on-premises servers. Before advising the smallest of organizations on a move to the cloud, I want to know what data they generate, how much there is, how big it is, and what they do with it. That knowledge, coupled with their internet connection capability, determines if there is a case for leaving some shared data or archive data out of the cloud. That’s before we’ve looked at legacy applications, especially where aging specialist hardware is concerned (think manufacturing or medical). I’m not saying it’s impossible to go full cloud, but the dream and the reality are a little different. Do your due diligence wisely, despite what your friendly cloud salesperson says.

Fire your engineers
Millions of IT pros have not been made redundant because their organizations have gone to the cloud. They’ve had to learn some new skills, for sure. But even virtual servers and Infrastructure as a Service (IaaS) requires sizing, monitoring, and managing. The cloud vendor is not going to tell you that your instance is over-specced and you should bump it down to a cheaper plan. Having said that, I know organizations that have slowed down their hiring because of the process efficiencies they now have in place with cloud and/or automation. We don’t seem to need as much technical head count per end-user to keep the lights on.

Virtual desktops
Another early cloud promise was that we could all run cheap, low-specced desktops with a virtual desktop in the cloud doing all the processing. Yes, it sounded like terminal services to me too, or even back to dumb terminal + mainframe days. Again, this is a solution that has its place (we’re seeing it in veterinary surgeries with specialist applications and Intel Compute Sticks). But it doesn’t feel like this cloud benefit has been widely adopted.

Chatbots are your help desk
It could be early days for this one. Again, we haven’t fired all of the Level 1 support roles and replaced them with machines. While they aren’t strictly a cloud-move thing (other than chatbots living in the cloud), there is still a significant amount of hype around chatbots being our customer service and ITSM saviors. Will this one fizzle out, or do we just need to give the bots some more time to improve (knowing ironically that this happens the best when we use them and feed them more data)?

Build your own cloud
After being in technical preview for a year, Microsoft has released the Azure Stack platform to its hardware partners for certification. Azure Stack gives you access to provision and manage infrastructure resources like you’d do in Azure, but those resources are in your own data center. There’s also a pay-as-you-go subscription billing option. The technical aspects and use cases seem pretty cool, but this is a very new thing. Have you played with the Azure Stack technical preview? Do you have plans to try it or implement it?

So, tell me the truth
One thing that has become a cloud truth is automation, whether that’s PowerShell scripts, IFTTT, or Chef recipes. While much of that automation is available on-premises, too (depending on how old your systems are), many Software-as-a-Service (SaaS) solutions are picked over on-premises for their interoperability. If you can pull yourself away from GUI habits and embrace the console (or hand your processes off to a GUI like Microsoft Flow), those skills are a worthwhile investment to get you to cloud nirvana.

I’ve stayed vendor-agnostic on purpose, but maybe you have some vendor-specific stories to share? What cloud visions just didn’t materialize? What’s too “bleeding edge” now to trust yet?

19 Comments

Efficiencies have reduced our data center floor area requirements, and we've reduced our data center operators from a staff of 25 that operated 7x24 to two people who are here 8x5. 

Our DC server hardware growth was out of control eight years ago, and we'd run into cooling, electrical, and floor carrying weight capacity that had shut down our application expansion and growth.  Since VM and UCS came, we've eliminated MANY servers, lightened our financial load for cooling, staff, and lightened our physical floor load weight requirements immensely.

Part of the staffing reduction and electrical / cooling savings we've seen came from retiring our on site data centers and moving to managed data center space from an outside vendor.  They manage space & security & cooling & electrical needs 7x24, resulting in our decreased need for in house data center support.

Our VAR's are noting they're seeing a return to more dedicated hardware platforms instead of towards VM / UCS solutions, and the model is much like Google's & Amazon's, where servers are a commodity item bought in quantity, pre-built and mounted in entire racks complete with Cisco Nexus 9K's at their top.  Customers are buying these a rack at a time and rolling it into a DC and the Azure-style VM-like deployment is fully compliant with an external cloud or an internal solution.  This was a surprise to me.  I'd expected VM / UCS chassis solutions to be the continued future, and it's not the case, per information our biggest VAR just shared with us.

So, no turning off servers.  No cutting network engineers.  So far.

We're keeping our apps internal, and only looking to the cloud for saving millions of dollars on Microsoft licensing.  I certainly see decreased performance in Outlook now that we use Azure instead of in house Exchange servers.  That example is continuing to keep us from moving our mission critical internal apps to the cloud--for now.

