Posts Tagged ‘Zero Page Reclamation’

I need to learn Norwegian!

14/08/2009

I need to be brief with this Weekly View as it is PL Junior’s birthday today, and Mrs. PL has made it quite clear that I am required to help setup for his party …and ensuring I don’t bore our guests with analogies of birthday cake slices and how they remind me of data deduplication.

But with guests coming over later this evening, I got to thinking again about our recent holiday.  As I said last week, this is the second year running that we went to the same resort in Malta on holiday.  Last year PL Junior as just learning how to swim and, as he was just turning two, wasn’t yet to the point where he would play with children he didn’t know.  Equally, as only about twenty percent of the folks holidaying in this resort are British, there were many children speaking a multitude of languages that PL Junior didn’t understand.  This year saw many changes, from PL Junior spending nigh on six hours a day in the pool and, interestingly PL Junior developed a relationship with three Norwegian children whom he played with non stop.  Did PL Junior suddenly become a fluent Norwegian speaker?  Nope, didn’t understand a word of what the three children aged three, five, and nine were saying …nor they him …but between my German, my wife’s French, and the Norwegian parents’ English we were able to find enough common ground to enable us to go out to dinner together and generally enjoy one another’s company.  PL Junior continues to ask for Markhus, Khristof, and Nikolay and when we are going to go back to Malta to play with them …and Mrs. PL and I have a standing invitation to come to Norway with PL Junior to ski in their family chalet.

What does this have to do with data storage and protection?

The sale of enterprise [e.g. EMC Symmetrix] and modular storage [e.g. EMC CLARiiON] ‘crossed over’ during the last recession in 2000 / 2001, with modular storage sales overtaking enterprise for the first time and, since then, modular storage sales have continued to rise whilst enterprise have continued to decline.  There are many reasons for this, but customers realised that they could achieve similar performance and reduced complexity with modular storage at a price significantly lower to enterprise.  We are about to see a similar ‘paradigm shift’ as enterprise and modular storage converge.  You can tell your friends that you heard it here first, and there are storage products on the market today which represent the future converged enterprise/modular arrays as I’ve described …think IBM XiV and EMC Vmax …and trust me when I tell you that other major vendors such as NetApp, HDS, and HP aren’t far behind.

As the storage arrays converge we will be able to provide customers performance, high utilisation, and decreased management complexity without sacrificing quality and, perhaps most importantly, at a competitive cost.  But as this convergence occurs, how will customers be able to differentiate one solution from another to satisfy their business needs …or will they all look the same?

No, I can assure you that, whilst they will seek to achieve the same outcomes, how they get there will continue to be different.  IBM XiV solves the grid architecture question by using a massively parallel SATA array and using software to allow storage to be ‘written‘ across all the drives thus giving high performance to all applications equally …whereas EMC takes a different view with Vmax useing mixed solid state drives, fibre channel drives, and SATA drives with software to ‘promote‘ and ‘demote‘ storage as application demands require.  Equally, storage vendors will seek to differentiate their products with features …whilst they will all provide thin provisioning, vendors are now going to war re how their thin provisioning works.  To wit, they go to great pains to explain ‘chunklets’, or how large a thin provisioned block they will use.  It is now being argued that use a chunk too large and you may negate the very applicability of thin provisioning long term, whereas use a chunk too small and you may overrun the onboard storage cache and negatively affect performance.  HDS uses 42 MB chunk sizes to allow for their thin provisioning solution to execute zero page reclamation, thus ‘reclaiming’ up to 30% of previously allocated storage back into the storage pool as fully usable …thus equating to cost savings …whilst EMC uses 768K chunklets as this is optimised for performance on their systems.

Who is right and, more importantly …what to do?  Firstly, they are all correct …and all incorrect!  The solutions will all technically ‘work, however what matters is what the customer is trying to accomplish and the business issue(s) we are trying to solve.  This is where Computacenter comes in as our vendor agnostic yet vendor selective consultancy practice enables us to firstly understand how each of our vendor partners accomplish things like grid architecture, thin provisioning, and chunklet size …and secondly, which vendor or vendors can demonstrably solve the customer business issue(s).  Just like PL Junior and our new Norwegian friends, there are folks who can translate what is being said at the bits an bytes level into real world cost reduction and the solving of business issues.

