Working in IT 3.0: xp_cmdshell

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Working in IT 3.0: xp_cmdshell

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

BTW when I said lifecycle management, I was talking about the Cloudforms product itself having poor lifecycle management tools, AKA you can't export/import some things, only do backup/restores and some of the configuration is manual GUI clicking.

Generally speaking, a private cloud is only concerned with deploying the OS and does very minor configuration (configuring hostname, IP, possible root keys / passwords). It doesn't have anything to do with actual application deployments. In all these products, you need a different solution (or custom add-on configuration) to manage application stuff.

Tab8715 posted:

How come? Aren't they all essentially accomplishing the same goal?

I'm coming from a "pet" background and I'm a little confused when I hear we need to make cattle. When I hear an application needs to be scaled I think of increasing vCPUs, RAM, I/O or splitting database, services, to run on multiple servers.

Well, it's like anything else - Oracle skills don't all transfer over to Postgres which don't transfer to mysql, even though they are all relational databases. Same deal with Nagios / Openview / Icinga. General knowledge and ideas transfer, but expert / SME stuff doesn't since it's all application specific.

Bhodi fucked around with this message at 20:18 on Feb 19, 2015

# ? Feb 19, 2015 20:11

Adbot: ADBOT LOVES YOU

# ? May 25, 2024 05:09

evol262: Nov 30, 2010; #!/usr/bin/perl

Bhodi posted:

BTW when I said lifecycle management, I was talking about the Cloudforms product itself having poor lifecycle management tools, AKA you can't export/import things, only do backup/restores and some of the configuration is manual GUI clicking.

Cloudforms/ManageIQ itself has issues since most of the development effort has gone towards open-sourcing it and trying to get knowledge from the original owners. It's improving, but...

Bhodi posted:

Generally speaking, a private cloud is only concerned with deploying the OS and does very minor configuration (configuring hostname, IP, possible root keys / passwords). It doesn't have anything to do with actual application deployments. In all these products, you need a different solution (or custom add-on configuration) to manage application stuff.

Any cloud is only really concerned with deploying the bare bits. Including most PaaS stuff. Actual application deployments can be handled any way you want, but common ways are through software lifecycle management tools, configuration management and private/public package repositories, or creating images which already have the software you want, using config management tools to pump in what's necessary to get it running. Especially if the software you want isn't available in repos and you can't create one for whatever reason.

# ? Feb 19, 2015 20:18

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

And all this poo poo is a complicated ecosystem and the amount of words being thrown around on the last two pages should give a hint of why Amazon's EC2 and Azure are eating everyone's lunch. Only the largest companies can afford to hire people to untangle and manage it all. And you get serious economies of scale by going to a specialist company.

Fortunately for people like me, there are a lot of business reasons that require data to stay in-house but C-levels still want the convenience of cloud provisioning - and so this entire market sprang up to try and fill that business need. It's still new though, so it's morphing wildly quarter to quarter and nothing has really settled out yet, and a lot of the software is still bleeding edge.

Edit: it's also INCREDIBLY lucrative for companies who are eager to promote their (universally lovely) private cloud products and services.

Bhodi fucked around with this message at 20:28 on Feb 19, 2015

# ? Feb 19, 2015 20:24

evol262: Nov 30, 2010; #!/usr/bin/perl

Bhodi posted:

And all this poo poo is a complicated ecosystem and the amount of words being thrown around on the last two pages should give a hint of why Amazon's EC2 and Azure are eating everyone's lunch. Only the largest companies can afford to hire people to untangle and manage it all. And you get serious economies of scale by going to a specialist company.

Fortunately for people like me, there are a lot of business reasons that require data to stay in-house but C-levels still want the convenience of cloud provisioning - and so this entire market sprang up to try and fill that business need. It's still new though, so it's morphing wildly quarter to quarter and nothing has really settled out yet, and a lot of the software is still bleeding edge.

You need to do all the same poo poo on EC2 and Azure, by the way. They're eating everyone's lunch because a lot of startups in seed/early series funding don't have the budget or need for buying a bunch of physical servers, along with managing dead drives, cabling, cooling, power, etc. They can get machines on Azure/AWS for peanuts (comparatively), and many just stay there instead of getting their own private infrastructure.

The complicated bits of "private cloud" (SDN in particular) haven't even come up on the last two pages.

# ? Feb 19, 2015 20:28

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

That's so incredibly not true I don't even know how to phrase a response. I think you're living in the redhat bubble and so you haven't seen the enormous shift of people going "gently caress it, just set up a bunch of EC2 instances and VPC it into our network". It's so dead easy it's not even comparable to trying to get openstack or cloudforms to actually work, it's not even about cost anymore, it's about spinup time and being agile and poo poo

# ? Feb 19, 2015 20:30

MagnumOpus: Dec 7, 2006

Bhodi posted:

Fortunately for people like me, there are a lot of business reasons that require data to stay in-house but C-levels still want the convenience of cloud provisioning - and so this entire market sprang up to try and fill that business need. It's still new though, so it's morphing wildly quarter to quarter and nothing has really settled out yet, and a lot of the software is still bleeding edge.

This is the real reason to just jump in feet first. The pace of development is furious and the market is a bear pit but just picking one combination of platform/tools and learning it will give you enough of a semantic understanding to make sense of what's occurring. Disruptive developments (e.g. Docker) are coming along with a frequency that means just about anything can happen in a few quarters. No matter how it all shakes out the future is going to look very different from 2010 and earlier and you're going to want to have augmented your skillset with something virty, cloudy, or devopsy.

# ? Feb 19, 2015 20:33

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

Seriously EC2 was growing at 400%/yr 5 years ago, even while tons of articles were being written about how expensive it was if you left the servers on all the time. People didn't care about the cost. At all. People still don't, not really.

