Cloud Computing: Let's kick this off

4saken

[H]F Junkie
Joined
Sep 14, 2004
Messages
13,147
I figure this forum will probably be a little more business/industry related, rather than hobbyist as the Virtualization forum is(which has and is been a great resource for years here). But, lets face it, the enterprise is going this way whether people want to admit it or not. So let's maybe start by just checking in on who currently is heavily involved on a day to day basis with AWS, Azure, GCE, docker, DevOps, orchestration, config mgmt, etc, etc. Of course not all of the latter technologies are bound to cloud computing, but they are a necessity when moving to cloud services. I hope we can avoid turning this into a on-prem vs cloud flame fest, there are merits and valid use-cases for both still.

I'll start. I am a manager of operations for a large company that is currently in phase 2 of a full migration off of 'legacy' virtualization/SAN/NAS/datacenter infrastructure to AWS. We are leveraging about 65% of AWS available services, from EC2, S3, Lambda/Api Gateway(serverless), dynamodb, autoscaling,etc, etc.

Where we are at:

  • Creating our AWS base environments via Cloudformation scripts(VPC, Gateways, DBs, Bastion hosts, DNS, etc)
  • Completely replaced our VMware horizon/view implementation with AWS Workspaces, this one was easy
  • Migrated 45 TB of image data(600mil+files) using Lambda/S3/DynamoDB(for image metadata). This was a big project and very challenging, but extremely fun and gratifying to complete.
  • Written our own Ruby based orchestration system utilizing Chef, Terraform, Consul, RabbitMQ, to allow our developers and QA teams to spin up complete production environments for testing or devving, along with scheduling to turn them off after hours or when done. This has been a multi year project and still ongoing.
  • Forklifted 20% of our static non config mgmt friendly VMware infrastructure so far, going for 100, utilizing AWS vm-import method, this works surprisingly good.
  • Adopted a DevOps mentality. We had big walls between Ops and Dev, and our Ops team actually were the first ones to start getting involved with all of these projects. Getting some of the traditional devs to embrace these tools has been the challenge, but we are getting there.
  • Migrated our Email system, Zimbra(blech), to AWS, built to be fault-tolerate(multi-AZ hot, multi-region cold). This was a huge challenge and about 20TB of data
  • Converted about 75% of our on-prem applications to be deployable via Chef through our orchestration systems. This has allowed us to quickly recreate for disaster recovery or new deploys.
  • Adopted a Blue/Green code deploy mentality. We can spin up a new cluster of production systems, smoke test, QA verify during Day, click a button when verified and using HAproxy and Consul instantly swap to new code base and retire old code, and revert back as fast if bugs are found.
  • plenty more going on here but thats a start

One of our biggest challenges has been changing the mentality of our IT dept. We had a few folks adamantly against this technology, I was one about 4 years ago, but have come around in a huge way. I absolutely love playing with all the new code/technology/services and I think in the last 2 years of this project I have learned and grown more as an IT person that I had in the previous 5 years of work combined.

Who else out there is working or moving to this space?
 
To add another slant to the question, I am curious to know what steps can be taken early on to avoid problems in the future and to avoid the many hours of work that were described by the OP. What is the best way to start from scratch?
 
To add another slant to the question, I am curious to know what steps can be taken early on to avoid problems in the future and to avoid the many hours of work that were described by the OP. What is the best way to start from scratch?

There truly isn't a one size fits all when it comes to this type of migration. I think covering the bases of CI/CD are absolutely necessary. Any new deploys of systems and software should be scriptable and repeatable. We spent about a year before our production move to AWS just creating Chef cookbooks around our current environment so we could deploy to any platform we required and bring systems back on-prem if necessary or in event of DR.