What sometimes happens, however, is that the most applicable solution could actually involve more than one vendor …but as we’re trying to reduce complexity, not introduce more many customers decide to go ‘sole vendor’.  Again, this is where Computacenter come in as …just as we parents helped PL Junior and the Norwegian children understand one another …Computacenter can provide consultancy, implementation, and support services which span multiple vendors thus providing customers with the best of breed solution at a fair cost without increasing complexity.

Mrs. PL and PL Junior have already stated their intention for us to holiday with our new Norwegian friends next year, and I plan to work a bit more on my Norwegian before we go away.  In the interim, please don’t hesitate to contact me if I can be of any assistance in helping to articulate the ‘why Computacenter’ for you  …our customers.

Have a great weekend.

-Matthew

Click here to contact me

Advertisements

Is optimisation really that difficult?

15/07/2009

I absolutely love working for Computacenter generally and what I do for a living specifically.  I get to work with some amazing people, meet with interesting customers, work on challenging solutions, see fascinating technologies from vendors …what’s not to like?!  That said, I know that the most important job I will ever truly have is that of father and husband …and that time is the most precious and finite commodity we have.  I can easily and without reservation name the two best days of my life thus far; getting married to my wife Sarah [aka Mrs PL] and the birth of our son Louis [aka PL Junior].

I’m quite sure that I am not alone in this …and I always smile when someone boots up their laptop and their family is smiling back as the screensaver, but let’s be honest …none of us really truly knows how much time we have, and yet we often spend it like it was in limitless supply.

Now, I sometimes get the balance wrong, but truth be told I would like to spend more time with my family as I’m sure we all would.  Having a child changes things forever, and I have noticed that I am much more ‘precious’ with my time since PL Junior was born almost three years ago.  Indeed, I am always on the lookout for new and more efficient ways of doing things and one of the biggest lessons I’ve learnt in the past three years is that trying to find a half day here or a full day there to ‘make up’ time is next to impossible and has a low probability of success.  What does seem to be possible, and I’ve had much success with, is finding more efficient methods of doing things which ‘give back’ 5 minutes here or 10 minutes there …add them up and you will find the half day, full day [or more] that you were looking for.

Don’t believe me?  Why don’t people use VCRs anymore …because Sky+ is much more efficient and better quality.  When is the last time you heard the digital squelch of a dial-up modem …broadband gets faster with each year, and WiFi has made access remarkably simple in the home …and the time we spend on the internet trying to get information that much more efficient.  Do you still have a Sony Walkman or play records on a turntable  ….nah, me neither as iPods, iTunes, and Spotify have made downloading and listening to music ridiculously easy and far more efficient than popping down HMV to buy the latest ‘Take That’ album which will only play tracks in succession anyway.

What does this have to do with Storage and Data Protection?

I have been amused recently to have some vendors remark that they have heard I am ‘in love’ with this or that vendor’s storage technology.  For the record [no pun intended], I’m not ‘in love’ with any technology …I don’t give a badgers backside if vendor A can move a data bit faster from Edinburgh to London faster than vendor B.  Truth is ‘speeds and feeds’ are not really what impress customers, and I can see why vendor A’s solution would be applicable for a customer under certain circumstances and vendor B’s solution would be applicable given others …it is futile, in my opinion, to turn such things into ‘religious’ arguments …I’ve got a religion, and it’s not storage.

What I am ‘in love’ with is how a vendor solution can demonstrably reduce a Computacenter customer’s costs and optimise their business infrastructure.  The recent launch of our Sharpen Your Business initiative is hugely useful to our customers as it shows them how to demonstrably lower their IT costs without sacrificing functionality or putting their business at risk.