# ? Feb 19, 2015 20:34

evol262: Nov 30, 2010; #!/usr/bin/perl

Bhodi posted:

That's so incredibly not true I don't even know how to phrase a response. I think you're living in the redhat bubble and so you haven't seen the enormous shift of people going "gently caress it, just set up a bunch of EC2 instances and VPC it into our network". It's so dead easy it's not even comparable to trying to get openstack or cloudforms to actually work, it's not even about cost anymore, it's about spinup time and being agile and poo poo

That's exactly the market case for any private cloud solution.

I've only been at Red Hat for two years, contrary to the opinion of every person in SH/SC. CERN's talk which famously references Pets v. Cattle came out before I got to Red Hat. The idea that it's so "bleeding edge" the terms are undefinable is absurd.

You can set up a bunch of instances on AWS and VPC it into your network and pretend that it's basically an ancillary vCenter and run traditional virt poo poo on it if you want to, but we aren't talking about that. Making horizontally scalable applications is still AWS best practice and probably always will be, and all the same arguments about pets v cattle and swarm management also apply on AWS. A lot of the software for doing this was developed specifically for handling spikes on public cloud infrastructure.

The difference between "running machines on someone else's infrastructure" and ~cloud~ is what you're missing. You're doing the first. It doesn't sound like you're doing the second. Not to make a "True Scotsman" argument, but there are very defined work patterns and best practices.

# ? Feb 19, 2015 20:37

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

We're splitting hairs at this point but I understand the point that you're making. But I think you're conflating the "Cloud Philosophy" - which is developing applications that scale sideways and with elastic demand - and "Cloud Products", which are applications that provide the foundation (and APIs) to allow you to that from within the application as well as lowering the barrier to provisioning down to clicking a button in a GUI. I'm talking about the latter. It may enable the former, but it's not the same thing at all.

# ? Feb 19, 2015 20:41

syg: Mar 9, 2012

So my company has decided to give us each $5k per year to spend on whatever training or conference related things we want. I have no certifications now as every time I sign up for a virtual classroom (Security+, CCNA) I don't have enough time to actually finish it before it runs out. I'm not even sure I want or need certifications per se. I'm the systems architect for a large company, working mostly in networking(cisco), vmware and storage. I'd say my network skills are around CCNP level and my vmware around VCP but I'm not sure how I'd ever have the time to study all the components of these exams and write them.

How important are certs like this for someone in my role? I don't anticipate hopping around the job market much at this point, I'm 30+ with a family and unless something bad happens with this company I'm not heading anywhere except possibly into management in the future. Am I better off using this money to go to an IT conference or two (vmworld/cisco live) ?

# ? Feb 19, 2015 20:42

MagnumOpus: Dec 7, 2006

It's getting a little hot in here let's talk about something less contentious like the definition of DevOps.

# ? Feb 19, 2015 20:44

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

More certs = more better

No, a conference is not going to forward your career (unless you are meeting people who will happen to attend the same conference and are really spending $5k on networking)

Edit: I'd be scared of a "systems architect" who only wanted to get a Security+ or CCNA though, considering those are entry level certs. Maybe aim higher?

Bhodi fucked around with this message at 20:46 on Feb 19, 2015

# ? Feb 19, 2015 20:44

evol262: Nov 30, 2010; #!/usr/bin/perl

Bhodi posted:

We're splitting hairs at this point but I understand the point that you're making. But I think you're conflating the "Cloud Philosophy" - which is developing applications that scale sideways and with elastic demand - and "Cloud Products", which are applications that provide the foundation (and APIs) to allow you to that as well as lowering the barrier to provisioning down to clicking a button in a GUI. I'm talking about the latter. It may enable the former, but it's not the same thing at all.

Part of the "cloud philosophy" is also building infrastructure services which scales sideways and with elastic demand. Which means building infrastructure which can scale with the "click of a button in a GUI" (despite the overweening popularity of API-driven creation, from AWS's API, Rackspace's API, Cloudformations, or other), and that means that your images should be doing a lot more than getting a "SSH key and hostname" then getting passed on to something else.

I'm not going to dispute that getting Openstack or Eucalyptus or any other IaaS product running internally is a nightmare for most people, despite improvements. But the number of people I talk to or customer cases we get from clients who have wanted the "cloud" then don't know what to do with it is incredible, and it's a good indicator that people who are using the "cloud" (of any kind) just for provisioning VMs by clicking a button are probably better served by wrapping VMware's .NET API or using Cloudforms or something to do it on traditional virt, not "cloud" at all.

The "cloud philosphy" and "cloud products" are inseparable, in that you cannot possibly make good use of the latter without understanding the former, since the majority of "cloud products" (including Amazon's) do not support FT or HA or the other moving parts which businesses traditionally rely on. In order to get a reliable application from any cloud provider, public or private (ease of provisioning a private cloud notwithstanding), you almost need to follow the "cloud philosphy" or risk outages which are unacceptable to most businesses, and would have been completely unacceptable to the kind of businesses I worked for before I came to Red Hat.

Companies that don't do this end up bringing up a bunch of pets in one AWS AZ and making GBS threads themselves and complaining about how the "cloud" is unreliable when that AZ goes down or their instance disappears (which still happens).

evol262 fucked around with this message at 20:52 on Feb 19, 2015

# ? Feb 19, 2015 20:49

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Ooh, I like this controversy! Here's some responses to things.

Japanese Dating Sim posted:

Wait, what? Really? And yeah, I did read all of the rest of your post; I've just never picked up any hint of that at my position (which just occasionally requires tinkering with our vSphere). That's kind of a bummer to hear.

It's not going to happen tomorrow. The landscape is going to look drastically different in ten years, that's all, just like how ten years ago people couldn't have fathomed the idea of fully virtualized infrastructures running on commoditized SAN storage. I wouldn't generally recommend VMware to someone looking to further their career prospects because by the time someone's put the years of effort into becoming an expert in the technology, the market for it will look much dimmer.