Virtual desktops rule in most of my network--if that's was you'll call Citrix solutions.  I work in health care and have around 14K Citrix users.  I still rely on a full PC and a full laptop, but there are multiple apps & troubleshooting tools I rely on installed locally.  Maybe that'll change in the future, but I'm skeptical.

Chatbots are not part of our future, so far.  It's due to reliability needs, security concerns, and reachability.  On the other hand, we use wireless tools that are much like Star Trek's communicators--touch your badge, say "Call Dr. Johnson", and Dr. Johnson and you have a two-way voice session in real time.  It's pretty reliable now, but was a BIG problem getting to this stage.  It doesn't scale well to large groups of communicators in close proximity to each other.  Depending on the AP to which they attach, eight simultaneous conversations depletes the AP of all its resources, and no other wireless communications of ANY type are supported.  That's harsh, but it's reality.  We rely on Skype instead of chat bots, and it still has technical glitches that remain challenging to identify and correct.

We're more likely to build our own cloud than use the Internet based cloud for our mission critical apps.  It's because security and availability are king, here in the Health Care industry.  And NetPath proves the Internet cloud is NOT as reliable or predictable as AWS and Microsoft would have one believe.

Level 20

We've had some real losers with some projects that tried VDI... don't get me wrong we still have some but I can think of two projects off the top of my head where the engineers tried the vdi and hated it.

Level 21

Wow, this is a lot to comment on!

Cloud technologies have certainly been a huge disruption in the industry.  They have caused a lot of confusion and there has been a lot of misunderstandings propagated to try and sell these services.  This has left many companies not sure of what to do or how to do it.

My advice for any company that is in a situation where they are not sure of what to do or how to do it would be to find a trusted partner that specializes in cloud and work with them.  I am admittedly biased as that is what our company does but I also think it just makes good sense.  At the end of the day a trusted partner will be able to assess your current situation and make recommendations on which cloud solutions might be a good fit and which ones might not be and ultimately help you create a road-map to reach those goals.

Moving to the cloud doesn't change the fact that you still have systems and services that need to be configured and managed in perpetuity, it just changes how you do it and where those systems/service reside; this is a very common miss-conception about cloud services.  Again, this is where a trusted partner can help; they can augment your IT staff by providing the cloud expertise by helping setup and manage as much or as little as you are comfortable with.

As far as Azure stack is concerned, we have worked with it.  In fact we were a very early adopter before it was even called Azure stack.  We work very closely with Microsoft in their CSP program and manage hybrid environments on both Hyper-V and Azure with data replicating between the two.  We definitely will be setting up Azure Stack to further our Hybrid hosting capabilities.

Level 13

I may have to chat with you about the badge communicator thing.  I have been told that is coming to a site we support here and I will need to have solid SolarWinds monitoring for it and the supporting infrastructure.  I don't know the ETA for it here though.

MVP
MVP

Everything in IT cycles and each new technology is the bees knees, or whatever the latest cool phrase is. That said it's easy to get caught up in the hype of things and move too quickly (as a consumer - if you are building products slowing down can be just the thing to sink a company) These are good comments that remind us to slow down and evaluate - honestly - what we need before moving to cloud or VDI or SDN or, or, or. . . There's always a cost to being on the "bleeding edge" of networking and not just the high cost of the equipment before it becomes mainstream. Consider the cost of being the "beta tester" for the companies that are still designing and working out the kinks in their products and offerings. Consider the training costs and the load on personnel - IT staff and the users. Believe me, I'm not saying don't do it, whatever it might be today, but I've seen far too much money spent and far too many employees damaged because of lack of real, honest, thorough preparation and investigation.

IM me via Thwack, I'll share anything I can.  In my environment, multicast had to be enabled on all VLANs that contain the messaging servers and end clients before this was successful.

If you happened to have a very small number of SSID's, it might not be a bad idea to create a new one and dedicate it to this technology.  Better still if it has its own unshared DHCP scope and VLAN.

We use Cisco 5xxx and 8xxx WLC's, and the technology works across LAN and WAN, within reason.

If the vendor comes to set up a POC in your location, don't let them just set up one or three AP's with only one SSID on them.  Require them to use your existing infrastructure so the communications badges have to fight it out with the rest of the clients for radio spectrum and signal strength.