Picking up on this theme, I would argue that our customers in the current climate are both carefully controlling their expenditures as well as seeking ways to gain back time …in other words, trying to make sure that every minute they spend is useful to the bottom line without sacrificing quality or customer service with their customers in turn.  Equally, I’m quite certain that were we to walk into a customer meeting and state ‘I can show you how you could spend more time with your family without sacrificing any quality at work, in fact you might be able to increase your quality’ we would get their attention!

We often talk about the ‘what’ when it comes to technology …and I’ll put my hand up in that sometimes it can be confusing as to what, exactly thin provisioning, virtualised backup and zero page reclamation are for example.  Watch this space as I am working on some collateral which should, hopefully, make this a bit less confusing.

More importantly, however, we’ve begun to speak in earnest about the ‘why’ to consider implementing technology, for example the Sharpen Your Business initiative…and, to be fair, the decision makers in our customers want to know both why from a cost benefit perspective and from a return on investment perspective our solution makes sense to implement now.  We’ve plenty of collateral in this area, and again …watch this space as more will be released over the coming weeks / months.

I’d like to also see us ‘cut to the chase’ with customers around key storage technologies and how they can save them time by increasing their efficiency, understanding that the collateral exists regarding the ‘what’ and ‘how’ should the customer wish to know more.

Why automated storage provisioning? Because we can automate highly repeatable storage provisioning tasks, allowing you to reallocate storage admins to more meaningful work …because we’ve saved you time.

Why virtualised backup? Because we can optimise your backup environment from a pure tape environment to a virtualised backup environment, allowing you to reallocate backup admins to more meaningful work … because we’ve saved you time.

Why thin provisioning? Because it is a more efficient method of storage allocation and will allow you to buy far less storage following implementation and perhaps no new storage required for several years to come …and because we’ve saved you time.

Why data deduplication? Because you are storing band backing up multiple copies of the same pieces of data, and data deduplication can remove all of this inefficiency allowing you to reallocate backup and storage admins to more meaningful work … because we’ve saved you time.

Why ZPR [Zero Page Reclamation]? Because we can give you back 30% of the storage you think you are using but aren’t actually, thus saving you money and giving you in year ROI.

Just a few ‘starters for ten’, and comments always welcome for more!

Time is precious, let’s help our customers save every minute they possibly can by optimising their datacentres.

What comes after books?

14/07/2009

I recently received a Sony eReader [PRS-505] from Mrs. PL and PL Junior for Father’s Day and, knowing me as you probably do by now, I love gadgets and anything technical that will help me save time.  However, I must admit that I was incredibly sceptical when I received the eReader at first.  Now, I had read more than one article which pitted the Sony eReader against the Amazon Kindle and other like devices …but not matter how greatly improved they said the electronic ‘ink’ is which comprises the secret sauce of these eBook readers, I am very much in love with the visceral experience of reading a traditional paper book.  The sound of the pages turning, the smell of the paper, the different fonts each publisher chooses to use, the sense of accomplishment of looking at a 900 page tome and thinking ‘I read that!’ …not easily replicated in the world of eBooks.  I know what you’re saying …but you’re a data guy!  You are constantly banging on about how to increase efficiencies with storage!  Yep, I know …all arguments that Mrs. PL had used with me before when she tried to convert me to the ways of the eReader.

More than that, Mrs. PL also likes to read …a lot …and between the pair of us I suppose I have lost sight of just how cluttered our house has become with books we’ve read.  Always happy to lend or give to a friend, would never dream of throwing a book out.  Throwing away a book is tantamount to sacrilege in both Mrs. PL and my extended families, but I think the warning signs that we had too many books to be sustainable came when we contemplated building PL Junior’s cot out of used books [but you could use a hardening shellac to make the pillars from old hardbacks!] and Mrs. PL looking at five bedroom houses [we live in a 2.5 bedroom house at the moment in North London] …not because we have other little PL Juniors on the way, sadly, but because Mrs. PL was seriously considering converting a bedroom or two into floor to ceiling library space.