Trends via Indeed: http://www.indeed.com/jobtrends?q=vmware&l=

Tab8715 posted:

How do you feel about health of future of hybrid clouds?

Hadoop might be 2011's buzzword, but if it taught us anything, it's that the future of computing is putting the compute power where the data lives. For situations where there's huge amounts of on-premises data that don't make sense to migrate, or where data protection requirements make this a risky proposition, there will continue to be on-premises systems.

I don't think highly of "hybrid cloud" solutions in the sense that we're never going to see systems seamlessly moving between on-premises datacenters and the cloud the way that we see virtual machines flying between hosts in a DRS cluster. However, I do think that most businesses will run their systems on a mixture of public and private clouds.

orange sky posted:

Private cloud is an on-prem cloud, but it could be two on prem data centers linked. See for example Azure Pack (not Azure, Azure Pack). It brings elasticity and self service to your Data Center, if you wish to have those.

I know you didn't ask for my opinion, but I'll give it to you: Public clouds will be used a lot by startups and companies/individuals like that but the big companies who can afford the CAPEX for on prem solutions will always prefer that. The thing is, people might start using public clouds a lot because it's cheaper, but sooner or later there'll be a huge hack on Microsoft/Amazon's public cloud servers and we can all foresee what that will bring.

That's why I think that Hybrid Cloud has it's space. Hybrid Clouds will be used, but carefully. You can use online storage for backups, DRaaS, something like that, but you try to limit as much as possible having everything outside your company. That's just my 2 cents though.

Also, SaaS. I think that's going to be huge. IaaS, not for the big boys, they'll still spend a lot of money on equipment.

I'm going to back it up here and mention two things that are tremendously driving the growth of public cloud:

Software as a service: as people migrate historically internal applications to hosted versions, those applications disappear from the datacenter. The jobs running those applications move to other companies who are overwhelmingly putting their greenfield applications in public cloud. IaaS is not directly used by many midsize organizations, but basically every outsourced SaaS application drives the growth of IaaS behind it.
Remote work: Remote hiring is moneyball for tech. Whatever savings you can scrounge up by moving applications on-premises is going to be offset when you have to pay $180k for the people to manage them in a major metro area. You can get a lot done by hiring where the cost of living is lower.

For companies that are so large that it isn't a challenge to staff their datacenters with very specialized and commoditized roles in order to run at the efficiency of the public cloud players, people will absolutely continue to do run their own infrastructures until it makes sense not to.

MagnumOpus posted:

I'm not sure I agree that private clouds are irrelevant. There are an awful lot of enterprises with ERP architectures that won't or can't move everything into a public cloud. For example my current client is an ivy league college and while they are trying to move as much as they can into AWS, they will never get all of their data sources migrated off of their current on-premises systems and can hope for a hybrid cloud at best. There are as well a great many such enterprises with private clouds who will not even attempt the move to public offerings due to lingering fears about public cloud intrusions. I think it may be premature to call time of death for the private cloud vendors. The market segment probably won't grow that much but VMWare would not be the first relic to stick around well past its prime.

Having built a VMWare private cloud though I would not jump at the chance again. Between the various cloud implementations and the industry dialogue about DevOps we're in a sort of IT development singularity that I don't see settling down in 2015 at least, and lots of cool things are coming of it. I know I'd rather be focusing on that side of things, but I also know guys who are VMWare storage experts who love polishing the same skillset to increasingly impressive levels. I guess what I'm saying is you can pick your poison this year.

There are lots of companies running very large private clouds for various reasons, and I certainly don't want to imply that they're irrelevant for everyone. Walmart Labs is running 100,000 cores on OpenStack Havana. Pretty much every insurance company you can think of is running petabytes or exabytes of data through local processing systems that are largely running on some kind of private cloud. These things exist, but in a decade they will be a niche confined to certain industries, like how some companies have their own HVAC departments but most people outsource their maintenance contracts. You can point to tons of verticals like high-frequency trading where experience with InfiniBand and user-space networking stacks are absolutely critical to their day-to-day operations. Maybe "irrelevant" was the wrong word, but it's certainly not a paradigm that's going to apply to the majority of businesses and their applications.

go3 posted:

its another tool in the box and you're nuts if you think everyone is just chomping at the bit to jump onto AWS/Azure

The market isn't driven by what people are excited to adopt. It's driven by what people have to adopt in order to stay competitive. This is the very definition of disruptive technology. (Shoot me if I use that adjective again in this conversation.)

In-house IT departments can't compete with the elasticity and rapid turnaround of public cloud. Despite the enterprise messaging race being fought between Blackberry and Windows Mobile a few years ago, shadow IT and the rise of BYOD meant that iPhone and Android became the two major players for enterprise mobile technology. AWS and Azure are Bring Your Own Infrastructure. People are going to use the tools that make them most productive, not the tools that some soon-to-be-irrelevant technology manager decreed are The One True Standard.

# ? Feb 19, 2015 20:52

Thanks Ants: May 21, 2004; #essereFerrari

rafikki posted:

Goddamn that Lenovo debacle.
http://arstechnica.com/security/2015/02/lenovo-pcs-ship-with-man-in-the-middle-adware-that-breaks-https-connections/

How on earth did someone think that was acceptable?

# ? Feb 19, 2015 20:53

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

Honestly, I don't think anyone cares about this anymore except for you and me. E: and Misogynist!

evol262 posted:

I'm not going to dispute that getting Openstack or Eucalyptus or any other IaaS product running internally is a nightmare for most people, despite improvements. But the number of people I talk to or customer cases we get from clients who have wanted the "cloud" then don't know what to do with it is incredible, and it's a good indicator that people who are using the "cloud" (of any kind) just for provisioning VMs by clicking a button are probably better served by wrapping VMware's .NET API or using Cloudforms or something to do it on traditional virt, not "cloud" at all.

