Archive for the ‘Cloud storage’ Category

Boil the kettle, data rationalisation and reduction could take a while.

23/07/2010

UPDATE Mon 26 July: My interview with The BBC World Service ‘The World Today’ programme covering this topic aired this morning.  Click here for the full thirty minute podcast, or here for just my interview excerpt.

I thought perhaps I would begin this Weekly View with a quick experiment …now, you’ll need a kettle for this exercise …and in this context we need this to be a kettle of the electric variety … so if you don’t have one, or reading this blog where ‘lectric kettles may be a foreign concept, here’s a picture of one which will suffice for now.

Okay, ready?  Great.  Now, I’d like you to go and boil your kettle seventeen and a half times.  It’s okay, I know it takes a bit for each boil.  I’ll wait.  See you in a few minutes …or if you’re feeling generous, mine’s a PG Tips with milk and two Splendas.

Right …all done?  Great!  You’ve just expended as many greenhouse gases as you would by sending an email with a 4.7 megabyte attachment.

That’s right, campers … boiling your kettle 17.4 times consumes as many resources (electricity, water, and the like) and produces as much greenhouse gas as sending an email with a 4.7MB attachment in a traditional IT environment.

Source: Life-cycle analysis carried out by Mark Mills of Digital Power Group, translated into kettle boilings with help from the Energy Savings Trust [UK].

Now, I know what you’re thinking as I was thinking the same thing when I first read that statistic …what?  How can this be?!

Without getting overly geeky on the topic, the short answer is that traditional IT environments tend not to be overly efficient at scale and we’ve known for quite some time that the traditional IT infrastructure …server plus storage plus network plus operating system plus application …tends to be siloed with each individual component connected physically to the others with wastage and efficiencies lost between these connections and within the physical devices themselves.

And, to be fair, traditional datacentres don’t fare much better …indeed, the datacentre industry as a whole has reached parity with the airline industry in CO2 production with 2% of all man-made CO2 comeing from computers and communications technology.

Source: Green IT: A New Industry Shock Wave by Simon Mingay [Gartner]

What’s this got to do with Data Storage and Protection?

I suppose that there is the obvious ‘think before you send emails with 4.7 meg attachments’.  I’m somewhat bemused …well, saddened really …that with the green revolution of the past ten years or so I now get emails with the tagline ‘Think before you print!’ with pretty green trees from just about everyone these days.  But what about having a tagline which gently asks the user …’Do you really need to send this, and, if so …please consider an alternative method rather than sending the attachment.’  Or, ‘Think before you send!’ for short.

Email has been the bane of many a storage administrator’s life as it has morphed from a purely text based system …remember SNDMSG, CICS, and OS/400? …to the rich text and inefficient file distribution model we find today.  Why do people attach large files to email and send them to a large distribution list?  I suppose the short answer is …it’s easy and they would argue they’ve more important things to be getting on with.  Fair enough.

But this isn’t a blog post having a whack at email and email vendors …and we should consider the fact that the ubiquity of smart phones, high definition cameras, et al mean we’ll continue to create ever larger files …indeed, we’re uploading 24 hours worth of video to YouTube every minute up from 12 hours a minute just two years ago …so how do we reduce the amount of resources we’re consuming …electricity, datacentre space, and people to run the show cost dosh you know! …and the CO2 we’re creating when we need to share files with others?

I think there are five answers which are worth considering.

1. Introduce proactive limits for users.

Let’s face facts, reactive limits with users tend not to work and/or are quickly circumvented to keep the business moving.  Telling users ‘your email mailbox cannot grow beyond 20MB or we’ll cut you off so you can’t send/receive email’ rarely works in my experience.  Rather, we need to evolve this theory to be proactive.

For example, I use a great cloud based application called Evernote.  I could write a whole post on just how great this app is …it allows me to take notes anywhere I am on my MacBook Air, iPod, Blackberry and keeps the notes and related notebooks in sync so that where ever I am, all of my notes are up to date without me having to do anything.  Brilliant.

