Storagebod Rotating Header Image

More Vintage Stuff

Recently I’ve been spending time thinking about what DevOps really means to my teams and to me. A lot of reading has been done and a lot of pondering of the navel.

Tne most important conclusion that I have come too is that the DevOps movement is nothing new; the second conclusion I have come to is that it can mean pretty much what you want it to and hence there is no right way to do it but there might well be horribly wrong ways to do it.

As a virtual greybeard; I started my IT career in the mainframe world as a PL/1 programmer but I also did some mainframe systems programming. As an application programmer, I was expected to do the deployment, support the deployment and be involved with application from cradle to undeath.

As a systems programmer, we scripted and automated in a variety of languages; we extended and expanded functionality of system programs; user exits were/are an incredibly powerful tool for the systems programmer. VSAM and VTAM – the software-defined storage and networking of their time.

We plagarised and shared scripts; mostly internally but also at times, scripts would make their way round the community via the contractor transmission method.

Many DevOps engineers would look at how we worked and find it instantly familiar…although the rigorous change control and formalised processes might freak them a bit.

So as per usual, the wheel has been re-invented and re-branded.

I’ve boiled down DevOps and the idea of the Site Reliability Engineering function down in my mind to the following –

Fix Bad Stuff
Stop Bad Stuff Happening
Do Good Stuff
Make Good Stuff easier to do
Keep Developing Skills

It turns out that my teams are already pretty working in this way; some folks spend more time dealing with the Bad Stuff and some spend more time dealing with the Good Stuff.

DevOps could be a great way to work; you might find that you already are on this journey and don’t believe anyone who tells you that it is new and revolutionary.

It’s not!

Time to Shop?

So have you taken my advice and started to build your own storage arrays? I certainly hope so; it’s a lot of fun and could really save your organisation some money.

Still I hope that you’ve bought some spares; thoroughly checked the code, made sure that all your components work together and have thoroughly checked out the edge-cases.

You’ve sorted out your supply chain; made sure that you can get to your data-centre at all hours and told your boss that you are happy to be phoned up and shouted at until something broken is fixed.

But you’ve saved your company a bunch of money!

And it was totally worth doing!? You might have to hire another person or two to help out but they’ll be geeks too and you’ll have great fun. You’ll be competing with Google, Amazon etc with your team’s engineering chops. 

Of course, there are a lot of nay-sayers who say you can’t do this and you can’t do it at scale. They are both right and wrong.

This is a thought-process that a lot of us working in the larger-scale enterprises go through a lot. Every now and then we’ll go though the exercise and work out that the costs don’t stack up over the four or five year technology life-cycle we tend to work in. Certainly when you start to factor in the cost of support and the potential risk profile that a Business is willing to sign up to; you just don’t save enough at Enterprise scale. It might be different if you are Google or Amazon and working at a completely different scale. Building your own works at the small scale and the hyper-scale.

But the big vendors are very worried about this trend; you might simply just move wholesale into Public Cloud and then they are going to see very little of your money if you go into one of the big Cloud companies who do their own engineering. I’m finally beginning to see this fear being reflected in the traditional vendor’s pricing; their four year costs are getting closer to the cost of building out excluding the ongoing costs.

Vendors who supply both software-only and appliance versions of their infrastructure generally price their appliance versions at a lower price point than software-only with bring-your-own-hardware. This allows them to maintain revenue at the expense of pure margin; the big analysts tend to report on revenue and market-share as opposed to raw profitably; it seems to be more important to be the biggest and not necessarily the most profitable.

Buying whitebox is not a game that most Enterprises are in yet; whitebox almost takes you to a world where servers are consumables like pen and paper; this is how the hyperscale works. In the Enterprise, we might not be too far off this; if Enterprises start to change their depreciation cycles for compute tiers, it is entirely possible that the software layer becomes the real capital investment and not the tin.

Dell, HPE and the likes simply become the equivalent of a Ryman’s catalogue.

 

Time to Build?