This may be absolutely true but it doesn't stop your (or anyone else's) sales guys from happily selling the product. I mean, I can speak to our use, and we are using the product not for it's API at an application level (because it's both bad and useless for elastic scaling in 3.0) but for it's ability to track what groups provisioned what and print out useful reports that we can give to VPs to provide chargeback finances, as well as providing a baked-in GUI we can offer to internal dev groups to easily provision hosts and do development on. That's something a wrapper around vcenter can't provide for us and so there's a useful business case for Cloudforms for us, even if getting it to work is like wrestling a bear.

Bhodi fucked around with this message at 20:58 on Feb 19, 2015

# ? Feb 19, 2015 20:55

syg: Mar 9, 2012

Bhodi posted:

More certs = more better

No, a conference is not going to forward your career (unless you are meeting people who will happen to attend the same conference and are really spending $5k on networking)

Edit: I'd be scared of a "systems architect" who only wanted to get a Security+ or CCNA though, considering those are entry level certs. Maybe aim higher?

Those two cert programs were from 6-7 years ago when I was a sysadmin. The lack of free time is still an issue now though. That's the other problem, because I didn't write CCNA yet I've got to take the time to study all the obsolete and rarely used protocols and interfaces I will never see in my life, before I can move on to CCNP and higher. VCP would be a bit more straight forward as I know I would learn some useful stuff from that program, but again the time constraints.

# ? Feb 19, 2015 20:56

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

I can only say that if you've got a networking focus, a CCIE looks pretty drat good and directly translates to cash. Only you can decide if you're far enough in your career "not to bother" with getting certified. And maybe find a more laid back job where you have a more time to do personal growth stuff?

# ? Feb 19, 2015 21:01

evol262: Nov 30, 2010; #!/usr/bin/perl

Bhodi posted:

Honestly, I don't think anyone cares about this anymore except for you and me.

Misogynist might care, I guess.

Bhodi posted:

This may be absolutely true but it doesn't stop your (or anyone else's) sales guys from happily selling the product. I mean, I can speak to our use, and we are using the product not for it's API at an application level (because it's both bad and useless for elastic scaling in 3.0) but for it's ability to track what groups provisioned what and print out useful reports that we can give to VPs to provide chargeback finances, as well as providing a baked-in GUI we can offer to internal dev groups to easily provision hosts and do development on. That's something a wrapper around vcenter can't provide for us and so there's a useful business case for Cloudforms for us, even if getting it to work is like wrestling a bear.

Because I work on both ends (RHEV and Openstack), I'm exposed to a lot of people who got sold the product and don't like it. There's a relatively large pushback internally against the Openstack hype train as being, well, a hype train, and that we should be trying to convince a lot of the customers who've decided they're ok getting off of VMware that they should be looking at RHEV instead, since it fits what they're already doing and lets them ease into the "cloud philosophy" bits as they get ready.

Not gonna defend Cloudforms/ManageIQ in any way, since I don't work on it and mostly don't work with it. Everything that I'm saying is from an Openstack perspective and Openstack business case.

Reporting and accounting (telemetry is the term we use) is super useful in Ceilometer and presumably Cloudforms as well. I'm pretty sure that some some VMware product also does this, but I get lost in the vAcronym soup.

# ? Feb 19, 2015 21:02

psydude: Apr 1, 2008

syg posted:

Those two cert programs were from 6-7 years ago when I was a sysadmin. The lack of free time is still an issue now though. That's the other problem, because I didn't write CCNA yet I've got to take the time to study all the obsolete and rarely used protocols and interfaces I will never see in my life, before I can move on to CCNP and higher. VCP would be a bit more straight forward as I know I would learn some useful stuff from that program, but again the time constraints.

If you work for a VAR, MSP, or any other company that sells your services (directly or indirectly), it's probably beneficial for you to suck it up and play the game. If you're just doing internal support and plan on staying put, then spending it on some courses may be a better use of your time. That being said, having a CCNA is an excellent hedge against unemployment.

# ? Feb 19, 2015 21:03

FISHMANPET: Mar 3, 2007; Sweet 'N Sour
Can't
Melt
Steel Beams

I also care because I just switched jobs and am now having an existential crisis that is basically boiling down to not knowing what I want to be when I grow up.

Also I'm now in a weird position of running a doggie day care. We have 1000 servers that we host and we want to treat them like cattle so we just give them patches and call it a day, but every one of those servers is a pet to someone else.

I certainly see the value in "cloud" and the new things it can do, but I don't see how it replaces a lot of the things we already do. But maybe my perspective is skewed since everything I work with is focused on managing people's interactions with their desktop than making ~apps~.

I've tried to puppetize a Samba server, I've tried to puppetize a CUPS server, both of them caused me to basically run into a situation where the answer was "puppet can't do that" and I was left holding the bag. I feel like someone's always going to want a print server and/or a file server, what are they doing in this glorious cloud future? Or in 10 years will I look at the people doing that with the same bewilderment that I do people right now in environments with 50 machines and no Active Directory?

# ? Feb 19, 2015 21:06

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

evol262 posted:

Reporting and accounting (telemetry is the term we use) is super useful in Ceilometer and presumably Cloudforms as well. I'm pretty sure that some some VMware product also does this, but I get lost in the vAcronym soup.

Sort of; v* applications are losing ground bigtime as they try and slew the entire company into the services role which vCenter was never really intended to provide. So now you have a company who is creating products that run on top of vCenter (vCloud director, vWhatever, vWhizbang) but that are inferior to products other companies have already created. The increasingly complicated licensing requirements, wildly changing APIs, and internal siloing and dissension isn't helping their case, since the products can't even offer tighter coupling into vCenter (because the vCenter group has told them internally to gently caress off and deal with published SOAP specs). There are still some cool tools coming out with vCenter 6.0 but all the offerings on top of it that VMware are offering are pretty lovely and have a very anemic uptake rate, because let's be honest, vCenter does what it does pretty damned well and for a lot of companies it's all they need.