Create real-world situations where many badges are in the same physical area, like an E.R. receiving/triage area.  Imagine having a bad situation where a train or plane is involved in an accident, and many E.R. staff are present at once, all associated to just one A.P. because it's the only one within 40'.  Then see how well the technology works.  Their POC will not cover this, and the worst performance is when multiple badges are associated to the same AP.

Early versions of AP's had insufficient resources to handle more then 8 or 12 badge conversations, and the AP would drop all other communications until resources were freed by the badges moving out of range.

Ensure your low throughput rates (1 Mb/s and 2 Mb/s) are disabled, which helps force clients to move to closer AP's as people move about.

Oh, I could go on about what we've learned.  Where some users have problem-free use of this technology, they swear by it, they don't know how they lived without it, and how they could move forward if it failed.  Where it isn't reliable, test staff have simply thrown the $3000 badges on the ground in frustration, breaking them to pieces.

I continually remind staff to practice their down-time procedures, since no WLAN is as reliable, secure, or fast as a wired LAN.  Some day their service will be unavailable, and they MUST still be able to serve patients.

Level 13

You are bringing up many things that popped into my mind with this deployment.  I need to find the document about it I was shown a month ago, but it sounds like the exact technology they were going to deploy at this site, for the same reasons.

Level 13

Hey thanks for sharing your experience. I think people underestimate how popular Build your own Cloud will be. We saw similar things when we started playing with virtualisation in the late 90s. Suddenly we were consolidating servers like mad.

Level 13

VDI feels like a hammer looking for a nail, when sometimes you actually need a screwdriver.

Level 13

I think Azure Stack will be massive. It's great to see partners keeping up to speed with the latest tech and showing the way with Cloud best practices.

Level 13

Absolutely - with new and shiny comes risk, so as a consumer you have to be very careful throwing all of your eggs in a new product basket. Easier for larger companies to sink some time and money into a test bed, proof of concept etc, but there's still risk in making that decision to adopt into production when the product/solution is very new from the vendor.

It may simply boil down to an opportunity for companies to train their employees to be experts in something new, and then setting them loose to make it all work.

The trick will be to:

  • Identify what services & technologies will be appropriate for any part of a company's I.T. environment
  • Accurately predict the future so resources aren't wasted on technologies that will be here a small number of years, then replaced by something better.  I'm thinking of how my company nearly went down the ATM path, which would have wasted a lot of money without any gain for our environment.  And now that technology has fallen by the way side, and we're better off not having used it.
  • Finding budget and selecting the right training for the staff, to enable them to meet the challenges and succeed
  • Selecting the right VAR & Trainers to ensure the best path is selected

Turn off your servers: yeah, some. Mostly old legacy stuff as delivery systems and apps change. Not really cloud-specific, though - we've been reducing footprint for years.

Fire engineers: nah. Last time I checked, people with appropriate skillset are still required to administer systems, regardless of their location.

Virtual desktops: non-issue. VDI has a bunch of use cases, but our business isn't one of them. Plus, it just shifts the responsibility of support.

Chatbots: this is vapor and sounds like a Buzzfeed article title. Self-service is a more compelling use case for thinning the tier-1 herd.

Build your own cloud: this will happen eventually. Maybe sooner than later.

All these discussions are just SO BROAD when you don't take business-specific use cases into account, honestly. It's good to see the resulting conversation in the comments since these varied cases are fleshed out.

In addition to your comments I had the "pleasure" to get data out of the cloud because the provider was bankrupt.

as cool and shiny some new companies might seem, choose a market proven solution that will last for a while and gives you enough time to get your data.

Test the solution thoroughly

It is no fun at all to pull out 3TB of Data over a limited timeframe when all other customers are also trying to get their data out.

Or being stuck with 500GB of encrypted Mail Archives because the customer forgot to export the decryption key when the provider shut down their service.

Level 13

Ouch!

Level 13

Great points. Now if we could only develop an accurate crystal ball and sell it, we'd be billionaires.

Level 13

I agree! I like to think that my role here is to structure a skeleton of a topic and then we can chat about everyone's real world experiences in the comments. The community here has a wealth of knowledge of what has worked and what has not worked in reality and I love bringing those stories out.

I love your phrase:  "Embrace The Console!"

RE3Ko6.gif

Level 13

VDI -- Ugg.