2.031: manual or automatic?

My, but hasn't EMC's FAST announcement generated a lot of discussion this week?

I've been very busy out in the land of social media, answering questions on Twitter and clarifying things for a broad range of bloggers – most supportive, others well, not so much.

In the midst of a rather respectful tete-a-tete with Pete Gerr over on his HDS "Ars Indicium" (the art of information) blog, I suddenly had a revelation about what distinguishes Symmetrix vs. the USP-v (et al).

It's the different approaches we each take to addressing customer problems.

Now, to be sure, we actually sell to many of the exact same customers, often competing head-to-head for business. So you'd figure we're both seeing the same requirements from these customers. We each have our own well-established technology and storage platforms, and we both get our drives and components from pretty much the same place. I'll stop short of saying that our software does the same thing, though, because this is where it is that I suddenly realize we really differentiate.

But it's not what you think. No, it's not about Virtual Provisioning vs. Dynamic Provisioning or SRDF vs. TrueCopy.

I think it is really something more fundamental than that: Hitachi Ltd. builds tools that customers can use to solve problems, while EMC provides automation to solve those same problems.

We're automatic, they're manual.

Let me explain what I mean…

manual vs. automatic

My revelation was borne of a discussion with Pete about whether or not Hitachi had "delivered the promise of FAST first." Pete referenced a post from Hu Yoshida where Hu asserted that Hitachi's Tiered Storage Manager did the same thing as FAST – moved data to different tiers. (If you read Hu's post, you'll see that he and I actually had a respectful dialog as well – surprising as that might be)

My challenge back to Pete was that TSM, for everything it could do, was not AUTOMATIC. TSM requires users to set up monitors and then review results to determine potential candidate LUNs for relocation – an entirely manual process (except for the actual move). FAST, on the other hand, is all about Fully Automated Tiered Storage – it can do the monitoring and analysis against admin-defined policies, and then effect relocations – all entirely on its own.

It was when Pete admitted that TSM indeed wasn't automatic that it hit me, for his response concluded with an assertion that Hu had also made:

"That doesn’t mean they [customers] want automation or additional products. I believe (the majority of) customers want to retain control over these migrations…

As you'll see if you read through the comment-debate at the end of Pete's post, he eventually accepts that automation is important, but that he thinks

"HDS believes customers do want and need some degree of automation; I think the difference is, again, the perspective with which we view the market, our customers, and the “right” time to introduce innovation."

He later continues:

"What I think is that HDS might hear what those customers are saying in a slightly different way than EMC does, and certainly would say that HDS reacts differently, not worse or better, not slower or faster, to what we hear."

That's when it hit me…

For the past several years, EMC has been cutting a path of automation through the challenges of the high-end market, while Hitachi has been selling path-cutting tools.

simple, intelligent modular storage

That's actually the vision statement for the Symmetrix development organization, and it's an aspiration that our customers have validated repeatedly over the past several years.

The track record of automation-optimized enhancements is notable, and each can be contrasted to the more complex, manual approach that Hitachi's engineers have taken with their implementations. Here are but a few:

