having seen what (capable) server admins did to perfectly innocent SAN or arrays, I can sleep comfortably and know:
Their incompetence is my job security.
There is more to storage administration than just allocating gigabytes.
Last week I had an interesting chat with Andy Warfield (CTO and founder of Coho Data). We started a debate about his latest article and the pros and cons of custom hardware design in modern storage products. However, the conversation quickly got side-tracked to another topic: the role of the storage admin, if any, in 2016. It …
I agree completely, but the last time I was looking for work (this time last year) the number of places that still maintained their own Storage team was getting pretty small. Only the real behemoths still keep specialized storage staff, and everyone else has folded it into the sysadmin role. Not just that, but they didn't regard my lack of storage administration as any big blocker because "it's easy to pick it up". *boggle*
You're right, to an extent. There have always been organisations, normally smaller businesses and public sector entities who have done without dedicated storage admins. With the simpler products that you get nowadays more and more organisations fall into that category.
Many of the vendors are still churning out the same technology they were ten or even twenty years ago. Yes, they have flash in there now, automated tiering etc. which make things easier but they're not admin-free. And you have to think about what happens when something goes wrong. For a dual- controller system, when a controller fails, performance generally goes down the pan. Yes, so the vendor will help out but they don't know how your business works. They don't know how to deal with the guy who's shouting that his application keeps failing because latency has gone through the roof.
Being a good storage admin isn't about being able to create a spreadsheet with some commands to cut and paste into a CLI (yes there are those who still do that). It's about understanding your systems, knowing how they operate and how they relate to the business and the end users, which ultimately is what IT is there for. It's about being able to identify risks, develop strategies to mitigate against those risks and have the ability to implement those strategies.
So yes, the traditional role of the storage admin is long gone, but storage is where your entire business resides. Shouldn't you have someone looking after it?
Scale-up/out isn't the interesting part, and performance tuning was a dying art since XIV first came out.
The interesting part of storage management is all about reliability and availability at the organization level: Mirroring, DR, even snapshot management. Don't see this part of the storage admin role going away anytime soon.
I don't think it's just storage. Like the article mentions, the dedicated, deep-knowledge server admin is going away as well. Part of this is automation -- tools like Chef/Puppet on the Linux side and PowerShell DSC on the Windows side are making it incredibly easy to stand up environments. The other part for the hardware and infrastructure nerds (like me!) is the cloud. I'm currently working on a project that's 100% in Azure, and it's a huge leap going from physical hypervisor hosts to hardware you can't see or touch. It's just as complex, but in a different way -- now you have to design your system to stand up to an environment that you don't control the downtime on, plus wade through all the marketing hype and figure out how things are actually accomplished.
I think the long term answer is to do everything you can to stay a generalist. I've worked almost exclusively for big companies that had the storage silo, the network silo and the server silo. Colleagues of mine have invested years of their lives learning the ins and outs of, say, VMWare or EMC or Citrix to the exclusion of all else. It's very easy to get trapped, because once you're the expert on a particular system everyone wants your knowledge...until they don't need it anymore!! My plan has been to get involved with a little bit of everything, and that has landed me in a systems architecture role. I do end user stuff, servers, storage, enough networking to survive, you name it. My next move is basically to become a platform-agnostic automation guru, because that's definitely where things are headed. Sysadmins for the most part will be taking care of thousands of machines and aren't going to be able to do it with manual changes.
.... you were already just a rack guy and never a system administrator.
It doesn't matter if the servers are virtual or physical, if they are next to you or on the other side of the planet - good administrators always "orchestrated" their systems with the tools available, even when they were no more than a shell and its scripting language, and a task scheduler.
Everyone I encountered who had no clue how to do it properly was just a rack guy who thought to be a syadmin - but never became one.
Also people that focus on a single "task" - be it applications, servers, storage, networking, whatever - and never try to understand what there is around to build the whole system, are other clueless workers trying to get a salary at the end of month, and nothing more.
If these types eventually disappear, no one will complain.
I think there is still a need for specialisation at large outsourcers or cloud providers. I've been contracting a few years ago at a large outsourcer and they had dedicated teams for Networking, Storage, Linux, Windows and Virtualisation.
Unfortunately these teams didn't always communicate and problem resolution wasn't always easy.
However, due to the scale (number of devices) and nature of outsourcing people had to automate whatever they could to cope with the workload. That was before it was called orchestration.
I was the one of the storage guys which by default is the smallest team and other teams instinctively pointed the finger at storage for all sort of issues, but mostly performance related.
I always had to work my way up when proofing my innocence (Storage ->Network(SAN or LAN)-> Server /VM/OS -> Application -> Network again ....)