Any half-way competent storage administrator or systems administrator should be able to build a storage array themselves these days. It’s never really been easier and building yourself a dual-head filer that does block and network attached should be a doddle for anyone with a bit of knowledge, a bit of time and some reasonable google-fu skills. I built a block-storage array using an old PC, a couple of HBAs and linux about five years ago; it was an interesting little project, it could present LUNs via FC/iSCSI and file-share via SMB and NFS. It couldn’t do Object but if I was doing it again today, it would.

And it was a single-head device but it was good enough to use as a target device to play about with FC and generally support my home devices. I only recently switched it off because I’m not running FC at home any more.

But if I could build a storage array five years ago; you can do so today. I am not that good a storage/server guy; I’m a tinkerer and dilettante. You are probably much more competent than me.

Another factor that makes it easier is that FC is slowly going away; it’s slow progress but iSCSI making headway for those who really need block, 10 GbE is coming down in price. I’m also interested to see whether some the proposed intermediate speeds of Ethernet have an impact in this space; many data-centres are not yet 10 GbE and there is still quite a cost differential but 1 GbE is not really good enough for a data-centre storage network but 5 GbE and maybe even 2.5GbE might good enough in some cases. And as FC goes away; building your own storage endpoints becomes a lot simpler.

Throw in commodity flash with one of ‘new’ file-systems and you have a pretty decent storage array at a cost per terabyte that is very attractive. Your cost of acquistion is pretty low, you’ll learn a whole lot and be positioned nicely for Infrastructure as Code tsunami.

If you do a great job, you might even be able to spin yourself out as a new flash-startup. Your technology will very similar to a number of start-ups out there.

So why are you sitting here, why are you still raising POs against the three or four letter name vendors?

Imagine never having to speak to them again, what a perfect world.

Tooling Up?

As the role of the storage admin changes; the toolset will change and we will need new tools that assist us. Modelling tools that enable us to work with our workloads are generally very expensive and often have rather dubious ROI models that allow us to justify them to our management. 

And you rarely need them but it you are heading into a refresh cycle for example and you don’t necessarily always trust the vendors to mark their own homework..a tool is useful, especially one that you can model your own workload on.

So this tool from the newly merged Load Dynamix and Virtual Instruments looks interesting; I’ve not had a play yet and am not sure of the limitations but free is hard to beat..

Workload Central 

Hopefully more vendors and the community will get involved and add more potential data sources..

Pestilential but Persistent!

There is no doubt that the role of the Storage Admin has changed; technology has moved on and the business has changed but the role still exists in one form or another.

You just have to look at the number of vendors out there jockeying for position; the existing big boys, the new kids of the block, the objectionable ones and the ones you simply want to file. There’s more choice, more decisions and more chance to make mistakes than ever before.

The day-to-day role of the Storage Admin; zoning, allocating LUNs, swearing at arcane settings, updating Excel spreadsheets and convincing people that it is all ‘Dark Magic’; that’s still there but much of it has got easier. I expect any modern storage device to be easily manageable on a day-to-day basis; I expect the GUI to be intuitive; I expect the CLI or API to be logical and I hope the nomenclature used by most players to be common. 

The Storage Admin does more day-to-day and does it quicker; the estates are growing ever larger but the number of Storage Admins is not increasing in-line. But that part of the role still exists and could be done by an converged Infrastructure team and often is. 

So why do people keep insisting the role is dead? 

I think because they focus on the day-to-day LUN monkey stuff and that can be done by anyone. 

I’m looking at things differently; I want people who understand business requirements who then turn these into technical requirements who can then talk to vendors and sort the wheat from the chaff. People who can filter bullshit; the crap that flies from all sides; the unreal marketing and unreal demands of the Business.

People who look at complex systems and can break them down quickly; who understand different types application interactions, who understand the difference between IOPS, latency and throughput.  

People who are prepared to ask pertinent and sometimes awkward questions; who look to challenge and change the status-quo. 