Thankfully Mrs. PL was the realistic and rational of our little tribe and made the leap to the eReader …gave me one for Father’s Day so she could guilt me into using it if needs be …and then proceeded to cull our house of better than 80% of the books we’ve read and don’t need any longer.  Fifty seven black bin bags full of books, at last count.  Has it made a difference?  Who knew that we had a dining room?!

What does this have to do with Storage & Data Protection?

I’ve had the eReader for a little over two weeks know and, truth be told …I love it.  Seriously.  It’s not perfect, and it does have some drawbacks [no native support for my MacBook Air …although I have a cunning workaround!] but being able to download the next ten or more books I want to read to the eReader is hugely useful and doesn’t take up any more space than the physical device does already.  Equally, after having used the device for a while, I am utterly convinced that this is the future for many magazines and newspapers …think about it, instant and automated delivery via your existing home or corporate WiFi connection …the device already knows what you have and haven’t read so it can automatically delete/replace based on your preferences …and all without the clutter of unread/read newspapers and magazines [also a bit of a problem in the Yeager household]!

When it comes to data storage, we have a very similar challenge.  Five or more years ago, heck maybe up to only a year ago …simply adding more storage capacity was an oft pursued storage strategy …and deemed perfectly valid when we had shed loads of cheap datacentre space and power wasn’t an issue.  Indeed, in the heady days before the recession many of our customers had a desperate need to stay ahead of their competition and so adding more capacity was the order of the day …trouble was, much of the capacity that was added was tier one / high end / monolithic / enterprise [however you wish to describe it!] storage.  Now, nothing wrong with this type of storage …but we don’t need it for 100% of our infrastructure!  We know that the average customer environment is comprised of 20% structured data [the important stuff …you know, the data which makes us money!] and 80% unstructured [MP3s, old or duplicate spreadsheets, joke emails …you know, the data which COSTS us money to store!].

I don’t think in the span of my career have I seen customers so evenly split into two camps …those that have run out of space and power due to continued purchase and provisioning of high end enterprise storage alone …and those that will.

Now, our solutions aren’t designed to go in and tell customers ‘you’re doing it all wrong!’ or that you need to ‘rip and replace’ …no, our solutions are designed to save customers money whilst helping them utilise assets they already have.  I wouldn’t dream of ever calling someone’s baby ugly, Mr. Burton!  But how do we articulate our solutions in a way which will resonate?

A few examples for those that know they have a problem NOW:

Problem: I’ve run out of datacentre space, my datacentre power is costing me too much …but I need to find more storage space with a tight budget!  And I have been buying nothing but enterprise storage and/or have more than one storage vendor already!

Solution: Let’s look at storage virtualisation such as HDS USPV or HP XP  to help us consolidate without disrupting the production business…we can create virtual storage pools universally available for our server hosts without needing to worry about what vendor badge the existing storage arrays have …migrate to thin provisioning so we only allocate what we are physically using …and give you in year ROI because we can use Zero Page Reclaim [ZPR] to return 30% or more of their existing allocated storage into usable unallocated storage.  And finally we automate the storage provisioning to greatly reduce ongoing manpower costs. Et voila!  Reduced power, cooling, space, and manpower whilst enabling business innovation to continue!

Problem: I’m running out of datacentre space as I’ve too many servers, my power and cooling for the servers is costing me too much, I can’t afford to hire more administrators to look after more servers …but I need to add more servers with a tight budget!

Solution: Let’s look at consolidating all of the existing server shares to one or more NAS devices such as NetApp, HP Lefthand SAN, or IBM N series …we can then create a universal NAS namespace using F5 Acopia so we can manage the NAS devices as one and make the NAS storage universally available for server hosts …and then virtualise the remaining servers with automated provisioning of storage AND servers to greatly reduce ongoing manpower costs.  Et voila!  Reduced power, cooling, space, and manpower whilst enabling business innovation to continue!

Problem: I’ve run out of datacentre space, my datacentre power is costing too much …I need to reduce costs and find more storage space with a tight budget!  And I’d like to have access to tier one functionality as I may need it …but I’d prefer tier three pricing!