VMWare's response to all this is fairly predictable; they're trying to force adoption by splitting off common and highly requested features (cluster to cluster migration of vms while on) into their own products or adding them into their "Cloud" offering instead of into vCenter

Bhodi fucked around with this message at 21:11 on Feb 19, 2015

# ? Feb 19, 2015 21:08

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

SubjectVerbObject posted:

All very interesting. I do business phone system support, and the trend for larger systems is virtualization. This is in part because everyone is doing it, and also because most of the big players don't want to be in the hardware game any more. What this means is that there are private cloud type offerings for large customers, which include either a complete hardware system with storage, network, server, etc, or a CD that installs everything needed on their existing infrastructure. For smaller customers, hosted solutions are becoming a thing, with IP phones on site, and everything else managed remotely.

My concern is that when I look for jobs, most of the time when I see 'admin the phone switch' requirements, those are secondary to windows admin, network admin and other main functions. I think this will get worse in the future, especially if this is just going to be another program running in the data center. I need to either specialize in some more specific functions (call routing, call center, IVR), or get more into the infrastructure side.

Very interesting that I have barely heard of AWS, but there are certs and training for it.

Future shock is here and job roles are getting more specialized as automation takes over the industry, and uptime of even cheap ISPs is more than good enough to facilitate moving business-critical services offsite. Generalist roles will be increasingly shifted off to managed service providers, consultancies, and pre-sales engineering gigs for the major players selling on-premises equipment.

I'm surprised that you've barely heard of AWS. Amazon is by far the world's largest hosting provider. Some metrics put them at double the size of runner-up OVH. Many players with tens or hundreds of thousands of server instances, like Netflix, run most of their infrastructure on AWS.

Docjowles posted:

Depends which of the 9 million definitions of cloud you choose. But to me, HA/FT is the opposite of cloud. Even if those VM's are hosted in someone else's data center. "Cloud computing" (again, to me) is IaaS. Massively scalable and elastic, relatively unreliable, extremely cheap.

Scale vs. management overhead is always a pendulum. Mainframes with gigantic, super-fast hard drives were expensive and difficult to scale, so we invented RAID as a means of consolidating independent disks into larger, faster volumes and scaled out our systems to lots of cheap x86 servers. These RAID arrays became hard to manage through central policy, so we started to aggregate them into SANs that virtualized the underlying volumes and made it easier to carve them up. These SANs got to be too expensive, so now we're moving to scale-out distributed storage platforms that are cheap but impossible to a layperson to debug. In the software realm, we kept making software slower and slower because hardware kept getting faster and faster, until one day we noticed Ruby on Rails was very slow and async, evented services weren't. We started the push to microservices and someday people will realize what a loving disaster it's going to be when they have to manage a legacy microservice where none of the original developers are left anymore. Technologies are constantly changing to align with the cost model as people find more efficient ways to distribute the money required to take something from an idea to a finished result.

I think we're very close to seeing people trying to build that kind of traditional monolithic reliability back into cloud computing systems. Platforms like Google Compute Engine and OpenStack now support live migration for pet VMs when the underlying hardware needs to undergo maintenance, just like in vSphere environments. Storage environments are starting to simplify and lose some of that AWS cruft like "your data isn't committed until it's in a snapshot."

# ? Feb 19, 2015 21:11

syg: Mar 9, 2012

psydude posted:

If you work for a VAR, MSP, or any other company that sells your services (directly or indirectly), it's probably beneficial for you to suck it up and play the game. If you're just doing internal support and plan on staying put, then spending it on some courses may be a better use of your time. That being said, having a CCNA is an excellent hedge against unemployment.

Not at a VAR so my company doesn't really care if I get any letters. The IT market in my city is god awful and there are lots of CCNAs without work or working the absolute lowest help desk job. In fact many of our help deskers here have CCNAs and don't know what a frame is. They all used brain dumps when they went through community college and never touched a network again. I almost feel like if you are in a senior generalist position you are better off with no certs and playing the "i don't do certs" card (as much of a pretentious cop-out that sounds) rather than having some low level ones that makes it look like you do play the cert game, just not very well.

Bhodi posted:

I can only say that if you've got a networking focus, a CCIE looks pretty drat good and directly translates to cash. Only you can decide if you're far enough in your career "not to bother" with getting certified. And maybe find a more laid back job where you have a more time to do personal growth stuff?

I'm about 50/50 networking/vmware-storage. I also have a lot of budgeting and project management duties at my current role so getting deep enough to warrant a CCIE wouldn't happen at this company.

I may be better off going for a VCP and a CCDA I guess since most of my networking is designing new builds.

# ? Feb 19, 2015 21:12

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

FISHMANPET posted:

I also care because I just switched jobs and am now having an existential crisis that is basically boiling down to not knowing what I want to be when I grow up.

Also I'm now in a weird position of running a doggie day care. We have 1000 servers that we host and we want to treat them like cattle so we just give them patches and call it a day, but every one of those servers is a pet to someone else.

I certainly see the value in "cloud" and the new things it can do, but I don't see how it replaces a lot of the things we already do. But maybe my perspective is skewed since everything I work with is focused on managing people's interactions with their desktop than making ~apps~.

I've tried to puppetize a Samba server, I've tried to puppetize a CUPS server, both of them caused me to basically run into a situation where the answer was "puppet can't do that" and I was left holding the bag. I feel like someone's always going to want a print server and/or a file server, what are they doing in this glorious cloud future? Or in 10 years will I look at the people doing that with the same bewilderment that I do people right now in environments with 50 machines and no Active Directory?

