Jul 21

Google Data Centers – a Glimpse into our Future

Enterprise Technology 3 Comments »

5 years ago, your average data center hosting 10,000 servers looked like a linoleum covered football field,  was wicked cold, yet had plenty of aisles where you could break a sweat due to warm air-current hot-spots. In the early century, air-convection planning and floor-space reconfigurations were considered ‘green’ strategies to reduce environmental costs.Google

Flash forward to, uh now, for those who haven’t seen the video of google data centers, this is an interesting glimpse into the world of massive-scale distributed computing on the cheap and the green.

First, some observations:

What you see:

  • Lots of modular components: starting with containers, basic power and plumbing conduits, heat-exchangers, and lots of open-case x86 kit crammed into high-density confined spaces.
  • An obsessive design focus on minimizing the Power Usage Effectiveness ratio
  • Centralized cooling and power infrastructure and many non-descript containers in a warehouse rack configuration
  • Vertical distribution of gear and workload (way beyond the scale of your average data center rack)
  • As noted in the exceptionally dry video dialogue, 45000 servers across 45 containers (1000 servers per container), in a single warehouse type facility
  • (now, close your eyes and imagine Google running on over 450,000 servers world wide and tell me if you’re going to sign up for a yahoo account tomorrow)

What don’t you see?

  • No raised floor, specialized rooms or floors built for data center infrastructure
  • The classic air conditioned hockey-rink full of heat generating equipment
  • Large volumes of ambient space being cooled
  • Centralized UPS / Power Fail-Safe equipment
  • No cooling or power machines competing for compute space
  • Vendor labels (notice, no labels whatsoever)
  • Specialized hardware – in fact the hardware is engineered to be dead simple and uniform (which is relatively easy to do if your business runs massively distributed custom application code)

If you’re watching the Google drip-feed of information about their data centers, the innovation is absolutely incredible. The latest news about the chiller-less data center in Belgium, you see geography, climatology, IT infrastructure, and geographic load-balancing strategy all in play, with a goal to load-balance work away from Belgium on hot days, and not rely on a single chiller. That’s seriously more innovative than creating warm/cold air convection cells in your raised-floor meat-locker. Maybe my earth sciences degree will come in handy for IT work after all…

Next, if you take a look at the major IT infrastructure vendors, we now have a horse race to emulate big modularized x86 kit in a box, ready for rapid deployment and consumption. Sun arguably pioneered the concept several years ago with the modular data center, and now we have pods, pods, and more pods.

In the larger picture, these innovations seem to validate Friedman’s semi-futuristic speculation in Hot, Flat, & Crowded, of the convergence between power grid and IT infrastructure. Granted, it will take years, if not decades to purge legacy systems and applications, but for your average enterprise planning on ‘massively virtualized’, this isn’t a far-stretch of the imagination to see where centralized infrastructure could possibly take hold for large parts of IT infrastructure. Absolutely, there are scores of dependencies and assumptions which must be true for this to work.

The Google drip feed also shed’s light on why cloud computing is more than just a link to the next best data center. As a buddy of mine put it, ‘maybe the people who want to build their own cloud are missing the point’.

Tagged with: • • • • • • • • • • • • •
Jul 08

Nothing Says Love like Cold Hard Cash: EMC wins the Data Domain Bidding War.

Enterprise Technology No Comments »

The biggest drama in the 2007-2009 storage industry is over. NetApp loses Data Domain.

This is very bad news for Quantum and FalconSTOR, leaves NetApp in a less than desirable position to market, and puts EMC in a strong technology position.  With a capital infusion of 2.3 billion, EMC is going to lead aggressively and move more kit than we can collectively imagine with the Data Domain acquisition.

For Quantum, their main channel was EMC through a not too shabby product line up (if you exclude the Franken-EDL with Deduplication), which extended through the Dell relationship to the mid-market, with a collective wide and deep market reach. That’s going away.

For FalconSTOR, the EMC relationship presumably ends with EDL. EMC still pushes EDL due to the fact it’s a solid product and does a great job at being a non-deduplicated VTL. But nontheless deduplication for backup data storage on disk is usually a deal-maker when you do the capacity sizing and are getting 8:1 deduplication or better.

For NetApp, it’s back to square one with kind of late-to-market VTL+deduplication offering that has yet to even scratch the powerful legacy of the NAS product lineup.

For EMC, things just got a lot more interesting. EMC has acquired arguably the most ‘proven’ deduplication feature set in the field (if you could instances and years in production). While some real engineering will be required for Data Domain deduplication to legitimately play in the primary storage space, the collective deduplication, virtualization, and security capabilities of EMC (Data Domain, VMWARE, RSA), position EMC with a mad toolkit for the cloud storage game. And in the meantime, a simple and proven product offering to market in the small-mid-enterprise market for VTL/NAS deduplication is ready to roll.