But here’s where it gets even better …it’s free.  Provided I don’t exceed a 20MB monthly limit, of course …and therein lies the true genius in my mind.  Evernote resets my 20MB limit at the beginning of each month so, providing I don’t exceed the 20MB in a month …sorted!  This is the type of proactive limit I’m thinking of for users …we give you a limit and then count down monthly to zero.  Stay in your limits, you’re good to go …exceed them, we charge you more on a graduated basis.

2. Rethink service levels for workloads.

So what are Evernote doing with the 20MB that I created last month …it doesn’t get deleted from my syncronised notes as they remain available to me, so what gives?  To be honest, I’m not quite sure …my guess would be they move the data previously created to a lower tier of storage, such as dense 2TB SATA drives, or even archive.

To be fair, I don’t much care.  I don’t notice any performance degradation and I get to carry on consuming the service for free.

Perhaps this is the key to the answer with our users …we’ll keep your data in a highly performant architecture for one month and then demote to a lower less performant tier thereafter and reset your limit.  And we won’t ‘charge’ you unless you exceed your limit.

3. Introduce datacentre technology optimisation in the form of virtualised datacentres [VDC]s.

I’ve talked a lot about VDCs in previous posts starting with this one and many more since, so there’s no reason for me to labour the point more here other than to say that what VDCs deliver is optimisation by removing wastage as well as increasing business agility.

How much optimisation?  Chad Sakac, Vice President of VMware Strategic Alliance and the general in charge of ‘Chad’s Army’ of VCE vBlock gurus, blogged in 2009 about the potential benefits of deploying a vBlock against deploying technology silos.  An excerpt follows below:

  • 30% increase in server utilization (through pushing vSphere 4 further, and denser memory configurations)
  • 80% faster dynamic provisioning of storage and server infrastructure (through EMC Ionix UIM, coupled with template-oriented provisioning models with Cisco, VMware, and EMC)
  • 40% cost reduction in cabling (fibre / patch cords etc.) and associated labor (through extensive use of 10GbE)
  • 50% increase in server density (through everything we do together – so much it’s too long to list)
  • 200% increase in VM density (through end-to-end design elements)
  • Day to day task automation (through vCenter, UCS Manager and EMC Ionix UIM)
  • 30% less power consumption (through everything we do together)
  • Minimum of 72 VMs per KW (note that this is a very high VM/power density)

Now, I say potential benefits as, at present, these numbers have been derived from product datasheets and lab work by EMC …however we at Computacenter are looking to put more substantive quantitative analysis behind these benefits (and those of other VDC variants such as NTAP SMT, HDS UCP, ‘open VDC’) as we deploy VDCs with our customers locally in the UK.  Watch this space.

4.  Use alternative distribution tools for large attachment distribution and filesharing.

I really try not to use email as a file distribution these days, preferring instead to use cloud applications such as Dropbox to share large files with others such as internal, customers, and our vendor partners.  Now, this isn’t perfect as a) in the absence of my using encryption I wouldn’t wish to use this for sensitive corporate data, and b) it does have a ‘hard stop’ limit where I can only store 2.5GB for free with no reset limit like we have with Evernote.

But using tools such as Dropbox, uploading personal photos to Facebook instead of emailing them, if I must send an attachment trying to shrink it by converting to PDF or similar …every little helps!

That said, I accept that I’m a geek by trade and we need to find ‘easy’ ways for everyday users which replace email as a distribution system without increasing complexity.

After I’ve done that I’m planning to sort world peace, famine, and poverty.

5. Rethinking how we create data.

Only about 20% of the data we create today is ‘text’, with rich media [high def videos, pictures, etc.] representing well over 50% or more of the new data being created.

Equally, the text data we are creating is rarely just text …by the time we save it in MS Word or similar we have increased the file size with the formatting and related metadata, and many users choose to use PPT to communicate ideas such as business plans and so on …a large file type if ever there was one …and that’s without even adding any pictures, charts, or videos to the PPT.

Again, I’m not having a go at the use of PPT or MS Word …but I do believe we are going to have to begin to think about how we create data so that the data ‘footprint’ is smaller in addition to the optimisation and alternative distribution models we’ve discussed above.