One day your department is going to be too slow deploying changes to a Samba server because of all the pets, and the department asking for it is going to say "you know what, gently caress it" and buy a Dropbox for Business subscription that lets them point and click to share with someone in another department. Then, you'll have one less pet. The next week, you'll have another less pet.

# ? Feb 19, 2015 21:12

Gucci Loafers: May 20, 2006; Ask yourself, do you really want to talk to pair of really nice gaudy shoes?

Thanks Ants posted:

How on earth did someone think that was acceptable?

I'm sure there was money involved...

Bhodi posted:

Honestly, I don't think anyone cares about this anymore except for you and me. E: and Misogynist!

While I'm largely "cattle" illiterate I find this discussion highly engaging.

syg posted:

So my company has decided to give us each $5k per year to spend on whatever training or conference related things we want. I have no certifications now as every time I sign up for a virtual classroom (Security+, CCNA) I don't have enough time to actually finish it before it runs out. I'm not even sure I want or need certifications per se. I'm the systems architect for a large company, working mostly in networking(cisco), vmware and storage. I'd say my network skills are around CCNP level and my vmware around VCP but I'm not sure how I'd ever have the time to study all the components of these exams and write them.

How important are certs like this for someone in my role? I don't anticipate hopping around the job market much at this point, I'm 30+ with a family and unless something bad happens with this company I'm not heading anywhere except possibly into management in the future. Am I better off using this money to go to an IT conference or two (vmworld/cisco live) ?

I've been to a few conferences and I guess they're okay at least when it comes to IBM Events. There's definitely a focus to assist issues with larger clients. Everyone was shocked when I told them I worked on IBMi without any virt. All of the labs were enormous in complexity which while cool not necessarily useful for my projects. On the other hand, I did meet some really cool techs and there's big networking benefit.

# ? Feb 19, 2015 21:13

Zaepho: Oct 31, 2013

One comment on the whole Public Vs Private cloud vs Virtualization vs etc debate.

I see a lot of calls for "Develop apps that scale horizontally" which is awesome. Until you realize that the overwhelming majority (aka virtually all of them) of Line of Business apps are developed out of house and simply don;t work that way. It's rare that the primary LoB applications are actually written in house UNLESS you are a SaaS company. Then by definition you are selling that service and are writing it in house because it didn't already exist somewhere.

At least that's been the way I've seen it from the corporate side and my consulting experience with companies from tons of different industries and all different sizes.

"Private cloud" is the buzzword, Hybrid cloud even more so for these guys. Private cloud means VMware/VCenter or Hyper-V/VMM&WAP if they're really advanced. Hybrid means putting some virtual servers in the public cloud that they treat 99% the same as their internal VMs.

Why is it this way? The businesses. You have a bank. They have a Teller system which is usually pretty well load balanced/HA'ed that they bought, an ATM backend that they bought, They have a core/batch processing system that they bought because it takes a team just to keep up with building the code to handle all the regulator changes, they have an Online banking system that they might have written (this is customer facing so usually more customized), they have a Wire Processing System, they have systems (yes plural) for interacting with the Fed, they have Image Capture/Item Processing systems. These are all important and are critical to the functioning of the business but none are particularly large systems and none of them have been created with elasticity in mind. Hell the vendor wants to charge you as many licenses as possible as much as possible so why should they be able to scale down?? trying to create any of this stuff i house opens you up to all sorts of regulatory issue and tons of gotchas. When I worked at a bank, if our interface to mastercard had less than 95% uptime for a particular day, we got charged $10,000. There's no room there for spending the time figuring out how to build the system and dealing with whatever bugs may come up causing outages. Not a place for Agile/Scrum methodologies there.

I should shut up and quit rambling here.

# ? Feb 19, 2015 21:17

Gucci Loafers: May 20, 2006; Ask yourself, do you really want to talk to pair of really nice gaudy shoes?

syg posted:

Not at a VAR so my company doesn't really care if I get any letters. The IT market in my city is god awful and there are lots of CCNAs without work or working the absolute lowest help desk job. In fact many of our help deskers here have CCNAs and don't know what a frame is. They all used brain dumps when they went through community college and never touched a network again. I almost feel like if you are in a senior generalist position you are better off with no certs and playing the "i don't do certs" card (as much of a pretentious cop-out that sounds) rather than having some low level ones that makes it look like you do play the cert game, just not very well.

Where are you located?

# ? Feb 19, 2015 21:18

evol262: Nov 30, 2010; #!/usr/bin/perl

Bhodi posted:

Sort of; v* applications are losing ground bigtime as they try and slew the entire company into the services role which vCenter was never really intended to provide. So now you have a company who is creating products that run on top of vCenter (vCloud director, vWhatever, vWhizbang) but that are inferior to products other companies have already created. The increasingly complicated licensing requirements, wildly changing APIs, and internal siloing and dissension isn't helping their case, since the products can't even offer tighter coupling into vCenter (because the vCenter group has told them internally to gently caress off and deal with published SOAP specs). There are still some cool tools coming out with vCenter 6.0 but all the offerings on top of it that VMware are offering are pretty lovely and have a very anemic uptake rate, because let's be honest, vCenter does what it does pretty damned well and for a lot of companies it's all they need.

I think it's actually vCloud Director that does telemetry, but complicated licensing, complete lack of understanding of their customer base and what they want, and internal messes with greenfield deployments and interop have always been the case, with the vSphere 5 licensing debacle (when it first came out and offered about half the memory entitlements) aptly demonstrated.

FISHMANPET posted:

I also care because I just switched jobs and am now having an existential crisis that is basically boiling down to not knowing what I want to be when I grow up.

Also I'm now in a weird position of running a doggie day care. We have 1000 servers that we host and we want to treat them like cattle so we just give them patches and call it a day, but every one of those servers is a pet to someone else.

Are you at Wells Fargo down on 2nd Ave?