Poor Communication (Silos) forced me to learn a lot about the entire Solution, to the point where I could pinpoint issues outside of the storage domain. The other teams didn't necessary like it.
So to the point - Specialisation will stay for a while. The people that tell you that orchestration, self healing,self optimising, automatic provisioning will take everyone's job are full of BS. The tools available are not good enough (yet) and as soon as your Orchestration/Automation Single Pain of Glass Dashboard has a hiccup, you're back at the CLI - or alternatively a support ticket with the vendor(s) which will take weeks to resolve.
There will be storage admins with the large cloud providers and the small to medium businesses never could afford a storage admin in the first place. Their Generalist Datacenter person may become a Contracts administrator - whose most important job is to uncover hidden costs in the cloud.
For years now, there have been tools attempting to automate storage management (vmware mgmt as well). I was on a contract a few years ago and the site was implementing the HP automation tools that promised to reduce the IT staff personnel. While that did happen, most of the staff that became redundant moved over the HP Tools team after the implementation phase was over and HP professional services left. It really was a beast to manage.
My own story is about 10 years ago I knew very little about storage. When I looked at an array I saw it did RAID, had redundant controllers, had fibre channel connectivity, and that's about it. I didn't understand the differences much beyond that. My background before was mostly servers, and some networking. I had dabbled a bit(very little bit) at one company, mostly following directions from the formal storage admin (this was about 12 years ago). I saw them planning the data layouts in excel and visio at the time. I thought to myself "I don't want to do storage, pain in the ass".
Then I came across 3PAR and well I suppose the rest is history. I have learned a lot about storage in the past decade, and things run pretty well. Managed 3PAR systems across 4 different companies over the past 10 years, currently have 4 arrays at two data centers. I would never call myself a storage admin, though I do do storage admin related things I suppose (nobody else in the companies I have worked at did storage stuff). The thing I perhaps learned that is the most important lesson is that features or performance alone is only a small part of the equation when it comes to storage. I've certainly had issues here and there with 3PAR in my years using them and so am MUCH more conservative with storage management today than when I started. I have had multiple people at HP(and 3PAR before) tell me that they have never known someone who operates 3PAR arrays as well as I do. I may not of ever become a 3PAR person if it wasn't for NetApp refusing to lend a system to do an evaluation 10 years ago.
Storage still makes up a very small % of my time, networking generally even less(switches especially, hardly ever touch them, I am the primary SME for switching, load balancing, firewalls, VPNs, storage, vmware, servers etc at my company ~$300M annual revenue). I realize I do a lot of different things, more than most people in my field(I've never met anyone else like me but I have not tried to either). So I am probably more of an edge case (or extreme edge case) rather than a run of the mill systems person.
(system admin/engineer/etc for roughly 20 years now)
Did you get your disconnection notice?
Mine came in the mail today
They seem to think I'm disconnected
Don't think I know what to read or write or say
Glossaries injected daily
Words and numbers spell out the price to pay
Its simply states that, "You're disconnected, baby"
See how easily it all slips away
This obviously doesn't take into account all the HA/DR solutions as well as the complexity of managing a proper high end array like a VMAX. Storage performance is also another aspect which cannot be conveniently scripted away.
Perhaps at the low/middle end of the storage market, a SysAdmin can get away with provisioning a few LUNs due to disk pooling taking away some of the thought which used to go into provisioning, but this has been the case for several years now and is not new news.
The future is storage virtualisation. But thats another 5-10 years away from completely transforming the market. Will we still need storage admins? I think you'll need to check how complicated SDS solutions are compared to disk arrays today, you'll need to know about the Linux OS, clustering, different data protection methods, metadata manipulation and a bit of scripting to boot as well as all the networking/pooling/performance skills you need today. Check out CERN's implementation of Ceph in a ~30PB cluster for a giggle at people who think data storage wont be a specialist job especially at 'Big Data' scale.
And for the most part we are better than a lot of our current sysadmins at being sysadmins. We are but 4 guys managing 10Pb of block and NAS. Sure automation and self provisioning are something everyone's getting excited about but like some of the other posters have said someone has to 'watch the box.' They systems are all just big servers and they don't take care of themselves all the time. I've been doing this for over 10 years and I've never met the PHD storage admin. We're just sysadmins (none with a CS degree mind you, like almost the entire infrastructure team.) Automation is great but the Sysadmin who has hundreds of virtual servers and doesn't know much about the underlying infrastructure of those VM's is kind of a problem isn't it? There's already a gigantic problem with app and dev people who don't know the first thing about their dependencies so they can't trouble shoot anything. There has to be a balance between automation and understanding what you're working with and what it can and can't do.
Biting the hand that feeds IT © 1998–2022