Plus, EMC has done a great job of acquiring companies and not screwing them up over the last several years. That is of course if you look past the Data General acquisition and the 2nd-class citizenry of the CLARiiON line, which persists to this day as a result of an ego-driven acquisition. I’d speculate EMC plays somewhat softer and smarter these days, and is going to make this work and take advantage of the investment.

Tagged with: • • • • • • • • • • • • •
Jun 25

Why This, Why Now?

Industry No Comments »

The mission?

To research, spot, and explore trends and their real world impact as they apply to the next generation of IT infrastructure and enterprise computing.phase shift

Why this blog?

Because there aren’t many independents covering this space in a holistic way. Plenty of ‘company blogs’ but not so many IT infrastructure people. There are plenty of niche bloggers and evangelists who are super deep in various specialities, such as virtualization, storage, deduplication, data protection, and cloud, but few who are looking at the wholesale change underway for IT infrastructure. At first I thought I was late to the party but upon further inquiry realized the party is just starting…

why now?

There’s a lot of hype in IT, it keeps the spending and budgeting up.

Cloud is hype city without a whole lot of focus or clarity yet. I just look at the vast inefficiencies in classic IT infrastructure versus the radically innovative core infrastructures of Amazon, Google, etc and think the cloud gateway to these architectures will crush IT as we know it for some data types, some types of companies, in many different situations. Labor arbitrage for enterprise infrastructure management could become irrelevant over time if next generation architectures flatten out skill requirements.  Why now? It’s timing.

WAN Network architectures are fatter and cheaper. Cheap petabyte scale infrastructures are do-able and you don’t necessarily have to hack together custom built gear and code to make it work. And most importantly you have base of technologies for virtualization, grid, deduplication, all rapidly becoming commoditized, and now cloud emerging as the gateway to super efficient architecture. This is more than just the next wave of Moore’s law and new feature sets or gadget buzz. This is a phase shift.

Why should people read it?

To keep up, to contribute, to challenge ideas, to address a plethora of good questions. Because the whole idea we are used to and for many of us the basic paradigm of distributed computing “as we learned it” and is staple to our entire careers, may in fact be changing. Were talking about 20 years of one model and entering into the next 20 years of the new new new thing. And the practical impacts to us, well that’s really yet to be seen but it’s a fair bet certain roles, functions, and technologies will begin to disappear in the next 5 years as we shift gears into the next 20 year cycle of IT infrastructure.

Who should read it?

Technologists, engineers, architects, entrepreneurs, strategists, designers, scientists, researchers, and thinkers who share a common curiosity of what is going to happen to IT infrastructure as we know it.

Tagged with: • • • • • •
Jun 15

The Problem with a One Hit Wonder

Enterprise Services No Comments »

legend

A large project can make good people go deaf in the echo chamber, in many ways due to the sweat equity put into a project and the amount of attachment to the project most people can’t shake. Sometimes it temporary, other time’s it can derail entire careers. It’s human nature. In the services business, a bare necessity is to re-leverage your work and experience where possible. But where do you separate the valuable service offerings from a one-hit wonder, and how do see the difference?

The last big project often creates a fixed paradigm that all things future will fit the mold. Once you’re in the echo chamber, new opportunities take on similar characteristics. I can say this, because I’ve been there and have over time figured out some of the self-check warning signs of a one-hit wonder project:

1) The project took ‘all hands’ to deliver

2) Minor miracles and acts of God were instrumental in pulling it off

3) Majority of the work was custom

4) The customer dictated deliverables

5) The customer was ‘pleased’ with the work but can’t figure out what to do next

The problem is that the last big kill, may be a one off and if you try to repeat the experience, you run a good chance of alienating your customers and wasting a lot of time pushing the wrong solution to market. The good news is that you can sill harvest goodies from a one-hit wonder (like experience, intellectual property, consulting methods, lesson’s learned).

Tagged with: • • • • •
Jun 14

Phase Shift

Industry No Comments »

In this market, we are seeing striking change not only in customer behavior, but in the global competitive landscape for enterprise technologies and services. Recently this dawned on me: the IT infrastructure market is changing faster than we may know. A severe market downtown, survival driven customer behavior, increased global competition, and multiple disruptive technologies hitting the scene simultaneously,  all pointing to a fundamental shift in our game.  My position?

Traditional IT services methods for selling and delivering will quickly become a thing of the past, and to compete and grow in the market, innovation and measured-risk are prerequisites.

The coalescence of virtualization, deduplication, cloud infrastructure, and maturing WAN architectures is collective game changer. The whole idea of customer data sitting in the data center forever is going to change.  For now, companies are taking these technologies on one by one, but it’s not long until the combined force of these will yield truly innovative products and services.

This blog is about what we do to evolve our methods, and in doing so keep up with a big shift in underlying infrastructure technologies.

Open kimono, and why be paranoid about divulging good ideas, when you might receive good ideas in return?

Tagged with: • • • • • • •
preload preload preload