FISHMANPET posted:

I certainly see the value in "cloud" and the new things it can do, but I don't see how it replaces a lot of the things we already do. But maybe my perspective is skewed since everything I work with is focused on managing people's interactions with their desktop than making ~apps~.

I've tried to puppetize a Samba server, I've tried to puppetize a CUPS server, both of them caused me to basically run into a situation where the answer was "puppet can't do that" and I was left holding the bag. I feel like someone's always going to want a print server and/or a file server, what are they doing in this glorious cloud future? Or in 10 years will I look at the people doing that with the same bewilderment that I do people right now in environments with 50 machines and no Active Directory?

Puppetizing file/print servers is kind of a no-win. Puppet (and augeas) are often lacking in these cases. Someone probably will always want a file server and a print server. In the glorious cloud future, everyone puts their files in spideroak or S3 buckets or whatever.

Really, no, people will need pets and traditional types of servers for quite a long time. Until someone convinces your company to move to files in the ~cloud~ with Google Docs or Spideroak Blue or some future WharghrblJS frontend that does it all for you and creates a Postscript to replicator driver that lives in the ~cloud~

There are no answers for these things, but the market is moving towards them becoming a limited, curated part of your environment with lower administrative overhead instead of having racks full of poo poo in the datacenter, so that instead of QE knowing that SQQER2U346 is running YourApp v7.1 but SQPZR9U282 is YourApp v7.1.0.1.2.3 and needing some admin to go deploy new versions on them, they get just get new versions of whatever they want. Or ideally, that checkins and checkouts automatically run integration suites and pass it along to QE's portal that says "press a button to bring up an instance with v7.1.0.1.2.4 and click the green checkbox if it passes", at which point your environment starts shutting down the old ones and rolling out the new ones.

# ? Feb 19, 2015 21:19

FISHMANPET: Mar 3, 2007; Sweet 'N Sour
Can't
Melt
Steel Beams

No I stayed at the U, just different department. Took a job that's 100% running SCCM because I like SCCM but now I'm not so sure about that anymore, and also wondering if this is a direction I should continue on, or pivot somewhere else.

# ? Feb 19, 2015 21:21

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Zaepho posted:

One comment on the whole Public Vs Private cloud vs Virtualization vs etc debate.

I see a lot of calls for "Develop apps that scale horizontally" which is awesome. Until you realize that the overwhelming majority (aka virtually all of them) of Line of Business apps are developed out of house and simply don;t work that way. It's rare that the primary LoB applications are actually written in house UNLESS you are a SaaS company. Then by definition you are selling that service and are writing it in house because it didn't already exist somewhere.

At least that's been the way I've seen it from the corporate side and my consulting experience with companies from tons of different industries and all different sizes.

"Private cloud" is the buzzword, Hybrid cloud even more so for these guys. Private cloud means VMware/VCenter or Hyper-V/VMM&WAP if they're really advanced. Hybrid means putting some virtual servers in the public cloud that they treat 99% the same as their internal VMs.

Why is it this way? The businesses. You have a bank. They have a Teller system which is usually pretty well load balanced/HA'ed that they bought, an ATM backend that they bought, They have a core/batch processing system that they bought because it takes a team just to keep up with building the code to handle all the regulator changes, they have an Online banking system that they might have written (this is customer facing so usually more customized), they have a Wire Processing System, they have systems (yes plural) for interacting with the Fed, they have Image Capture/Item Processing systems. These are all important and are critical to the functioning of the business but none are particularly large systems and none of them have been created with elasticity in mind. Hell the vendor wants to charge you as many licenses as possible as much as possible so why should they be able to scale down?? trying to create any of this stuff i house opens you up to all sorts of regulatory issue and tons of gotchas. When I worked at a bank, if our interface to mastercard had less than 95% uptime for a particular day, we got charged $10,000. There's no room there for spending the time figuring out how to build the system and dealing with whatever bugs may come up causing outages. Not a place for Agile/Scrum methodologies there.

I should shut up and quit rambling here.

Banks are not likely to run in the public cloud for a litany of other regulatory and legal reasons, so they're probably not the best example.

Most LoB apps don't scale horizontally, but most LoB apps don't scale at all, period, and don't need to. They're running on virtual machines in the datacenter today after being P2V'd off someone's Windows 2000 desktop, and their successors will continue to run on virtual machines in the cloud tomorrow. These apps are the next big growth area for cloud providers, and they know it. They're a big driver to build more stability into networking, and into the hypervisor platform where it makes sense (live migration, etc.).

# ? Feb 19, 2015 21:22

Gucci Loafers: May 20, 2006; Ask yourself, do you really want to talk to pair of really nice gaudy shoes?

Misogynist posted:

Ooh, I like this controversy! Here's some responses to things.

It's not going to happen tomorrow. The landscape is going to look drastically different in ten years, that's all, just like how ten years ago people couldn't have fathomed the idea of fully virtualized infrastructures running on commoditized SAN storage. I wouldn't generally recommend VMware to someone looking to further their career prospects because by the time someone's put the years of effort into becoming an expert in the technology, the market for it will look much dimmer.

Trends via Indeed: http://www.indeed.com/jobtrends?q=vmware&l=

That indeed graph is fascinating.

# ? Feb 19, 2015 21:23

evol262: Nov 30, 2010; #!/usr/bin/perl

Zaepho posted:

Why is it this way? The businesses. You have a bank. They have a Teller system which is usually pretty well load balanced/HA'ed that they bought, an ATM backend that they bought, They have a core/batch processing system that they bought because it takes a team just to keep up with building the code to handle all the regulator changes, they have an Online banking system that they might have written (this is customer facing so usually more customized), they have a Wire Processing System, they have systems (yes plural) for interacting with the Fed, they have Image Capture/Item Processing systems. These are all important and are critical to the functioning of the business but none are particularly large systems and none of them have been created with elasticity in mind. Hell the vendor wants to charge you as many licenses as possible as much as possible so why should they be able to scale down?? trying to create any of this stuff i house opens you up to all sorts of regulatory issue and tons of gotchas. When I worked at a bank, if our interface to mastercard had less than 95% uptime for a particular day, we got charged $10,000. There's no room there for spending the time figuring out how to build the system and dealing with whatever bugs may come up causing outages. Not a place for Agile/Scrum methodologies there.

