Storagezilla

To the eXtreme, with Pat Gelsinger

Before the chaos that is EMC World I sat down with EMC COO Pat Gelsinger to discuss todays product launches and talk about EMCs product strategy going forward.

We begin with the acquisition of XtremIO and then carry on from there.

-----

Lets begin with XtremIO (Extreme-Eye-Oh), why was it important to own that?

Well, before we focus on specific products it’s worthwhile for us to look at Flash as a technology. It is dramatically higher performance in certain cases but it’s also dramatically more expensive, so the goal is to use Flash for performance while understanding that the largest amount of data is going to remain on hard disk drives. (HDD) So while Flash is a very disruptive storage technology it needs to be correctly applied. A little bit of flash can go a long way in performance.

With that in mind, we have laid out a comprehensive strategy to apply "a bit" of flash throughout the data center: in the server with VFCache, as an appliance networked into servers with Thunder, as a dedicated All flash array with XtremeIO and into all our arrays by converting our HDD based storage arrays to hybrid arrays using both Flash and HDD.

In your view, what are the differences between what EMC define as Network Flash such as Thunder and Array Flash such as XtremIO?

VFCache & Thunder are server-side offerings without array features. XtremIO is a storage array with data services like replication, native data deduplication and High Availability. Depending on the use case, the customer would choose which is a superior tool for what they’re looking to achieve. As one press article noted, EMC is the Baskin Robbins of Flash, we offer a compete set of superior technology options to cover every use case.

Baskin Robbins?

An Ice cream franchise which offers over 1000 flavours.

If the vast amount of storage will be provided by Hybrid arrays, what is left for a specialized All Flash Array (AFA) like XtremIO?

And now we can turn to product specifics, we see AFAs being used for specific and targeted use cases where extreme performance is required, such as for clustered databases, high performance trading and thin clients. Those are potential targeted use cases, are meaningful and are critically important for our customers.

Do you see this competing against Exadata from a performance standpoint?

When it comes to that we have multiple tools depending on the customer use case. Obviously one is our Greenplum Analytics Appliance, another is VFCache/Thunder in combination with our hybrid arrays, yet another is an optimization solution with DBClassify. However, XtremIO is clearly another tool to bring extreme performance to our customers.

There’s a large number of startups in the All Flash space and they’re moving in a lot of different directions. How did EMC make it’s choice in a crowded market and why move so fast?

It’s crowded because it can create so much disruption. We pursued XtremIO as we saw it as the most differentiated technology amongst everyone in that market. It’s scale out block storage designed to leverage the unique capabilities of solid state media. While early, we are quite comfortable that moving quickly allowed us to get the pick of the litter.

Some of the raft of start ups in that space will get bought up, others will just dry up and blow away but EMC is in that space and driving the disruption as it occurs.

Lets move on to VMAX. VMAX Service Provider announcement last week and a new addition to the VMAX family with the VMAX 40K this week. I’ve always thought of VMAX as ‘the mainframe of storage arrays’, who is the target customer for VMAX these days?

Unquestionably VMAX is the most powerful and trusted array in the industry, designed as a mission critical, high QOS storage product.

Mission critical databases, mainframe applications and mammoth scale environments in industries such as financial services, telecommunications, transportation, etc. have and continue to prefer VMAX in those environments.

Referring to it as the "storage mainframe" given its scale, reliability and longevity is certainly appropriate. However, given the range of the VMAX 10, VMAX 20 and now VMAX SP, we see that customers also want the high quality of service value of VMAX in other setting as well.

Thus, this family of offerings allows us to reach new price points and satisfy new use cases with the brand and quality that only VMAX has to offer.

VNX has been very well received since launch and gaining ground due to ease of management with Unisphere and things like VNXe kicking ass in the channel. What input did you have into that product family before launch? There seems to be a different approach with VNX then there was back in the CX/NS days.

VNX is a huge success in the marketplace and we expect today's announcement along with a strong roadmap to continue that momentum. Mid range competitors have felt that success and we are now the fastest growing mid range product.

When I joined EMC two and a half years ago, the strategy for VNX was already well underway. My contribution was in encouraging the team to accelerate while assuring they had the focus and support to deliver. They did and I'm proud of them for it.

Isilon, growing out of vertical segments and into primary storage. How do you see Isilon progressing over the next few years?