In any large IT infrastructure organisation; there are two teams who can generally look at their systems and make significant inferences about the health, the effectiveness and a difference to the infrastructure. They are often the two teams who are the most lambasted; one is the network team and the other the storage team. They are the two teams who are changing the fastest whilst maintaining a legacy infrastructure and keeping the lights on. 

The Server Admin role has hardly really changed…even virtualisation has little impact on the role; the Storage and Network teams are changing rapidly, many are embracing Software-Defined whilst the Industry is trying to decide what Software-Defined is.

Many are already pretty DevOps in nature; they just don’t call it that but you try to manage the rapid expansion in scale without a DevOps type approach. 

I think many in the industry seem to want to kill off the Storage specialist role; it is more needed than ever and is becoming a lot more key…you probably just won’t call them LUN Monkeys any more..they’ve evolved!

But we persist…

Reality is persistent

I see quite a few posts about this storage or that storage..how it is going to change everything or has changed everything. And yet, I see little real evidence that storage usage is really changing for many.  So why is this? 

Let’s take on some of the received wisdom that seems to be percolating around. 

Object Storage can displace block and file?

It depends; replacing block with object is somewhat hard. You can’t really get the performance out of it; you will struggle with the APIs especially to drive performance for random operations and partial updates.

Replacing file with object is somewhat easier, most unstructured data could happily be stored as object and it is. It’s an object called a file. I wonder how many applications even using S3  APIs treat Object Storage anything other than a file-store, how many use some of the extended metadata capabilities? 

In many organisations; what we want is cheaper block and file. If we can fake this by putting a gateway device in front of Object Storage; that’s what we will do. The Object vendors have woken up to this and that is what they are doing. 

But if a vendor can do ‘native’ file with some of the availability advantages of a well-written erasure coding scheme at a compelling price point, we won’t care.

And when I can boot from Object Storage..call me.   

All new developers are Object Storage aficionados?

I’m afraid from my limited sample size; I find this is rarely the case. Most seem to want to interact with file-systems or databases for their persistence layer. Now the nature of the databases that they want interact with is changing with more becoming comfortable with NoSQL databases.

Most applications just don’t produce enough data to warrant any kind of persistence layer that requires Object or even any kind of persistence layer at all.  

Developers rarely care about what their storage is; they just want it to be there and work according to their needs. 

Technology X will replace Technology Y

Only if Technology Y does not continue to develop and only if Technology X has a really good economic advantage. I do see a time when NAND could replace rotational rust for all primary storage but for secondary and tertiary storage; we might still be a way off. 

It also turns out that many people have a really over-inflated idea about how many IOPs their application need; there appears to be a real machismo about claiming that you need 1000s of IOPS…when our monitoring shows that someone could write with a quill pen and still fulfil the requirement. Latency does turn out to be important; when you do your 10 IOPS, you want it to be quick. 

Storage is either free or really cheap?

An individual gigabyte is basically free; a thousand of these is pretty cheap but a billion gigabytes is starting to get a little pricey.

A Terabyte is not a lot of storage? 

In my real life, I get to see a lot of people who request a terabyte of storage for a server because hell, even their laptop has this amount of storage. But for many servers, a terabyte is a huge amount of storage..many applications just don’t have this level of requirement for persistent data. A terabyte is still a really large database for many applications; unless the application developers haven’t bother to write a clean-up process.

Software-Defined is Cheaper? 

Buy a calculator and factor in your true costs. Work out what compromises you might have to make and then work out what that is worth to you. 

Google/Amazon do it, so we can too?

You could but is it really your core business? Don’t try to compete with the web-scale companies unless you are one..focus on providing your business with the service it requires. 

Storage Administration is dead?

It changed, you should change too but there is still a role for people who want to manage the persistent data-layer in all it’s forms. It’s no longer storage…it’s persistence.

Mine is the only reality?

I really hope not…

 

 

 

 

Technology Live and a Little More…

Last week, I was at A3 Communications’ Technology Live event;  it’s a smaller event where a group of journalists, bloggers and analysts are briefed by three or four companies. Good fun, a chance for awkward questions to be asked and generally good-humoured banter. 