Solution: Let’s look at consolidating the storage to a grid storage architecture such as IBM XiV …we can migrate from existing tier one arrays to a grid storage architecture with tier one functionality but at greatly reduced cost, up to 25% of what tier one would normally be in some cases …and it uses SATA drives, which consume 97% less power and cooling over traditional tier one FC drives …whilst also giving us thin provisioning so we only allocate what we are physically using …and give you in year ROI because we can use Zero Page Reclaim [ZPR] to return 30% or more of their existing allocated storage into usable unallocated storage.  And finally we automate the storage provisioning to greatly reduce ongoing manpower costs. Et voila!  Reduced power, cooling, space, and manpower whilst enabling business innovation to continue!

I could give you more examples, but I’m running out of space [and you may be running out of patience!] so I want to talk about the other half of the customer stack …those who haven’t run into the same problems …yet.  What do we say to them?

Well, Mrs. PL and PL  Junior went out and bought me a tool which they knew would allow me to keep only the 20% of books I truly need whilst giving me a much more efficient way to read and store the other 80%.  They knew there was a problem there, although I was loathe to admit it …but am very much loving our ‘newly’ discovered dining room in addition to my eReader!

We need to engage our customers to discuss their business problems now and not wait  until they issue an RFP.  What if they say ‘we’re not bothered about storage optimisation at the moment as enterprise tier one storage from [insert vendor here] is so cheap it’s more cost effective for us to add to our [insert vendor here] arrays’?

Fair enough.  On balance, over the next month or three months or even six months they might be right.  But what about after that?  Do we know what their 20% of structured data is and how fast their storage is growing per annum?  Can we extrapolate these figures to show them when they will run out of datacentre space and/or be consuming more power than a small Yorkshire village?  Can we marry this data to the total cost of ownership manpower costs to show them how expensive this storage will be to maintain in the future?

You bet we can.  And we absolutely should.  We won’t win them all, but I would be surprised if we didn’t find something that we can help them with …automated storage and server provisioning, data deduplication, reducing backup windows, consolidating server shares, virtualising their servers …to name but a few.

I’m up for it if you are … and I bore everyone I meet at cocktail parties, wine tastings, Waitrose, my dry cleaners something silly with how proud I am to work for Computacenter and how insanely great our solutions are.  Let’s get out there and talk to all of our customers about how we can Sharpen Their Business and save them money by optimising their storage now.

So you want to thin provision storage …

12/07/2009

Did you know that the human brain doesn’t actually need you to spell correctly in order to discern context?

I cdnuolt blveiee taht I cluod aulaclty uesdnatnrd waht I was rdanieg. The phaonmneal pweor of the hmuan mnid, aoccdrnig to a rscheearch at Cmabrigde Uinervtisy, it deosn’t mttaer in waht oredr the ltteers in a wrod are, the olny iprmoatnt tihng is taht the frist and lsat ltteer be in the rghit pclae. The rset can be a taotl mses and you can sitll raed it wouthit a porbelm. Tihs is bcuseae the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a wlohe.

What does this have to do with Storage and Data Protection?

Firstly, please allow me to apologise if you had any challenges reading the previous paragraph as I am categorically not trying to take the michael as regards dyslexia or related conditions.  Indeed, my sister is dyslexic and I have nothing but respect for the way that she hasn’t allowed this to slow her down in life or studies in the slightest.

But the question remains, where is the link between the opening paragraph and data storage?

The best storage device yet invented is the human brain, although admittedly it sometimes isn’t as performant as we would like after a night on the tiles, and we have much to learn about storage specifically and technology in general from studying neuroscience.

You may recall that I have talked about thin provisioning in this blow previously, affectionately known as the ‘Northern Rock’ method of storage allocation internally here at Computacenter, thin provisioning is a block level storage technology which allows us to more effectively control the storage allocation to hosts by only allocating [or provision, hence the term thin provisioning] that which is physically required.  Great.  That clears it up, simple right!  See you next week.