Which has me thinking …it’s time for a nice cuppa before Mrs. PL needs my help setting the table for dinner with she and PL Junior …the highlight of my week!

Have a great weekend and remember your kettle the next time you send an attachment.

-Matthew

Five reasons why cloud isn’t a four letter word.

16/07/2010

This is the second post in the Federated Service Provider series, click here to go to the first post.

Growing up my parents had two aspirations for me …one, to become a doctor and two, to play the piano.

Now, I studied neuroscience in university but didn’t go to medical school …long story, with the jury still out as to whether I get partial credit or ½ a point for that aspiration …and as for playing the piano, well I can play anything brass and was … okay, still kinda am …a band geek having been in marching band, concert band, and jazz/Big Band band.  Heck, I could probably even manage film classics on the vuvuzela.  But piano?  Erm, no …I would most certainly lose a contest to Chuck Hollis no matter how many drinks you gave me.

Why?  Two primary reasons, really …first, I could never could get to coordination required with feet, hand, pedals blah blah …and second, I never practised enough.

Now, before you go castigating me for not practising enough …pianos were expensive to own when I was a kid and keyboards hadn’t really hit the mainstream.  Yeah, okay, there were these things called ‘fingerboards’ with which you could ‘practise’ playing piano but really?  Cardboard fingerboard on the dining room table?  Exactly.  No self-respecting Atari 2600 playing kid is going to pretend playing piano …sorry, ‘practise’ …when there was Tron to be played in the absence of a real piano to practise with.

I hadn’t thought of any of this very much until very recently when I saw amazing piano tutor apps for the Apple iPad.  Would having had an iPad …admittedly much cheaper than a piano and also capable of doing much more than being just a piano …have changed things for me such that I would have become a concert pianist?

This blog post is by no means about the iPad …and gosh knows the iPad does evoke interesting responses from people …I’ve heard everything from the fan boy addict ‘I’ll stand  in queue overnight to buy one’ to people expressing hatred tantamount to ‘Steve Jobs is Hitler and Chairman Mao’s love child’ …and everything in between.

Look …the iPad, just like any technology, is important only inasmuch as what you are going to do with it.  And some are taking the view that perhaps better to test in the field and form a quantitative view as opposed to an emotive qualitative view.  Indeed, Eversheds have embarked on just such a qualitative analysis with Computacenter.

What’s this got to do with Data Storage and Protection?

Coming back to the topic of this series, the Federated Service Provider model, customers have begun to take steps into understanding how cloud offerings might be able to help transform their environments by executing limited trials …the Daily Telegraph trial of Google Apps is a good example of this.  And yet the market has been somewhat ‘surprised’ that customers haven’t adopted cloud technologies further.

Frankly, I’m not that surprised as cloud offerings offer a different consumption model more analogous to say electricity or water delivery …tell me exactly how much this will cost me by metering my consumption …versus more traditional IT delivery models where the individual components are purchased, integrated, and consumed.

But cloud isn’t a Federated Service Provider model, so what is?

1. Defining and implementing a service catalogue is the key to unlocking the internal service provider.

We have been hearing since the 1970s that we need to align data to business value and that IT should be a service provider to the business.  The challenge is that as we have distributed systems away from centralised systems such as mainframes, it has become more and more difficult to accomplish this as we have developed ‘silos’ of technology …servers, storage, network, desktop …all operating more or less independently although interconnected.  The business says ‘give me reliable data storage, and make it as inexpensive as possible!’ and so we replied ‘Tier One is 300GB 15K drives!’ …we replied with a technical purchasing strategy that the business could understand and we could implement, but it doesn’t align data to business value.

Another way to accomplish this is the definition of a service catalogue; codifying what throughput is required by bands [e.g. 30K > 50K IOPS], the Recovery Time Objective, Recovery Point Objective, retention period, deduplication, and so on such that you can then give this package a name …we’ll call is SLA1 to keep things simple …and then give the business a flat price per user per annum.  We then keep going …SLA2, SLA3, SLA4 …and so on until we have enough packages to satisfy the business requirements.

