Sears is Turning Shuttered Stores into Data Centers | Hacker News Reader

Sears is Turning Shuttered Stores into Data Centers

datacenterknowledge.com

135 points1SockChuck87 comments

This makes sense. Most of the spaces Sears is closing are obsolescent for retail uses for several reasons. There is a vast over-supply of retail space in general, partially because the boom generated construction for the sake of construction, partially because the bust reduced consumer spending and partially because of the internet.

Retail shopping has also changed. Malls are being replaced by power centers and urban core retail is being revitalized. And there's Walmart for household items and Home Depot for white goods and Dicks for athletic equimpment etc.

One of the biggest changes is the way in which stores are placed relative to catchment. Driving patterns have changed and interstate access is more important than a high traffic corner for big box retail. Sears locations were based on the old model.

That said, their locations on older arterials tends to correlate with high levels of utility infrastructure. A collapsing retail micro-environment may mean surplus utility capacity. Which reminds me that another trend making Sears commercial buildings obsolescent is the radically improved energy efficiency of modern retail design - big boxes have skylights, etc.

[edit] A bit more about power supply. Local power companies sell power. That's how they make money. Like the Sears stores are surplus retail space, they often have capacity in the wrong places. They will cut deals in exchange for a long term return - it's why they run lines to a site in the first place.

Lots of good points. K-Mart is especially ripe for conversion since, in addition to your points, their stores are two large for most retailers to move into and too small for a Wal-Mart or Home Depot.

National Big Box retailers rarely move into existing buildings unless the real-estate market is tight and entitling new development or re-development exceptionally difficult. Otherwise it is more cost effective to demolish and build to suit. They have corporate planners, architects, and lawyers who do this sort of thing day in/out.

In another life, I worked for an architect who did nothing but Walgreens stores. Once their analysis showed that a corner met their traffic count, it didn't matter how much the land cost. The pharmacy would pay for it in 30-90 days. That's why there will be a CVS across the street.

Id love to see / figure out their numbers.

I forget what it's from, but your last sentence reminded me of this (paraphrased).

> McDonalds and Burger King. McDonalds has it hard - they have to do market research, explore new areas, test whether they work. Burger King has it a little easier - they just follow McDonald's around.

In the mid 1980's, I heard that story about Subway. Which as a business strategy makes even more sense - offering a clear alternative to a market with a high demand.

Reminds me of the story of the master sailor hired to help navigate a sailboat across the Pacific to Tokyo. He satisfied the boat's owner of his qualifications, and joined the crew.

During the entire crossing, he drank heavily, slept until noon, crawled above decks, stared at the sky, and told the helmsman "point 10 degrees north". Or some days, "point 10 degrees south".

The owner was a bit concerned, but things seemed to be going well. And the ship arrived spot on to Tokyo harbor.

On landing, he asked the sailor, "how did you do it? The whole trip, you never touched a sextant, looked at the compass, or checked the satellite navigation?" "Oh", said the sailor, "when I came up on deck, I looked at the airliner contrails, If the were to the north, I headed us south. If they were to the north, I headed us south."

That's a great story!

But wait... He steered them away from the contrails? Not toward them? Please help me understand.

Longer trip. More billable hours. Made up at the end.

(Wups! )

It's from the book Fast Food Nation. McDonalds was said to spend millions on satelite imagery and computer time to figure out the best places to put franchises. Burger King and subway just followed them around.

Those analyses are interesting when the proposals project a 10-20 year use of the structure, and include the demolition and disposal costs of the building itself.

Still, it's ridiculous that Sears fumbled so thoroughly information age. Their background as a catalogue store should've positioned them perfectly for online shopping - effectively just an expansion of the catalogue concept. Having a physical storefront in tandem with a catalogue is fantastic for returns and other problems, and their full appliance service model is also still an impressive product even today.

Sears was positioned to do far better than it did, even with its anachronistic department stores. The leadership just completely fumbled the information age.

They had a head start, too: https://en.wikipedia.org/wiki/Prodigy_(online_service)

Part of the problem is that most of the things Sears sells are things people are used to getting their hands on first. When Sears was mostly a catalog company people simply didn't have the option -- order the dress and get the sewing kit out if it didn't fit. The long gap between pure catalog Sears and online sales was one that trained shoppers to be just different enough that it no longer worked for them.

Rackspace's corp HQ is a converted old mall (the entire mall) in San Antonio [1].

[1] http://www.nytimes.com/2012/10/31/realestate/commercial/rack...

Wild. I used to hang out at Windsor Park in the 80s. I still have a Rubiks Cube I bought there.

Katrina refugees put the last nail in the coffin at Windsor Park

“Sears and Kmart never deployed wireless on the rooftops, so there’s no rooftop usage at present,” he added. “There’s tons of interest. I will put as many of the rooftops in play as I can.”

I wonder if they will consider covering excess space with solar PV panels? They could become a massive distributed power source, which would contribute somewhat towards the power requirements of the data centre below.