Isilon is on fire. It is a great product and EMC leverage has led it to 100%+ growth rates. The Mavericks release we’re talking about at EMC World will expand the use cases beyond the tradition media, medical & service providers cases that have been its hallmark to date. Given EMC's sales strength in these areas and in the enterprise, we expect its growth rate to remain extremely high.

In the long term, we expect Isilon to become the predominate NAS platform in the industry.

The last time we spoke, you were studying corporate finance and recently were on the EMC earnings call. How’s this new side of your job going?

EMC has been very good to me personally. I needed to go from silicon to system, from OEM to end user, learn about enterprise sales and support and fill in gaps like corporate finance in my knowledge base. Overall, EMC is doing great and I'm doing great at EMC. Joe has asked all of us on the leadership to step up to larger roles and its great to have those opportunities.

What are you reading at the moment?

Right now I'm readying “On China” by Henry Kissinger. A fascinating read for anyone doing business in Asia. I'm also readying “Thinking Fast and Slow” by Daniel Kahnman. On the spiritual side I'm reading The Man God Uses by Henry Blackabee.

What’s the most interesting or most fun piece of consumer technology you’ve bought recently?

I haven’t bought much lately! I guess work has been keeping me too busy.

The last piece of tech I bought was the iPad 3 or “The New iPad” to give it it’s proper title. The screen is great, the LTE support is noticeable but the thicker size/weight are a step back from the iPad 2, which is now my wife's and the gen one iPad, which is now my daughter’s. The world has clearly moved on past the PC as the nexus of innovation. For a guy who spent the majority of his career in the PC industry, it is both difficult to watch but also exhilarating for an entire new industry of innovation to emerge.

Thanks Pat!

Enjoy the show, Mark.

Data Domain Quotas & Compliance, for your health.

At home when confronted with a slow draining bathroom sink, after trying mechanical means to remove whatever the obstruction was and it not doing the job I turned to chemical means. Bucket, gloves, caustic soda. You know the drill. So, having made up the mixture I tipped it down the sink.

It was when I splashed the fingers on my right hand that I realised I had forgotten to put the gloves on.

To sum the sensation of the chemically burning flesh on my fingers up in a word, the word I would use would be piercing. The pain was piercing. And it was getting worse.

When it comes to chemical burns what you do in the first minute can be more important than any treatment you could undertake later as you’re limiting damage as it’s occurring. Within seconds I had shoved the burning fingers under cold running water and left them there for 20 minutes or so. People underestimate the amount of time it can take to wash off chemicals and how long before the full extent of chemical damage is visible. 24 hours later the area was still red & irritated looking, like a nasty sunburn, had I not treated it immediately I’d surely have lost flesh on those fingers.

Distraction & stupidity were the root causes in this case. I was thinking four tasks ahead when I should have been focusing on the one I was doing. We all have moments where we’re just too clever for our own good.

When we’re too clever for our own good in IT we go to our backups. They’re supposed to get us out of a really tight fix.

One of the many comments I heard from Backup Admins offline after my Be Cool post, where I got into the awesomeness that are Data Domain Management Trees, was

‘If we go your way and provide these MTree logical partitions to the Application teams and the Storage team and the Virtualization team, then tell them they can control their own backups and restores while using my infrastructure to do that, potentially aren’t I allowing them to fill my Data Domain appliance to the max with data which might be encrypted or just doesn’t dedupe at all?’

That’s a good question when I don’t have an answer for it. But it’s an excellent question when I do. And my answer for it is MTree quotas which we’ve announced today as part of our 5.2 Data Domain Operating System (DDOS) release.

MTree quotas allow the backup administrator to control & monitor the space utilization of the MTree logical partitions, thereby preventing negative effects to other users of the system as a result of unexpected data growth or a backup policy error.

We talked about the importance of immediate treatment, which just like treating a chemical burn is better than finding the system has reached an out of space condition due to a policy gone wrong and now everything using the system has ground to a halt.

Mtrees and Mtree quotas don’t prevent mistakes and accidents from happening. They allow you to deal with them when they have happened and do so in a way which protects other users on the system from being affected by them.

That’s important, since the uses cases for Data Domain systems expands in DDOS 5.2 with Compliance level retention. Where previously this functionality was available from EMC in Centera and VNX now it’s also available on Data Domain appliances with the Retention Lock Compliance option providing secure retention of file and email data. Along with the Retention Lock Governance you can have different classes of data retained for different amounts of time under different requirements all in the same system and leveraging appliance wide deduplication & compression savings as well as replication, encryption, snapshots and all the other system functionality.