We also, again, painstakingly(there is no easy route here), we started mapping our infrastructure, software, systems, services, so we could design our security/VPCs/peering/etc properly around it. There was a TON of ground work trial and error during our POC'ing phases. Honestly, the config management piece is probably one of the most important, regardless of on-prem or cloud platform to have nailed down. We went from taking a day or two in some cases to be able to provide lab environments to our QA/Dev teams to down to about 10 minutes with a click of a button on their end. I probably could talk about it for hours and what we have had to do and are still doing to improve this process. It's reminiscent of when we were moving from the physical world to the virtual world about 10-12 years ago. I am sure it will get easier as time goes on, but right now it's still quite pioneer land. That in itself has made it fun.
 
I'll just leave this here

die-cut-stickers.png
 
We are looking at a hybrid approach with high traffic high viability stuff in AWS/Azure/hosted SaaS and everything else on prem.
 
And the ancient and mighty pendulum of mainframe or terminal swings back and forth.

Every time our company experiments with cloud we regret it for anything other than temporary DEV environments, although AWS is doing better for us recently.
 
Last edited:
I agree with the sentiments expressed by FLECOM. If one has enough hard drives and an NAS, why would one want to use the cloud and trust data to someone else's computer.
 
Who else out there is working or moving to this space?

I've been in this space for a few years. I understand the need for the cloud and host my personal websites in the cloud, for a long time. I'm not a believer in turning over all control to a cloud provider. I've experienced a few failures with large suppliers of the technology. If you have redundancy, then I think it can be great for elastic loads.

If one has enough hard drives and an NAS, why would one want to use the cloud and trust data to someone else's computer.

Offsite storage is one reason.
 
Last edited:
I think some of the folks in this thread don't understand what we are talking about. This isn't about your average users use of cloud, this is enterprise level computing.
 
like dropbox running away from AWS?

The Epic Story of Dropbox’s Exodus From the Amazon Cloud Empire

sorry but I don't really buy the "enterprise" cloud... the cloud is great for companies that don't have the talent to run their own infrastructure

For every one there are 100's others running to it. It's inevitable regardless of your beliefs. Hence why AWS is making profits for the first times over the last year, and their growth estimates are through the roof.
It's not for everyone, for sure. IMO it actually takes more planning and talent by far to utilize properly versus on-prem/colo'd infrastructure. If you want to use it just like your on-prem ESXi/HyperV/Xen, you are absolutely doing it wrong. Just look at all of the big SAN providers and VMware in particulars bleak outlooks. Writing has been on the wall and now they are all rushing to catch up. Those folks will be dinosaurs in 10 years if there is no adaptation. I refuse to sit idly by and become one.
But i digress. Those who are vehemently against it, probably always will be. While the salaries of VM/San/legacy *nix admins keep going down, DevOps salaries are flying high due to high demand and lack of talent. DevOps of course isn't relegated to Enterprise cloud space only, but its definitely vastly geared towards the agility given with automation and the cloud and not having to have the capex spent ahead of time to "guess" your usage for the next 3-4 years.

In any case i was hoping this forum would end up being a resource for those doing the same thing, not a soapbox for ppl who "hate" the "cloud" or who do not understand it.
 
Last edited:
I work in higher ed and support around 100k users give or take depending on how you count users. Problem is most vended software doesn't support the cloud unless the only option is for them to host it (most likely badly) in the "cloud" we are investigation moving some things to AWS/Azure but at this point we are calling our strategy "cloud smart" and don't see a way to move 90% of what we run into the cloud. Mostly because it doesn't scale and licensing prevents us for having multiple servers running in a cluster. Higher Ed is a joy let me tell you.

My view at this point is that its great if you want to run a SaaS or PaaS company. If you are a large enterprise it has it place but moving everything in impractical for multiple reasons.

That's not even factoring our multiple Petabytes of storage that all the application use.
 
They dont? Companies dont have telecom departments that work with vendors to provide employees with cell phones?

Huh - who knew.

So a company would rather buy cell phone service from a provider than build their own radio towers? Wow, it's the same analogy I made earlier.
 
So a company would rather buy cell phone service from a provider than build their own radio towers? Wow, it's the same analogy I made earlier.

Is comparing a cell tower to a server the same thing? No it isn't. I get it - some people would rather outsource all their IT. But doing it in house and doing it right is cheaper.
 