Only kidding …what that means in English is that we know that data is really only zeroes and ones [‘on’ bits and ‘off’ bits, actually], so who cares if it is a PowerPoint or Excel spreadsheet or joke email …what we care about are the zeroes and ones of the data, which is what we mean when we say ‘block level’.

Okay, now we move on to ‘fat’ provisioning in traditional storage environments …an email administrator tells the storage administrator he needs 10 terabytes [10,000 gigbaytes], so the storage admin adds a little ‘fluff’ just in case and we’re now into say 12 terabytes …and here’s the thing …the email admin won’t use the entire 12 terabytes in one go, he’ll use that gradually … in actuality he asked for the storage he reckons he’ll need for his users over the next year or more.  But the 12 terabytes has been allocated to him and cannot be used for anyone else.  Multiply that over multiple storage arrays and multiple hosts and you can see just how quickly storage is traditionally allocated which won’t be used for a considerable period of time …if ever …but locked into a single host allocation unable to be used for anyone else.  This certainly puts a big dent in utilisation!

Wouldn’t it be great if we could let other people have access to the storage we’re not using ?

Well, yes …and that is primarily what storage virtualisation is all about …and thin provisioning is a great feature of storage virtualisation.  A feature, mind …not a product …and the better vendor storage offerings have storage virtualisation built in as a feature so that you don’t have to go out and but a separate product or products to facilitate this.

Put simply, thin provisioning will tell the server that I have access to however many terabytes I have requested but the storage array will only dole out the zeroes and ones the server needs as it needs them thus freeing up a serious amount of storage for others to use.  And before you go worrying about it, we monitor how quickly data is being created to ensure we aren’t caught out with everyone requesting the storage they believe they are entitled to simultaneously thus crashing the system …a la Northern Rock.

Fandabbydoozy!  Everyone should be using this, right?  Absolutely.  In fact, buy three and we’ll have a great quarter!  But Houston, we have a problem.  Budgets are tight, and storage virtualisation and thin provisioning are predicated on the avoidance of future spend …and many, if not all, of our customers are looking for in year ROI …game over?

No, it doesn’t have to be …Zero Page Reclamation [ZPR] to the rescue!  Oh dear …what on Earth is ZPR I hear you shout.  Remember the first paragraph and the fact that our brains only need to see the first and the last letter to discern context?  Thought I’d forgotten about that, didn’t you?

Storage is allocated in a ‘fat’ provisioned traditional environment much in the same way …the array ‘writes’ a zero page to the beginning and the end and leaves the bits in between blank so that it remains allocated to the server host and won’t be accidentally allocated to someone else.  But we don’t want that as it is inefficient, so we buy an array with thin provisioning on board but …and here’s the kicker …if we don’t clear those ‘zero pages’ when we migrate from fat to thin provisioning [shouldn’t that be called a storage diet migration, by the way?] the new thin provisioned array will view the migrated data volumes as ‘full and allocated’ even though we know this isn’t the case.

Enter ZPR which ‘erases’, or removes if you prefer, the zero pages and returns this storage back to the pool.  Hey presto …in year ROI, money for old rope, instant storage!

There are some arrays which do this for themselves when you migrate to them alone [e.g. IBM XiV] but there are also arrays like HDS USPV which allow us to virtualise storage from heterogenous arrays [i.e from other vendors] into a pool, thin provision, and use ZPR to reclaim upwards of 30% of storage which would have otherwise remained ‘allocated’.   ZPR gives us a great way to not only help our customers control storage spend in the future, but use their existing assets to achieve ROI now …today …immediately.

I can’t think of a more powerful solution to ‘kick the door in’ when we’re talking to our customers about why Computacenter and not someone else when it comes to their storage partner.

I’ll be running a webinar re ZPR soon and a Masterclass in August to help you understand storage virtualisation more fully, but please don’t hesitate to contact me or your friendly neighbourhood Solution Specialist if you would like to position storage virtualisation and Zero Page Reclamation with your customer.