It is a chance for some of the smaller and lesser known companies; some just pretty much unveiling from stealth to get their message across without some of the hype and hyperbole of the larger events you sometimes associate with the business. 

Companies like Scale Computing and their converged platform probably deserve to be much better known; targeted at the SMB and smaller user whose IT department is one person who actually has another proper job, quietly get on with things without press releases about yet further funding rounds and a gazillion dollar valuations. It is one of the few times when I’ve had a converged platform demonstrated where I’ve thought ‘well that makes sense for their target market’ as opposed to ‘shiny lights…but where’s the substance’. 

DDN are much larger and better known that Scale Computing but probably not as well-known as they should be; their HPC roots are allowing them to play in the scale-out and big data space. They’ve taken massive strides in hiding some of the complexity of their products; what was really a bit of an engineer’s product, now has some polish that really lends itself to the Enterprise.  If you are looking at tiering from primary storage to a secondary storage object tier; I think that you must have a look. 

Tarmin have been around for ages with their Gridbank Data-Defined Storage; it’s a really interesting concept but it’s one that I still struggle to find the use-case that will really drive it forward. A Swiss-Army knife of a product that might be lacking that one blade that would make it compelling; I feel that it’ll just need too much work to integrate into most application environments and I also have concerns about how easy it is to get out of if you decided that it was no longer the platform for you.

We also had OpenIO who are another Object Storage vendor in what is an increasingly crowded space; new to the game and building on-top of an open-source product. You pay for the support and not the product; obviously, it’s model that has worked well for some in the past but I feel that you really need some critical mass before it becomes viable. And there’s many alternatives out there now but it did look nice; hexagons instead of circles. It is also really easy to get up and running quickly; install vagrant if you haven’t already and then a couple of commands, you can quickly have an object store up and running. With Swift and S3 compatibility; it could be a nice entry point for developers to play with.

Earlier in the month, I was at BVE for my day job. I chatted to a few vendors but I really want to call out what I think is a perfect example of a company who are successfully building a business out of doing something extremely well in a well-defined niche. Object Matrix who are based in Wales do Object-Storage for media applications; they have spent a lot of time integrating with products like Avid and GrassValley, really understanding the business that they are in and building a successful company without mega-investments. And they are really nice people….who unfortunately support one of the weaker sides in the Six Nations ;-). 

There are many companies like some of the aforementioned who are doing great jobs for their customers who aren’t getting the recognition because they don’t play in the ‘glamour’ end of the market but I suspect some of them will still be around years after the Unicorns have turned out to be pit-ponies…

Perhaps you work for one; if so…get in touch, I’d love to hear from you. 

 

Skating with Cerberus

I imagine there was a sharp intake of breathe as Microsoft announced SQL Server for Linux and then a checking of dates. And yet it makes perfect sense, a very sensible strategic move for Microsoft.

My question and I know I’m not the only person asking this is; what is the future of Windows in the data-centre? If SQL Server runs well on Linux; there are a vanishing small number of workloads that I would want to run on Windows Server in a data-centre. Yes there are alot of third party applicatons that run on Windows and this is going to continue for many years but I do really wonder if Microsoft’s heart is really in the Windows Server business.

Microsoft appear to have decided that their future is in Cloud; not the Enterprise DC. I mean it’s always been questionable whether anyone sane would run Exchange and now you don’t have to; Office 365 takes care of that for you.

A lot of people like Azure and sure Microsoft would prefer you to run your cloud apps in Azure but if you want to run them elsewhere; they would like to still make money out of you. SQL Server on Linux will remove some of the friction for deployment in the Cloud.

SQL Server running on Linux also allows them to compete with Oracle in those data-centres that Windows is simply a grudging presence; there are certainly those who will have you believe that SQL Server is not Enterprise but many of those comments have been driven by the stigma of Windows. I work with DBAs who do both; for most workloads, SQL Server and Oracle are equally good.