The business gets SLA1 at £15 per user per annum, and IT becomes an internal service provider by ‘stocking’ against the requests made by the business per SLA package.

2. Virtualised datacentres allow you to automatically provision and deploy resources from the service catalogue, and introduce a fixed cost but fungible and elastic infrastructure.

I’m a geek, in case you haven’t noticed …be kind! …and I both need and want to know how data deduplication, thin provisioning, grid storage, fully automated storage tiering and the like work.  But to be fair, the real advantage to a virtualised datacentre, or VDC, is not the technical whizzy bits but, rather, that wastage is eliminated at each IT level from hypervisor through server, storage, network such that workloads can be automatically deployed and …perhaps most importantly increased and decreased elastically …without necessarily requiring any manual intervention.  Watch this space as we’re developing some demo videos to help describe this process more fully.

What the VDC does deliver, however, is a fungible resource …I don’t necessarily need to know what future workloads will look like, but I know they’ll be £1.35 per user per annum to deploy …capable of expanding and contracting without losing structural integrity.

3. I’m not moving my holy of holies into any cloud!

If we accept that only 20% of data stored is truly business meaningful …some call this structured data; ERP, customer databases, transaction processing, and the like …it is unlikely that any customer will ever go ‘all in’ with cloud storage, preferring instead to house the ‘holy of holies’ locally within their own datacentre.  And why not?  Storing this locally would help to mitigate many of the concerns regarding data security.

But how to stem the tide and control the bleeding of the data being created?

4. Help me connect my internal service provider to external service providers seamlessly.

Controlling cost as data grows inexorably is what VDCs will ultimately provide, in my opinion …through bridges to support the federation of internal and external service providers.

The corporate IT department will continue to be the internal service provider; looking after the corporate service catalogue, managing the corporate structured data housed in the VDC, and evaluating external service providers such as Computacenter Shared Service, Amazon EC2, Google, and the like.

The onsite corporate VDC will enable a customer to virtualise, containerise, and mobilise …effectively mobilising unstructured data created by users to external service providers where it can be housed less expensively than it can internally.  Ah, but ‘aren’t we just shifting the problem elsewhere and creating a tap which can never be closed’ I hear you ask?  Not necessarily, but I’ll save the possible answers to this query to next week’s post.

5. These aren’t the stack wars, they’re the stack battles …the real war will be on of APIs.

Some analysts have been watching the emergence of vendor VDCs such as VCE vBlock, NetApp SMT, HDS Unified Compute, IBM Dynamic Infrastructure, and HP Converged Infrastructure and stated that we are ‘in the midst of the stack wars’ as each rushes to gain dominance in the market for their particular ‘stack’.

I’m not convinced.  Rather, I believe the real ‘war’ will be one of APIs.

The founder of cloud provider Enomaly, Reuven Cohen, recently asked in a blog post ‘Do Customers Really Care About Cloud APIs?’  I believe that they absolutely do …and will …but perhaps not in the way that Reuven was thinking.

If I now have a corporate VDC, I will need something to ‘federate’ my internal and external service providers.  Something, in essence, to move the data seamlessly from my corporate VDC to my external provider …based on defined business rules …without disruption to my production business.  Why wouldn’t this be APIs coupled with policy based engines?  In other words, software which is smart enough to know how to connect to an external service provider …read ‘cloud’ …and move data from the corporate VDC to the cloud seamlessly through the use of a policy based engine.

But how do we keep this from becoming just a shift of the real issue to another source?

And, if this works and we start to move significant amounts of data …how do we move a petabyte of data from the external service provider back into the corporate VDC or to another service provider should we so desire?

Mrs. PL is giving me that look which means it’s time for dinner and we’ll need to save those for next week’s post but, until then, enjoy the sunshine and have a great weekend.

Accents and the language of cloud.

05/07/2010

I have always had a fascination with language, have studied several, and am always fascinated by the regional dialect or accent of places I visit.  But accents, and language in general, can be a funny thing.