FYI, I came out of banking, and very small (6 person) team managed 8k servers in retail locations, as well as the dev and QA systems that fed into the versions which ran on these. Along with the deployment infrastructure for getting major operation system upgrades (in-place, of course, with rollbacks) and application upgrades (same) out across 128k satellite links to Idaho and locations in Alaska that you'd need to fly a tech into to replace a drive.

They moved to Openstack not long after I left. It was easy for them, because I spent 9 months standardizing their build process and deployments, as well as software candidate promotion through dev->qa->prod, with A/B deployments and rollbacks. They went from 8k "pets" to 2k cattle in on-premises infrastructure and 8k machines in retail branches managed from the same config management manifests and the same repositories.

Devops can work extremely effectively in many, many environments.

# ? Feb 19, 2015 21:26

syg: Mar 9, 2012

Tab8715 posted:

Where are you located?

New Brunswick, Canada

# ? Feb 19, 2015 21:26

Bhodi: Dec 9, 2007; Oh, it's just a cat.; Pillbug

evol262 posted:

I think it's actually vCloud Director that does telemetry, but complicated licensing, complete lack of understanding of their customer base and what they want, and internal messes with greenfield deployments and interop have always been the case, with the vSphere 5 licensing debacle (when it first came out and offered about half the memory entitlements) aptly demonstrated.

Yeah, it's so funny, I feel like VMWare is a prime example of a company who hit one home run out of the park and has been living on that success ever since.

Zaepho posted:

I see a lot of calls for "Develop apps that scale horizontally" which is awesome. Until you realize that the overwhelming majority (aka virtually all of them) of Line of Business apps are developed out of house and simply don;t work that way. It's rare that the primary LoB applications are actually written in house UNLESS you are a SaaS company. Then by definition you are selling that service and are writing it in house because it didn't already exist somewhere.

YYYYYYYuuuupp

And it's hard to change since you have to change the development environment (and thus, the company) from the ground up. Which is basically impossible, because a big company that's lost its product-market fit is basically a startup with less ability to innovate. I've watched company after company dump money into sinkholes and break ankles trying to be agile, but they can't. It's Conway's law in a different form.

# ? Feb 19, 2015 21:28

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

FISHMANPET posted:

No I stayed at the U, just different department. Took a job that's 100% running SCCM because I like SCCM but now I'm not so sure about that anymore, and also wondering if this is a direction I should continue on, or pivot somewhere else.

You're living in the Satya Nadella era. I'm sure SCCM is going to get some really nice Azure management capabilities over the next several years. (Learn. Azure.)

In general, if you're learning configuration management on the Windows side, which tool you learn is going to be determined by where you want to use it. I don't see SCCM being unseated as a killer piece of software for managing changes across thousands of desktops. For servers, it looks like PowerShell DSC is a more forward-looking tool.

# ? Feb 19, 2015 21:29

Zaepho: Oct 31, 2013

Misogynist posted:

Banks are not likely to run in the public cloud for a litany of other regulatory and legal reasons, so they're probably not the best example.

Agreed, probably not eh best example but I knew it best having spent about a third of my professional life working Operations at one. I see the same basic pattern everywhere though. From small Oil and Gas to Medical to Tier 1 backbone providers, to companies the manufacture doors. too many little silo'ed (although often with some integration to other moving parts) applications on servers.

# ? Feb 19, 2015 21:30

bull3964: Nov 18, 2000; DO YOU HEAR THAT? THAT'S THE SOUND OF ME PATTING MYSELF ON THE BACK.

evol262 posted:

This isn't a rub. Private cloud installations should already have the hardware procured and installed, waiting for instances.

It's more of a failing on the business side when specing out clients/projects, but we've had instances where demand for resources as doubled or quadrupled in less than a month with no prior warning. I can't double my local infrastructure with 15 days lead time.

That's where we are having trouble. Our growth rate is outstripping our ability to get hardware in place.

# ? Feb 19, 2015 21:32

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

evol262 posted:

No, it doesn't. You do it the same way as always. You run your applications in different geographical locations so you can lose one and keep going in the other. Your applications are already built so losing pieces is not a critical problem. Go look at Netflix's "Chaos Monkey"

Chaos Monkey is great, but Chaos Gorilla is even better. Chaos Gorilla will, once every couple of months, pick an entire AWS region worth of production infrastructure and just pull the plug to see what happens.

# ? Feb 19, 2015 21:33

Adbot: ADBOT LOVES YOU

# ? May 25, 2024 05:09

Zaepho: Oct 31, 2013

Misogynist posted:

You're living in the Satya Nadella era. I'm sure SCCM is going to get some really nice Azure management capabilities over the next several years. (Learn. Azure.)

In general, if you're learning configuration management on the Windows side, which tool you learn is going to be determined by where you want to use it. I don't see SCCM being unseated as a killer piece of software for managing changes across thousands of desktops. For servers, it looks like PowerShell DSC is a more forward-looking tool.

I sure hope SCCM improves from the Server/Application perspective. It's great for base OS deployment, Patching, some Config monitoring and so forth but it could really use a Distributed application concept would be much appreciated. It helps that the entire suite is licensed together so it is much easier to advocate SCCM on servers if they've already got SCOM on the servers.

# ? Feb 19, 2015 21:37

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Working in IT 3.0: xp_cmdshell

«‹›2905 »