Is comparing a cell tower to a server the same thing? No it isn't. I get it - some people would rather outsource all their IT. But doing it in house and doing it right is cheaper.
Again, I don't think people understand Enterpise Cloud computing. It isn't outsourcing all of their IT, or necessarily any of it. We haven't let any standard infra guys go through our moves, they have adapted to build our cloud infrastructure to support our CI/CD initiatives, which in turn has also forced our Devs to clean up their code. It has forced a new way of thinking for all of our IT divisions. And now we don't have to worry about a guesstimated capex spend on our resources needed. We can burst as much as we want to the cloud, and bring it back in when its low traffic. There is so much more to it. The cell tower analogy fits ideally to how we are approaching Enterprise Cloud.
I don't even...
 
There really is a lack of understanding of what cloud computing is. It isn't just shipping your servers/applications off to some cloud provider and then firing half of your IT staff. Now there are many applications that are not suited for the cloud however as these software vendors are creating new or upgrading existing applications they are doing so with the cloud in mind. These new applications are being built for the cloud and can utilize the many benefits of the cloud. A good example is the ability to dynamically scale up and scale down based on demand. Think about a web front-end for a company's online store. Now maybe for 9 months out of the year they only need X number of web front-ends to handle the traffic however for the 3 months around Christmas they might need double that number due to the Christmas shopping season. They cloud provides them the ability to do this. If the company hosted the application on premise then they would have to invest in an infrastructure that could meet the peak load demands of that application even though that peak load is only for 3 months out of the year. Since most infrastructure is on a 5 year life cycle you rinse and repeat this every 5 years. Companies are wanting to shift that capex expense over to opex and the cloud provides them a way to do this. IT jobs are not going away. IT professionals just need to shift their skill set to DevOps as that is where the jobs are going. If you are in IT and have been for a few years then you know this always happens and you have to be willing to learn and adapt.
 
DEVOPS IN THE CLOUD

Just the newest and latest buzzwords from sales people trying to earn a living.

The cloud has its place - agreed for specific situations. Its not always cheaper or more secure or more anything else. Thats all I am saying.
 
We went from taking a day or two in some cases to be able to provide lab environments to our QA/Dev teams to down to about 10 minutes with a click of a button on their end.

I work for a VAR focusing on the datacenter infrastructure side of things (Storage/Networking/Compute) however I have been shifting more towards the automation/orchestration side. This is where we are seeing companies start their foray into automation/orchestration and even into the cloud. QA/Dev teams can eat away at the Ops team's time with so many constant requests of spinning up new QA/Dev environments or decommissioning old environments. I know in some cases it could take weeks for the Ops team to get an environment stood up for QA/Dev just due to workload. Giving QA/Dev teams the ability to spin up/down their own systems has drastically reduced wait times as well as reduced the load on the Ops teams. To me this is one of the best ways to get teams on board with the whole automation/orchestration idea.
 
While I'm not direct involved as ERP-guy the company I'm working for is partially moving to the cloud. Public and private clouds. On the public cloud there are use cases around peak computing needs here we just quick scale up and down computing resources to solve given tasks in R&D. Private cloud more when it goes to privacy of confidential data.

As such we drive a dual strategy, having our own DCs and add cloud-based solution and link them. My guess it the trend to the cloud will increase; as long legal and compliance are ok. Same reason as above to save capes.
 
... And sometimes I rent an EC2 for my own entertainment and fold on it. When I want to have some bigger CPU ...
 
We run data intensive crunching tasks probably every quarter. We just spin up some EC2 instances and let it go with it. Much cheaper than hosting something of equal speed internally and maintaining it.
 
We run data intensive crunching tasks probably every quarter. We just spin up some EC2 instances and let it go with it. Much cheaper than hosting something of equal speed internally and maintaining it.

Sounds like a great idea for quarterly accounting and reporting.
 
like dropbox running away from AWS?