Language continues to evolve, as do accents … I’m as amused to read the use of ‘bespoke‘ to describe bagels and ‘fortnight‘ to denote a period of time in American periodicals such as the New York Times, given these words weren’t widely used by Americans not more than ten years ago  …as I am to listen to my seventeen year old niece and her friends, born and raised in Edgware/Mill Hill both northwest London suburbs, use ‘like‘ as every other word in a sentence and ending the sentence with a rising inflection such that virtually every sentence is no longer a declarative statement but, rather, sounds much like a question.  Like, ya know?

I meet frequently with vendor partners, customers, potential vendor partners, analysts, et  al and, this week I met with a potential vendor partner where I got one of my three favourite ‘frequently asked questions’ when someone is meeting me for the first time.

After introducing myself … ‘Hello, I’m Matthew Yeager, Practice Leader for Data Storage & Protection with Computacenter. How are you?’ I was met with …

‘But …I thought you were American?’

‘Yes, I was born in the United States and have lived here for over ten years … I’m married to a woman born and raised in north London, my son was born here, and of my friends, colleagues, and contacts not one of them is from the USA. The United Kingdom is my adopted home.’

Now, this isn’t by design nor was it a conscious decision, it is just how it evolved over the past ten years.  During this evolution I will admit that my accent has flattened and my sympathetic ear has caused my speech to evolve such that I am sometimes asked, ‘But aren’t you [insert expected/anticipated nationality …American/Canadian/Irish/Other]?’

When my father came from his home in Dallas, Texas to London for Mrs. PL and my wedding he did remark, ‘I’m fascinated, your accent seem to be shifting towards London,’ I remarked that I didn’t think I had an accent … I was just pronouncing all of the syllables and calling the herbs not ‘erbs which is what the good people here tell me one should do.

Thankfully he saw the humour and we had a good laugh.

For the record, I’m not trying to sound like anyone other than me … and accents shift all the time, I’m afraid.  Although the mind boggles at just how Steve McClaren shifted to first Dutch when the manager at FC Twente and, now that he is manager of German side Wolfsburg …has he shifted again to German?

But I digress.

I have friends who emigrated from America to live in Israel and, when speaking English, now have a definitive ‘Israeli’ accent.  I also have friends, born and bred in the UK, who after having moved to the USA and after only a few years have begun developing broad mid-Atlantic accents and using words like ‘cell phone’ and *gasp* ‘soccer’.

If you listen to the Queen, Her Majesty doesn’t sound anything today like what she used to not thirty or so years ago.  Indeed, when is the last time you heard a BBC reporter or news reader use RP.

What’s this got to do with Data Storage and Protection?

Our customers and indeed the technology market have demanded that we evolve our language, to speak more of the business benefits of technology and how our solutions will help them align data to business value.

Equally, Mrs. PL often tells me that there is a stark difference between listening to someone and hearing them …this is usually just prior to my taking the hint and closing the laptop lid whilst we discuss the day’s goings on and PL Junior’s day at school.

It was with these points firmly in mind that I began reading a very interesting press release on Tuesday last, 29 June 2010, which reported that EMC was closing down its Atmos online storage company.  What?  The world’s largest storage vendor is closing down their cloud storage offering …but weren’t we led to believe the great saviour of all things technical …the vaunted cloud …was the natural evolution of storage?!

Not so fast.  I’m joking a bit as it always amuses me how quickly hyperbole enters any conversation regarding technology, but there may be some clues as to what is really going on here in the actual statement placed on the Atmos website.

“We are no longer planning to support production usage of Atmos Online.  Going forward, Atmos Online will remain available strictly as a development environment to foster adoption of Atmos technology and Atmos cloud services offered by our continuously expanding range of Service Provider partners who offer production services [emphasis mine].  We will no longer be offering paid subscription or support for Atmos Online services.  Any existing production accounts will not be billed either for past or future usage.   We will also no longer provide any SLA or other availability commitment.  As a result, we strongly encourage that you migrate any critical data or production workloads currently served via Atmos Online to one of our partners offering Atmos based services.”

