What if ARM isn't more efficient than the other Server processors?

Wired has a post on why Google will embrace cell phone chips (aka ARM).

I've blogged on the concept of little green servers built on Atom and ARM, but what if ARM isn't more efficient than an x86 server?  Here are some thoughts that are running through my mind.

1. ARM is built on RISC which is more efficient
1. The ARM architecture describes a family of RISC-based computer processors designed and licensed by British company ARM Holdings.
2. Intel has not stood still and its processors are focused on efficiency and they have embraced the micro server category with the Intel Atom
1. 
2. And focuses a lot on migrating users from RISC to Intel. http://www.intel.com/content/www/us/en/risc-migration/server-migration-transition-to-intel-based-solutions.html
3. Which would make you think Intel has focused on how it can outperform RISC when there is heavy IT loads.
3. Part of the ARM energy savings for mobile is the ability to shut down energy consumption during idle times.  But, if you have a highly utilized server with many VMs on it, when is the idle time?
4. IBM has shipped RISC server chips for years and it works for IBM well enough that there are no rumors of them to switch to ARM.
1. The POWER7 superscalar symmetric multiprocessor architecture was a substantial evolution from the POWER6 design, focusing more on power efficiency through multiple cores and simultaneous multithreading (SMT).^[6] The POWER6 architecture was built from the ground up for frequencies, at the cost of power efficiency and achieved a remarkable 5 GHz.
5. HP had PA-RISC.  Sun/Oracle has SPARC.

Intel's biggest margin business is its server processors and they will do anything to defend its market share.

The way some people write it is absolute that ARM's arrival will replace x86 servers.  What if they are wrong and x86 has a price performance that meets the market needs. Intel can survive longer than most of the ARM developers, except Samsung.  Samsung are the one guys who you don't want to under estimate.  Now if you think about Samsung being able to create new always-on servers for things like your home, media centers, cars, small offices, then that is different.  When you think of all the disconnected ARM processors in your home in your Routers is one example.  Should there be a better home server appliance that can connect your home devices.

ARM's potential in servers may be in new markets, not in the data center.

Now that Facebook has announced Iowa, where will they build next?

I forgot to post about the Facebook Iowa data center, and just sent this up a month late. :-)

The next interesting question is where Facebook will build next.  I have some theories and I'll send some e-mails to some buddies and we'll chat about it next week.  Then we'll see if I was right or wrong in about 6 - 9 months.

Speculating about Facebook's data centers in the media just makes things worse for everyone, except the media companies that are only interested in traffic.  I don't think anyone has found that an early media disclosure of their data center plans has helped them.

So what's in this new Facebook Iowa Data Center?

I meant to post this, but forgot to send it back in April.  Oh well, I need to put this up so I can make a point with my next post. :-)

**********************

It's been interesting watching the news on the new Facebook Iowa Data Center.  It was big news for many, but I knew about the Facebook Midwest Data Center back in January.

It's been so long ago, I can't even remember how I found out about the new Facebook data center.  I didn't share my discovery beyond a close group of data center friends where we discussed why would Facebook build a huge data center in the midwest given their large presence in Prineville and Forest City?  My friends couldn't come up with a good answer.  

The one idea I threw out is Facebook is going for an Active-Active strategy.  Prineville to Iowa.   Forest City to Iowa.  If Prineville goes down, fail over to Iowa.  Same for Forest City.  There is complete data redundancy between the East and West Coast to the midwest data center.

Now that Facebook Iowa Data center is public.

A New Data Center for Iowa

April 22, 2013

By Jay Parikh

Today we’re thrilled to announce that Altoona, Iowa, will be the home for Facebook’s newest data center.

There are more of us who can chat about what Facebook will put in its midwest data center and how it will interact with its other sites.

Ouch, Cloud may have holes in it, Rackspace stock drops, is reality catching up to hype?

Cloud, Cloud, Cloud.  It is the way of the future.   Yeh. Yeh, Yeh.

It looks the Cloud may be a bit over hyped.

GigaOm's Barb Darrow covered the Rackspace news.

Laggard Rackspace growth sparks concern: is there enough cloud biz to go around?

by Barb Darrow

1 HOUR AGO

No Comments

• A
• A

photo: Flickr / mnsc

SUMMARY:

There are lots of potential cloud workloads out there but there are also about a zillion clouds. Is there really enough paid work to support them all?

tweet this

Here’s the narrative that cloud vendors would like us to believe: there are infinite workloads flowing to clouds of infinite capacity. There’s enough business for all, keep moving.

Rackspace did add 4% more servers quarter to quarter.  So even though there may be some pricing issues, bottom line the server growth was only 4%.

 Total server count increased to 94,122, up from 90,524 servers at the end of the previous quarter.

Data Center Marketing Hype vs. Social Networks

One person recently asked me what is going on in the data center industry.  I said the big are getting bigger, the small are disappearing and the middle is trying to make it seem like things are great.  Those who make the most amount of noise aren't necessarily doing the best.  When a company is quiet with few press releases and presence at trade shows doesn't mean they are not growing. I found it interesting when you would give a small presentation at a conference like Uptime Symposium and half the room is full of your competitors. If you have something really good that sells, you don't want to tell your competitors how you market your product and how it addresses customer needs.

What got me to write this post is reading Chris Crosby's post on Pardon My Hyperbole.

Pardon My Hyperbole

 

Share on facebookShare on twitterShare on emailShare on pinterest_shareMore Sharing Services2

 

Everyone exaggerates. The big one that got away gets bigger with every telling. That one yard plunge you made for a touchdown in high school now stands at 50 yards, and getting longer, and of course you really did use to have a 30 inch waist. Hyperbole is not a bad thing. The advertising business is built upon it and so are most of our egos. Al Gore has made a career of it ($200 million at last count). In our everyday lives, we accept a certain amount of puffery surrounding most any assertion that we hear—let’s call it our personal plus or minus 10%—but sometimes we just have to jump in tell someone that “it’s time to pull in the reigns there cowboy”. I ran across just such a case the other day when I read someone describe data centers as “today’s steel mills”. While I agree that everyone has a right to use hyperbole to make a point, I think this guy’s abusing the privilege.

When I first read this I thought of the marketing over promises of what a product or service will deliver.  Most vendors know their customers will under utilize the product/service so its performance will be fine.  But the bigger players are working at a scale that challenges the limits of products and services.  What do you mean the product has issues at 80% load.  Well no one runs the product at 80% load most are aren't even at 50% load.  You mean your specifications aren't accurate.  Well no one has actually run our product in production at 80% load.  One of my friends was a the nightmare tenant in his colocation facility.  He would consistently push circuits to their 80% of rated capacity.  His landlord would constantly talk to him about the dangers of running the facility at 80%.  My friend knew he was paying for the capability so why not use it.

Here is a crazy idea for the vendors spend some of that marketing budget on listening to the experienced influentials.  Learn what issues they have with existing products/services and what they need in future products/services.  Oops, just shared what some of the smart people have figured out.  Hanging with the influentials to listen is worth more than trying to sell them something.  

One of the funniest stories is when a salesman cornered a data center executive and lectured him how he is making big mistakes not working with his company, one of the top companies in the industries.  You aren't buy into my over promise and under deliver market dominance strategy. :-)

Life changes when you hang around the really smart people to listen, learn, and socialize.  You start to see the reality of what works and what doesn't.

The big are getting smarter as well as bigger.  The small have no idea what is going on.  The middle is well, stuck in the middle.