So what’s left for Microsoft to do?

Well, if Microsoft announce AD Services running on Linux; you’ll really know that their heart is no longer in the Windows Data-centre.

Something’s missing?

Yes, I know…I’m getting very lazy about blogging; I’m still not sure if the industry is boring me or simply exasperating me so much that I cannot be bothered to vent my spleen any more. I suspect that it is a bit of both! This should be an interesting year for the industry with the mergers, takeovers and companies simply thrashing around trying to reinvent themselves. So apart from life still being somewhat stressful, I amuse myself trying to get my home-office perfectly set-up. I might even put up pictures once I have done so!!

Anyway the recent announcements from companies large and small around All Flash Arrays has temporarily pricked me awake; hopefully at some point soon, the All Flash Array Announcment will no longer be a thing, it’ll just be another array announcement. Flash will eventually subsume rotational rust as the primary storage medium of choice for all workloads; well until the next big thing comes along. Opinion as to when this is does vary from pundit to purveyeor but it is going to happen.

That time is not here though and perhaps it is still worth considering the best use of our storage capacity and how to get the most from it. And it seems that some vendors don’t really want to help us poor customers in this space.

If you ship an AFA variant of existing array and you either add new features that aren’t supported on the exisiting variant across all tiers of storage be it flash or rotational rust or vice versa; I want good architectural reasons as to why you can’t do so. Compression for example works very well on both traditional disk and flash; in-line deduplication is harder, so you might get a pass on the latter but not the former. If you want to try to convince me that your expensive Flash tier is actually as cheap as the traditional tier you also ship; you are going to have work extra hard to do so when competing with vendors who can actually enable features across all of their tiers.

I shall leave it to the reader’s imagination as to which vendor might be attempting to play this game.

2016 and Beyond…

Predictions are a mug’s game…the trick is to keep them as non-specific as possible and not name names…here are mine!

What is the future for storage in the Enterprise? 2016 is going to pan out to be an ‘interesting’ year; there’s company integrations and mergers to complete with more to come so I hear; cascading acquisitions seem likely as well.

There will IPOs; they will be ‘interesting’! People are looking for exits, especially from the flash market. A market that looks increasingly crowded with little to really tell between the players.

Every storage vendor is going to struggle with maintaining growth; technology changes has meant that it is likely that just to maintain current revenues that twice as much capacity is going to have to be shipped. Yet data efficiency improvements from thin-provisioning to compression to dedupe mean that customers are storing more data on less capacity.

Add in the normal year-on-year decline of the price of storage, this is a very challenging place.

Larger storage customers are becoming more mecurial about what they buy; storage administration has got so easy that changing storage vendors is not the big deal it used to be. The primary value these days of having some dedicated storage bods is that they should be pretty comfortable with any storage put in front of them.

As much as vendors like to think that we all get very excited by their latest bell or whistle; I’m afraid that we don’t any more. Does it make my job easier; can I continue to more with less or best case the same.

Data volumes do continue to grow but the amount of traditional primary data growth has slowed somewhat in my experience.

Data from instrumentation is a real growth area but much of this is transitory; collect, analyse, archive/delete…and as people start to see an ever increasing amount of money flowing to companies like Splunk expect some sharp intakes of breath.

Object Storage will continue to under-perform but probably less so. S3 will continue its rise as the protocol/API of choice for native object. Many file-stores will become object at the back-end but with traditional SMB/NFS front-ends. However, sync and share will make inroads formally into the enterprise space; products like Dropbox Enterprise will have an impact there.

Vendors will continue to wash their products in ‘Software Defined’ colours; customers will remain unimpressed. Open-source storage offerings will grow and cause more challenges in the market. Some vendors might decide to open-source some of their products; expect at least one large company to take this route and be accused of abandonware. And watch everyone try to change their strategy to match this.

An interesting year for many…so with that, I shall be off and wrap presents!

May you all have a Happy Christmas, a prosperous New Year and may your bits never rot!!