The Epic Story of Dropbox’s Exodus From the Amazon Cloud Empire

sorry but I don't really buy the "enterprise" cloud... the cloud is great for companies that don't have the talent to run their own infrastructure

You could build your own cell phone towers, too, but since I'm not in the business of running radio towers I think I'll just pay for the utility service.

Exactly. If business A's sole purpose is to produce and sell a doohickey or offer services B,C, and D, then the business could care less about replacing a hard drive or conducting any sort of hardware PM's, or refreshing hardware once every 5 years. All the business cares about is having the proper tools to carry out its day to day functions in the same manner as you'd expect your cell phone to "just work." As the IT guy, you'd argue that you need this hardware and infrastructure in place to keep the world spinning but what Amazon has done is put a price on storage AND compute power in one nice simple package that is easy to present to management and easy to manage by the IT department. Now all of the normal headache of IT is invisible to all parties - no more botched updates, downtime, swapping hard drives, maintaining vast amounts of DC space and rack space - everything is one simple easy OPEX cost for X amount of storage and Y amount of compute. As an added bonus, you get to take advantage of the economy of scale which translates to reduced prices AND you only pay for what you use - no more wasted CPU cycles on underutilized servers or storage arrays.

What dropbox did was smart - build a platform on top of someone else's infrastructure to get off the ground and then move to your own infrastructure once you have an established, stable product. In Dropbox's case, maintaining their own infrastructure makes sense as their entire business model essentially hinges on renting small portions of storage and network bandwidth to customers. Amazon was an unnecessary man in the middle shaving pennies off every megabyte they were selling.

The AWS or "cloud" model obviously does not work for everybody. If you have data or IP that's highly confidential and don't trust anybody else to store it, then it makes sense to bear the cost of standing up your own infrastructure as you've determined the risk of trusting it to anybody else as too great to the company's bottom line. When it comes to HIPAA or PCI compliance, I have no idea how that would affect this model but for all other cases, it's extremely hard to argue the "cloud's" value when your core business, in essence, does not rely on selling IT infrastructure such as Dropbox.
 
If you want, I'll rent you my extra capacity on my HTPC for your folding. ;-)

<Troll> How do I get in on this offer? I'm guessing you'll charge a significantly lower price since you don't have dual redundant power, connectivity, secure environment, etc. Yes? :-P </troll>

I always giggle when I see the comments "no cloud, someone else's computer." The brilliant thing about cloud is I don't have to pay for depreciation on that hardware. I use it when I need it and pay what I want for it. Spot pricing is very handy. If I only need 30 minutes of HPC, I only pay 30 minutes of it (or an hour depending on units) and don't have to keep that hardware around after I'm done or anticipate needing it beforehand. That's the main benefit for us - we can react nearly instantly where it took months to procure before.
 
<Troll> How do I get in on this offer? I'm guessing you'll charge a significantly lower price since you don't have dual redundant power, connectivity, secure environment, etc. Yes? :-P </troll>

I always giggle when I see the comments "no cloud, someone else's computer." The brilliant thing about cloud is I don't have to pay for depreciation on that hardware. I use it when I need it and pay what I want for it. Spot pricing is very handy. If I only need 30 minutes of HPC, I only pay 30 minutes of it (or an hour depending on units) and don't have to keep that hardware around after I'm done or anticipate needing it beforehand. That's the main benefit for us - we can react nearly instantly where it took months to procure before.

If you need 30 of HPC, price for you is £0.01.
 
Thanks for starting this group -- hadn't noticed it until now. After six years at Valve and 18 months at Amazon on AWS, I joined Oracle to work on the IaaS offering, which was announced at Oracle Open World last week. It's an exciting time!
Look forward to learning more about Oracles offerings. We have been getting cold calls from them lately on the PaaS/IaaS offerings, curious how they will compete.
 
I am not trying to be an ass and not trolling. I am honestly curious why you, mikeblas, are excited to work for Oracle?