This is the data protection infrastructure Backup Administrators can offer for Backup & Archive to Application Administrators, Storage Administrators, Virtualization Administrators and the Business Process people who define what information needs to be retained to meet compliance requirements.

And it’s the data protection infrastructure Backup Administrators can manage and report on to ensure that when asked ‘Are we backing up everything we should be backing up and are we retaining everything we should be retaining?’

They can answer ‘Yes’ with confidence.

The many faces of EMC Atmos.

While some competitors lament their ongoing failure to sell object storage systems and claim ‘it’s a market problem’ (It isn’t, the problem is that your product sucks) Atmos customers are busy getting real work done.

I’ll explore this topic further @EMCworld

The Advantages Of Flash In The Enterprise

Micron will be at EMC World next week. Here’s EMC’s Mark Sorenson and Micron’s Ed Dollar setting up some of the conversation which will continue at EMC World.

EMC World vPass

Not everyone can make it to Las Vegas for EMC World, but for those who can’t there will be live streaming over on vPass.

The first time EMC has ever streamed any EMC World content.

I always plan to do some coverage at the show but it never works out. Meetings pop up all over the place so I’m always in transit.  I can’t find an internet connection which isn’t being hammered or if I do the moment I sit down and open the laptop 20 people need to talk to me. Usually all at once.

So, I’ve given up.

I’ll be carry an iPad which won’t have a cellular connection, I’ll have a small camera and that’s probably it.

This year, no plans.

DD Boost for Oracle RMAN: Everyone wins.

Having had to spend a night sitting in a hotel (bar) with a bunch of other road warrior orphans due to shared late flight arrival and many missed flight connections I was speaking to a sales woman who’s job I found horrifying yet fascinating.

She sold Teddy Bears for a living.

Now, to me that conjured images of a Teddy Bear or two in a backpack going through customs but it turns out the Teddy Bear business is a leg breaking business. Competitors being known to resort to all sorts of chicanery to prevent stuffed product samples from making the best first impression. Love at first sight appears to be the prime motivator of sales to the target demographic of people under four feet high and it’s love at first sight which leads to unit orders that can be tens or hundreds of thousands in size. So sometimes, for potential retail customers in some places in the world, they send multiple samples using different shipping methods to ensure ‘Teddy doesn’t suffer an injury in transit.’

Yes, the Teddy Bear has a backup copy in case the primary is damaged or destroyed. By accident or design.

Having listened to all this it made what can go on in IT seem remarkably civilised. I don’t do cutesy and stuffing is there to be knocked out of people but there are some arguments that just aren’t worth having. One of the arguments I’ve been hearing more and more is how application owners aren’t using the prescribed backup processes and are doing it themselves to any bit of storage they can lay their hands on. This infuriates the Storage Administrator because large lumps of application backup data are showing up in many different places across the storage estate and frustrates the Backup Administrator who now can’t manage, or report on, all the data the organisation is holding on to. If they don’t know where it is they can’t see what it is, how long it’s being retained for, how it’s being secured and if it’s crossing any boundaries it shouldn’t as part of a replication group or whatnot.

What I’ve found is the approach at resolving these technical issues can quickly resort to people fighting their corner. The Backup Administrator ultimately answering to the data handling policy of the company, the Storage Administrator looking to optimise & minimise the storage spend and the Application Administrator looking to make sure they can recover exactly what they need exactly when they need it. The simple fact of all of this is these needs can be met with application awareness of the dedicated protection storage which Backup Administrators manage and monitor.  

With that in mind the first application to gain awareness of EMC’s Data Domain dedicated protection storage appliance is Oracle with today’s introduction of Data Domain Boost for Oracle Recovery Manager.

Oracle Administrators need more and more control over the different aspects of their environment as not only is the amount of information in their databases growing in the order of TBs, but they find they have less time to take backups, they have to satisfy retention requests for more information over the longer term and they want that data to be available online so they can get what they need when they need it and move on with their lives.