The emergence of the federated service provider model.

I think that this is a perfect example of what is happening today within technology generally and data storage offerings specifically; our customers encourage/insist/demand that we evolve offerings and the way in which data storage is consumed to help them align data to business value, so providers emerge with the advent of the whole new language of ‘cloud’.

But the adoption of cloud has been much slower than the cloud providers would like, with some providers such as Atmos shutting up shop altogether.  And being a cloud provider and service provider are two very different things.

Why?  I think there are a few reasons, but possibly the largest is that in actuality, one could argue …and I often do …that customers weren’t looking for a new language but, rather the evolution of language …a different ‘accent’, if you will.

Indeed, I would pose that what we are witnessing is not so much the emergence of cloud storage or ‘everything as a service’ …we are witnessing the emergence of a federated service provider model.

I’ll talk more about what I think a federated service provider model might look like next week, but a few reasons why cloud storage adoption has been slower than anticipated by cloud providers.

1. What is the cost benefit of placing data in the cloud?

This is probably the biggest reason that cloud adoption for storage has been slow, in my opinion.  Search just about any cloud storage provider’s website and I challenge you to find any statements, calculators, or baseline data which shows what the average cost of storing data for a customer onsite would be versus storing the data in the cloud.  Yes, I’m fully aware that ‘your mileage may vary’ and each customer will be different …but we as an industry need to give our customers real CBA [Cost Benefit Analysis] and TCO [Total Cost of Ownership] data both from our labs and from the field to help them efficiently calculate what the cost savings would be over a five year period when transitioning data from locally held to cloud.  Given the analyst statistics purporting that only roughly 25% of technology assets are likely to be owned by customers outright by 2030, one would think that the prize would be great in helping customers understand these calculations.  Watch this space as we are actively developing these very tools within Computacenter.

2. We’re not really placing data in the cloud are we …isn’t it more like workloads?

This can get somewhat complicated so I’ll go a bit deeper on this in the next Weekly View but, put simply, we can’t just say we’re putting data in the cloud.  We must understand that this data is part of an existing customer workload …which includes compute, network connectivity, application, possibly a hypervisor, recovery time objective/recovery point objective, cost, and so on.  In other words, attributes which must be taken into consideration prior to any consideration of placing the workload in the cloud …or anywhere else for that matter.

3. Help me understand which workloads are applicable for cloud adoption.

All workloads aren’t created equal and each will have different characteristics, as I’ve just described, as well as a different cost or value to the customer business.  Whilst some of the cloud storage providers have become a bit better at publishing how, exactly, their offerings can be consumed by customers I still feel that we could do better as an industry in helping to provide customers with tools and/or collateral which helps them determine which workloads could be migrated to the cloud to offer lower cost without a degradation in performance or service …and define what their workloads look like today.  This is, in my humble opinion, the role of a service provider and exactly where service and solutions providers such as Computacenter fit in.

4. What about security?

Admittedly government mandates and related legalities meant to govern data protection and security remain a bit confusing …I say a bit, learning a new language altogether is often less confusing and time consuming …but security is not an attribute which can be ignored.  But just what is meant by data security?  Does this mean where the data is held, the physical security attributes of the service provider datacentre, data encryption …and is that encryption at the host, from host to provider, throughout the entire stream …all of the above?  This can get very complicated quite quickly, but my belief is that understanding how to solve this issue starts with workload definition as discussed in point three above.

5. How would I use cloud based storage with what I have today?

Too often I have seen cloud storage offerings and ‘anything as a service’ offerings pitched as zero sum all or nothing propositions.  But it needn’t be so, and this is the core of my belief in a federated service provider model …helping the customer to understand how and when to consume external service provider offerings such as cloud and how to use these offerings in conjunction with their existing IT.

But we’ll talk more about that next week.

Until then, have a great week and if we can’t truly enjoy the World Cup now that England are out …perhaps we can enjoy the glorious sunshine we’ve been gifted instead!

Click here to contact me.