I know every company has its bad press and hatred, such as EA in the gaming world. However, for the last 19 years I have had to deal with Oracle the company and I can not imagine anyone "wanting" to work for that company. I deal with Oracle and their support of various products from UNIX/Linux/databases every day. Myself and my coworkers agree that we would rather eat four McDonald's BigMac's followed by a big-block of ExLax then hop on a trans-continental flight, than deal with Oracle.

Now, back to the topic. We have attempted migrating our infrastructure to AWS, even just parts of it, and found that the cost was astronomical compared to hosting it ourselves. Also, as others have mentioned, our customers do not want their data "touching" other customers data. Financial data on AWS is pretty much a no no for the company I work for. We deal in petabytes upon petabytes of financial data and AWS just could not deal with the data. AWS works for smaller companies with less data and a "spin up, spin down" work flow. If you run batch processing 24/7 you will go broke with AWS.
 
Last edited:
I am not trying to be an ass and not trolling. I am honestly curious why you, mikeblas, are excited to work for Oracle?

Well, there's about 135 thousand people at Oracle. It can't be all bad, right?

Large companies have microcosms. When I was at AWS, I was beyond miserable. I thought everything in the New York Times article was totally true -- and then some. But there were people I met and knew who were completely hppy. As long as their group stayed together and one manager or director was in control giving them a healthy environment, they were fine. These little microcosms were great ... until they popped and got reorged or merged or moved, or re-purposed.

I'm sure that, like Microsoft or Amazon or any other big company, there are really cool places in Oracle to work, and places that are not so much fun, even though the work is pretty much the same.

I know every company has its bad press and hatred, such as EA in the gaming world. However, for the last 19 years I have had to deal with Oracle the company and I can not imagine anyone "wanting" to work for that company. I deal with Oracle and their support of various products from UNIX/Linux/databases every day. Myself and my coworkers agree that we would rather eat four McDonald's BigMac's followed by a big-block of ExLax then hop on a trans-continental flight, than deal with Oracle.

You must be pretty young. When you get older, you'll realize that taking a shit is a privilege, not a right.

I think I was the 12th guy hired at Oracle's Seattle office. Maybe the 15th. We're over 350 people now, and we've got unlimited budget to hire. We've started writing a cloud service from scratch; we're building our own tools, our own data centers, our own internal services, our own external services, monitoring, back-ends. We're inventing our own processes, and building our own culture. We're a giant culture within Oracle by ourselves, doing things our way.

It's an absolute blast. I don't know why someone would not want to work there.

Now, back to the topic. We have attempted migrating our infrastructure to AWS, even just parts of it, and found that the cost was astronomical compared to hosting it ourselves. Also, as others have mentioned, our customers do not want their data "touching" other customers data. Financial data on AWS is pretty much a no no for the company I work for. We deal in petabytes upon petabytes of financial data and AWS just could not deal with the data. AWS works for smaller companies with less data and a "spin up, spin down" work flow. If you run batch processing 24/7 you will go broke with AWS.

There's lots of reasons to use the cloud, and lots of reasons not to. I don't think that those reasons correlate to company size too much; it's more about the specific project or application. If you're doing lots of large data consistently, you might be better off running your own data center because the storage amortizes better. If you have bursty workloads (you need 6 hosts sometimes, and 60 other times) or if you have bursty data needs (lots of online storage sometimes, push it to colder storage the rest of the time) then there are value propositions for you. Those applications exist in companies of every size. There are lots of very large companies, some with large batch processing needs, that successfully use cloud services. There are small companies with data processing needs that are a terrible fit for most cloud services.

It's really about the application, not company size.
 
I've worked as the Infrastructure Manager at both large companies ($7B) and small companies ($600M), and in both cases the cost of the cloud was drastically higher than on-prem infrastructure for our regular workloads.

I've used the cloud a lot for dev/test, especially for data processing, cold data storage and web apps, but in general I don't understand how companies are going to AWS and saving money. I've implemented open source tools and software on top of owned infrastructure for significantly less and achieved similar benefits. AWS/Azure (aside from pricing sweetheart deals for the first couple years) are consistently 15-30% more expensive
 
Back
Top