With the radical transformation of the data center storage landscape, and the transition from traditional block/file storage towards software-based / software-defined solutions, plus the advent of object storage and cloud integration, the bastion of traditional storage administrators is under siege from all sides. Is there a future for the storage administrator profession?
Why should we care?
So, why does this blog’s author -a nobody- gives a damn about storage administrators and why should it be interesting? Well, first of all I’m a human and I have to a certain (limited) extent a natural empathy towards people and their well-being. A few months ago, someone I know who works in a large company told me an interesting anecdote about how a colleague of them who works in the storage area approached them while they were deploying an hyper-converged solution and asked “so how’s the storage of these systems look like, is that block based? Are we going to manage it?”. Needless to say, I felt sorry for the storage person, because hyper-converged solutions are by essence making the storage administrator irrelevant. Lately, in another company, I was told about how a few storage admins were let go without any reason. And finally, I’ve also heard of a case when storage admin folks purchased Nutanix appliances for a PoC. Really?
Who’s targeting who?
But it’s not just about hyper-converged. Look at Tintri: to whom do they sell? To the VMware administrator! And they’ve done so since their beginnings. It’s not just Tintri anymore – see Enrico’s post on VM-aware storage. It’s not just that new storage solutions are (or not) VM-aware, but that they can be very simply deployed (generally in the range of minutes/couple hours) and managed out-of-the-box by any competent VMware administrator.
The VMware administrator perspective is that they are dependent on another person to provision storage for their infrastructure and that increases overall provisioning time. Considering also that a fraction of VMware administrators (and that was especially true a few years ago) were not versed into the lore of SAN switches configuration, zoning, masking etc, there is a portion of the profession that is all ears when simpler solutions can be used. But back to why they can be managed so easily – if one may say so.
Protocols and backplanes
Look at the backplane. The preferred connectivity for storage arrays was until a few years ago to rely on Fiber Channel connectivity. Look now. Most of the modern solutions we see around are based on Ethernet connectivity. Yes, some offer optional FC. But why should we care? If you don’t have an existing FC infrastructure, is it worth investing into one and having to pay for the in-house knowledge to operate and maintain it? And what’s the advantage of running on FC nowadays?
The advantage of a dedicated connectivity infrastructure (8 Gb, with now 16 Gb and maybe also 32 Gbps) is being gradually negated by the rapid pace in Ethernet connectivity speed increase (currently 10 Gbps, with 40 Gbps already used for backplane connections and 100 Gbps at the very near horizon). While on the technical side of things one can argue about the fact that FC is dedicated and may offer better security, there is no reason to think that Ethernet is not fit for the job provided that you apply common sense in properly isolating your management and transport networks. Also, the overhead in frame encapsulation and lack of packet prioritization is minor – the attempts to keep the FC protocol alive such as FCoE are (sorry to say so) dead and probably limited to some attempts to deploy FCoE in Cisco workshops a few years ago, when 10 GbE wasn’t mainstream. If you have to refresh your FC infrastructure and your storage infrastructure at the same time, it may be worth looking into the economics of refreshing vs moving to an Ethernet based solution.
It’s not just the backplane, it’s also the built-in features: do you remember the time when you had to create a RAID group, then provision LUNs on top of that, then eventually do some masking (and ignoring hypervisor operations)? Even ahead of this, consider the workload, determine which RAID level suits best based on the I/O model (if anyone ever cared about), which disks to assign.. Now, this is taken care of by the modern storage systems. Plus you generally get at least some level of caching. Let’s take the case of a Nutanix hyper-converged system. I create a storage pool, assign all the disks of my cluster, then create one or two containers on top of that. I don’t have to care about the underlying disk layout – and honestly, should I care?
Storage Administrators: are we doomed?
So pardon the digression and let’s get back to our topic: what’s in for storage administrators? That job is becoming irrelevant not only because of modern storage systems, but also because of the advent of public cloud storage solutions that leverage programmable APIs (think Amazon S3 – see Chris Evans comprehensive series). Programming/scripting skills will be required for these and even VMware administrators should jump on the train and educate themselves about Amazon Web Services – or they may have to look for a reconversion in 5 years.
But we still see job postings about Storage Administrator jobs! And traditional storage vendors are still selling hardware! Yes, this is true, let’s not forget the weight of legacy. Many large companies have still traditional storage arrays around. And upon pressure (from vendors, or from their storage silo/employees) they may still buy traditional storage. But consider that in many cases, the lifecycle of these is at best five years. What will these companies do in five years, and what will you do?
Storage Admins? VMware Admins?
Every serious company is weighing the cost of in-house IT and data center infrastructure versus public cloud infrastructure. While we can expect to see a mix-and-match of both (especially since the dependency on legacy applications may lie heavy), it doesn’t means that this will be a bloodless war without collateral victims. So, should the storage administrator evolve to become an IT infrastructure generalist? Maybe yes.
Even VMware administrators need to think about the future. VMware is alive and well, but even this inexpugnable fortress is under attack – not just frontal attack on the hypervisor front (KVM, AHV, Hyper-V) but also at its foundations by new types of workloads. And while you won’t run an SQL server or an Oracle DB (for now) on a container, who knows what the future reserves. Long time ago an Intel CEO wrote a book called “Only the paranoid survive”. Don’t take the threat lightly, these might be a modern embodiment of the treacherous Rabbit of Caerbannog.
For the sake of all we cherish I won’t quote Marc Andreessen but you get the picture. This is not an apocalyptic prophecy, it’s happening. Will you be ready?