DD Boost for RMAN allows Oracle Administrators to mange the entire process of backup to, restore from and offsite DR with Data Domain all from their existing Oracle tools. Boost integrating the RMAN client with the Data Domain appliance and bringing it’s unique advantages into play, such distributed segment processing which reduces the amount of data transmitted across a network link during the backup operation by checking with the Data Domain system that the data deduplicated by Boost isn’t already stored in the appliance and doesn’t need to be transmitted again. This can dramatically reduce the amount of time required to perform a full backup (Which Oracle recommend if the change rate is more than 20% between backups) and as such can really shrink the required backup window.

DD Boost for RMAN also provides application level automatic load balancing across multiple paths to a Data Domain appliance to ensure optimal information transfer performance during the backup operation. And it does this while also ensuring failover due to a path going down will not require a restart of the backup job. The job will just continue across the surviving interfaces.

And then we come to offsite replication. DD Boost for RMAN allows the Oracle Administrator to offsite their backups to additional Data Domain appliances and do so in a catalogue consistent manner. Their offsite copies, a product of replication between Data Domain appliances, will be fully visible to them from RMAN and can be recovered from directly.

So what’s the bottom line?

The bottom line is Oracle Database Administrators can now use unique features of the dedicated protection infrastructure Backup Administrators can make available to them from Data Domain without having to write crazy scripts or learn new tools. This is infrastructure the Backup Administrator has visibility into and can report on. In turn the unique advantages of this infrastructure in deduplicated storage capacity, replica management, encryption of data in flight/data at rest and the fact it’s online negate the need to use production storage to keep backups, which releases that storage to the Storage Administrator for other uses. There’s a very simple concept at work here.

When everyone involved can be productive, everyone wins.

Leadership is about being right.

A while back at one of these tedious workshop events held with members of different companies a group of us were asked to write down who we found inspirational during different periods of history. For the early 20th century a lot of people reached for that usual management touchstone, Churchill.

Since the question had nothing to do with leadership my answer was war photojournalist Robert Capa, who put himself in situations I wouldn’t to document the world being changed one bullet at a time. Having had to explain who Capa was and what his work meant to more than a few indifferent sniffs, I tuned out of the rest of the workshop and started thinking about the nature of leadership.

Leadership, true leadership as I see it, is both nourishing and destructive for the person involved.

Regardless of if it’s a corporate environment or otherwise the idea of the reluctant leader is a myth. Ambition is required for someone to move from a staff position to a leadership role but I think it requires more than one type of ambition.

The first type is clearly personal ambition. That’s the easy one. You’ll find personal ambition, the drive to be ‘the boss’ and take a wheelbarrow filled with money home with you at the end of every working week in a multitude of people anywhere you look.

I’m sure there’s more than a few of those people who have painstakingly nurtured an image of competence built on the backs of their direct reports, who carefully crawled up the lower rungs of the managerial ladder and have a corporate password something along the lines of 1mth3b0ss but there is a missing component which prevents their transition from staff to leader. It isn’t IQ, emotional intelligence or a good singing voice.

It’s another type of ambition.

Organisational ambition. 

To me organisational ambition is a genetic component of leadership engrained on a true leader. If your total ambition doesn’t extend beyond your own success to the success of the organisation and all of those it in it then you are not capable of leadership.

If you lack organisational ambition and pursue leadership you’re an empty suit. Nothing but a walking void of wants & needs with delusions of grandeur.

A true leader removes their ego (Not their instinct) from deliberations and makes the correct decision from the facts available at the time. They get it right a lot more than anyone else.

When presented with something rotten they’ll make the correct choice on how to handle it about 70% to 80% or more of the time. But when a leader gets it wrong they’ll U-Turn like they’re in a fighter jet, not an oil tanker, because it’s not just a matter of personal image it’s a matter of being right and making the right choices.

Think about what it is to really be on top of a successful organisation, I don’t think for one second that the people knocking on a leaders office door are there to tell them everything is great and they should head home while it’s still bright outside.

I think it’s more likely that they’ve just powered off their computer for the evening when that knock on the door brings them some Pandora’s box filled with all the horrors of the world to deal with. And if they get it wrong it’s going to be very bad for a lot of people.

I’m convinced that if you look at anyone considered to have displayed great leadership you’ll find a history of on going failure until a moment when it all changed for them. When you’re successful that history of failure is called ‘experience’ and it’s experience which increases a leaders accuracy in decision making.

If you become a great leader. If you have that required combination of personal and organisational ambition. If you’ve put the needs of the organisation ahead of your own as many times as you should have and waded through a sewer of decisions making the correct one the majority of the time, what you have to look forward to if you do everything right is obscurity.