• Symmetrix Dynamic Cache Partitioning – Symmetrix cache partitions can be dynamically created and resized under full workload and they uniquely allow the system to temporarily contribute unused cache across partitions (in accordance with policies). This automated sharing ensures that all memory is fully utilized, while still maintaining minimums to meet the needs of applications whose I/O patterns change over the course of a day.
• Symmetrix Priority Controls – similarly, relative priorities for specific LUNs and/or groups of volumes can be defined (as a policy), and the system will automatically assign resources and reorder work queues accordingly. These apply not only to disk I/O contention, but also to remote replication, ensuring the most important applications I/Os are serviced at a higher priority than others. And in the mainframe environment, Priority Controls integrate with SMS/WLM (I think Hitachi's may as well, not sure).
• Automated Storage Provisioning – since its introduction earlier this year on V-Max, virtually everyone who has used Auto Provisioning has acclaimed its logical and easy to use implementation – simplifying the task of creating, allocating and masking storage to server clusters (such as an VMware ESX cluster). The automation takes a process that requires almost 500 clicks using the Hitachi management GUI down to just 15 clicks on V-Max by leveraging automated discovery of LUNs, host initiators and ports and an intuitive user interface. The constructs are also extensible, allowing the addition (or removal) of new initiators, ports or storage to the relationship with just a couple of clicks (the process can be scripted by the CLI in a very logical manner as well).
• Secure Drive Erase – Symmetrix DMX and V-Max customers can elect to have drives automatically erased using DoD standards as part of the drive sparing process. Once a failing drive has been hot-spared, Symmetrix will perform data overwrite with patterns as per several standard specifications, ending with writing all zeros – all before phoning home for a drive replacement. The service enigneer can present the customer with a log showing whether or not all the data was able to be overwritten before taking the drive off-site (and for a small fee, these logs can be certified, where required).
• Virtual Provisioning – automation for Symmetrix VP allows storage admins to create virtual storage pools and to allocate LUNs (thin or pre-allocated) from them with about 80% fewer clicks and in about 1/3 the time required to perform the same task on USP-V. (I'll have more to say about this in a few days…now THAT'S a hint
• Fully Automated Storage Tiering – And finally, indeed, the key word here is FULLY AUTOMATED, where Hitachi's approach is anything BUT automatic.

I could go on, but I think you get the point.

Over the past couple of years I've challenged Hitachi's products on many occasions. It seems clear to me that Hitachi's developers afford customers with lots of manual tools to use with their arrays, often beating competitors to market with a feature. But in a world where information growth is 50, 60 or even 70% compounded annually, and in an economy that has significantly slowed (or eliminated) headcount growth in IT, the manual approach simply doesn't scale. Customers can't afford to spend hours allocating storage to an ESX cluster, and they don't have the manpower it would take to continuously monitor and hand-optimize their storage layout – the success mere existence of products like 3PAR and XIV are a testament to this new demand for Simple, Fast and Easy.

And I can assure you that the Symmetrix development team not only understands this, they have committed themselves to it – the Symmetrix of today is significantly simpler, easier, faster and more intuitive to manage than it was back in 2002. And I can assure you that it will keep on getting better going forward…because I am convinced that customers do in fact prefer automatic to manual – both for their automobiles and for their storage arrays!

 

technorati tags: EMC,Symmetrix,FAST,Fully Automated Storage Tieringi,Priority Controls,Cache Partitioning,V-Max,DMX,Virtual Matrix,Automated Storage Provisioning,Secure Erase,Virtual Provisioning,Hitachi,USP-V,Dynamic Provisioning

Comments

[Ex-HDS person]

Barry, correct at last - surprised it took you so long...

HDS provide techno-tools and then provide training and (variable) PS to set them up.

Often training courses and PS disagree on best practice so no two customers do things the same way....

[the storage anarchist]

Yeah, I guess I'm a little slow - thanks.

[Aran]

Barry,
That was a good discussion with Pete Gerr, I get tired of the petty bickering among competing bloggers and enjoyed the civil debate. Please do this more often; just because a competitor/blogger gets snippy with you doesn't mean you should lower yourself to their level. Stay above the petty bickering and continue to dispense with the solid analysis and comments that brought me to your blog in the first place.

As a happy EMC customer I have to disagree with you on one point. I like automation in my storage arrays, not my cars :-) I love to run my high performance turbo-charged car through all of it's gears on the winding mountain roads where I live, and hear the engine roar at high RPMs as I viciously push it through tight corners. Direct engine control FTW

Aran

[the storage anarchist]

Thanks for the excellent advice Aran.

And I was hoping someone would correct me on the point about the cars ;0)

[Jerry Zeisler]

Barry,
As a current EMC employee, when I get my 30 minutes or so in front of a customer, automation in EMC storage management tools is ALWAYS my topic of choice. Having spent 3.5 years with HDS prior to EMC, storage automation is without a doubt a key differentiator between us and Hitachi. And I discuss those very features you mention in your post. Most customers get it.