While you may never forget the company, the company will inevitably forget you. Only those who started the company, those who crash the company and those who pull a crashed company out of the ditch endure in it’s collective memory. The rest are forgotten.

And if you’ve been forgotten you’re just going to have to settle for being right the majority of the time.

(And settle for that wheelbarrow filled with money, of course.)

Data Movement At Exascale

One of the interesting technologies EMC is working on with the scientific community is PLFS, a checkpoint file system for parallel applications.

When running across thousands of CPUs on hundreds of nodes on a parallel file system, such as Lustre, you can assume something at sometime is going to fail. What HPC application authors tend to do is design their applications to take regular checkpoints to allow for application restart when such a failure occurs.

The two common ways of doing this are either to create a file per process and save your state across those or share a file between processes and save all the state to that.

However, the workload profiles of these checkpoint processes tends to choke the bandwidth of parallel file systems especially for multiple writes to a single file and you end up trading an aspect of checkpoint performance for protection.

What PLFS does is sit between the application and parallel file system and on the fly rearrange the writes in a sequence more conducive to optimising the bandwidth of the underlying parallel file system so you don’t suffer an order of magnitude loss of bandwidth.

It’s an interesting technology for the HPC market and something I’ll be watching develop.

Be Cool.

Restoring a backup of virtual machine image isn’t cool.

While you most certainly want it to work, and your organisation’s well being and your job may very well depend on it working, when it does work it’s still not cool.

Booting the backup of the virtual machine to a running state and then performing a migration off of protection storage and onto production storage as operation continues?

That’s cool.

And by nature of being cool it’s something we’re going to see more of. Not less.

That’s why random access disk won the battle for operational backup & recovery and tape didn’t. Now, people can get angry about that, which they do on a regular basis. People can blame EMC (Or me) for tape being out on the fringe, but it’s not my fault that an entire generation of suppliers, wracked by internecine warfare and pushing an increasing number of bizarre physical formats into a failing market didn’t look after people who were their customers and those customers left. If you let it that can happen to anyone in any business and in this case it did. People can even think LTFS is going to bobsleigh down hill on the corpse of LINCTAPE into a smoggy (Or is that cloudy) future, which most don’t while some do, but the fundamental truth is as an enabling technology for backup & recovery, random access media is cool.

When I look at a Data Domain system and I want to pick out cool enabling technology I look to MTrees.

Simplistically a Management Tree (MTree) is a logical partition inside a Data Domain system. What you choose to write into those partitions is up to you. Perhaps it’s your SharePoint backups, those virtual machines we were discussing earlier, the contents of Databases you bought from ‘Larry, Prince of Darkness’, or whatever else takes your fancy.

All stored in the same system but all logically partitioned.

And since they’re logically partitioned you can then replicate them differently using the 1:1, 1:Many, Many:1, Bi-Directional and Cascaded replication topologies Data Domain supports. The pre-transmission operation being either a periodic snapshot initiated by the system or a user initiated snapshot allowing for an application consistent snapshot. Either, or and all followed by transmission to the destination system(s).

In this way not only can you decide the value of the backup data by using partitioning but specify the optimum recovery scenario for that data just before it’s sent offsite.

And that’s cool.

Exploring FAST Cache Performance Using VMmark 2.1.1

It’s clear that we can replace a large number of HDDs with a much smaller number of EFDs and get similar or improved performance results. An array with 11 HDDs and FAST Cache outperformed an array with 20 HDDs without FAST Cache. FAST Cache handles the workloads’ performance requirements so that we need only to supply the HDDs necessary for their storage space, rather than performance capabilities. This allows us to reduce the number of HDDs and their associated power, space, cooling, and cost.

Tiered storage solutions like FAST Cache make excellent use of EFDs, even to the extent that 92% or more of our I/O operations are benefitting from Flash-level latencies while the EFD storage itself holds only 8% of our total data. The increased VMmark scores demonstrate the ability of FAST Cache to pinpoint the most active data remarkably well, and, even in a bursty environment, show incredible improvements in I/O latency and in the load that a cluster can support.  Our testing showed FAST Cache provides Flash-level storage access speeds to the data that needs it most, reduces storage bottlenecking and increases supported load, making FAST Cache a highly valuable addition to the datacenter.

via blogs.vmware.com

VMware take FAST Cache out for a spin and like what they find.