It's amazing to think of how the internet is changing cities and buildings. Reminds me of phone boxes that are being converted into other uses (such as electric vehicle recharging stations, mini-libraries and Wi-Fi hotspots) since they've been made obsolete by mobile phones.

http://www.engadget.com/2012/05/10/spectrum-interactive-wifi...

http://phys.org/news192197933.html

http://www.gadling.com/2012/08/19/british-phoneboxes-convert...

Somewhat tangential, but still part of the core argument:

Most of the issue with power is moving it. Generating power is not trivial, but it's easier than moving power, which is still very inefficient. Distributed solar panels on the roofs of Sears facilities might offset their own carbon footprint, but it's unlikely that the power they generate would have a significant impact on the surrounding area.

This is the reason we don't have giant floating solar panels in the ocean, for example.

Is there any reason why they couldn't do some sort of deal with the likes of Solar City? Sears has rooftop space, Solar City has the guys to install panels. They've done deals with other big companies in the past to cover their rooftops with panels:

http://en.wikipedia.org/wiki/SolarCity#Commercial_solar

I don't know what SolarCity would do other than install the panels.

Does SolarCity participate in the resale of power to the grid? I wasn't aware that they were an energy broker.

There's really no where to move the power, in this case.

The building itself could use it, and if there was excess the neighboring buildings could use it.

It's unlikely that solar panels on the roof with power the whole DC so I concede the point. Moving the power is probably a moot point.

The bigger issue for renewables than transmission (we already transmit hydro and some nuclear power long distances) is storage and damand-matching. Conventional and nuclear plants (as well as hydro and geothermal) are dispatchable power. The output can be dialed up or down to meet demand (faster for some means than others, hydro can spin on a dime, nukes and coal plants not so much).

For solar, wind, and tidal power, you get what you're given, and deal with additional demand by building overcapacity, through dispatchable power (conventional plants or biofuels in addition to hydro and geo), or by building storage.

Building a nation-sized battery is a pretty heady task. There simply aren't sufficient mineral resources for many electrolytes (including lead) to build sufficient capacity, though some abundant salts and metals may be suitable. Efficiency matters far less than scale and cost.

Floating solar hasn't been built because solar panels have until recently been more expensive than conventional sources, and more significantly, because marine environments are unbelievably harsh and electrical equipment will degrade quickly. It's far more cost-effective to simply build them on land. There's plenty of rooftops and unused acreage.

On the contrary, coal and nuclear thermal power plants take a long time to throttle up and down.

To increase the output of a thermal plant you need to increase the heat production, wait for the steam pressure to increase, then open the steam tap to increase the turbine power.

With the exception of New Zealand, which is ~80% hydro, most countries use coal for base-load, and hydro for peak generation.

Other peaking options are natural gas turbines and diesel generators, both of which are quick to start, and quick to throttle.

Additionally, Hydro power is easy to store - just leave the water behind the dam. Hydro will probably provide the national grid sized battery we will need to balance out transient generation like solar and wind.

On the contrary, coal and nuclear thermal power plants take a long time to throttle up and down.

It's the fact that they can be throttled. You cannot control the throttle on wind and solar. You can control the brake (how much of the generated electricity is released onto the grid). But if the motor ain't running you can't make it go. And if much of your power is dynamic and non-dispatchable, you've now got variance on both the demand and supply sides of your grid.

Hence: the need for storage. If you can take the surplus available at times when you don't need it and store it somehow, then you can match grid demand.

The problem with hydro is that there's simply not enough capacity. A small number of nations (New Zealand, Switzerland, Sweden, Brazil) have large hydro capacities (and relatively small populations for the most part). For the US, and much of the rest of the world, most available hydro sites have been built out.

Other options include geothermal. There's a vast potential under Yellowstone National Park, though it's politically untenable to even contemplate tapping it (a USGS survey of US geothermal resources didn't even mention it): http://pubs.usgs.gov/fs/2008/3082/

Iceland, Hawaii, Japan, and Kenya also have very significant geothermal potential.

I'm a little surprised by this. The article implies that shuttered Sears stores already have the power and data capacity to function as a data center, but I can't imagine why. Compared to similar sized buildings (office buildings or factories, mostly), Sears retail locations don't seem to have a lot of power- or data-intensive needs.

Sears does have some amazing infrastructure (lots of square footage, locations at key transportation points in densely populated areas throughout the country, huge parking lots) but those aren't really the kind of features that data centers are usually optimizing for. I'm sure they're finding creative uses for this infrastructure, but I doubt it's anything like a data center as we think of one.

It's not whether it's already laid, but how easily it could be added.

The article mentions that the target locations are standalone buildings with good fiber and power already connected. They mention 5 megawatts for one location.

I don't know why a Sears location would need lots of power (beyond tons of lights and powerful HVAC units), but I can see the need for quality and low-latency data: retail POS generates a large amount of transactional data, and you need low latency for database syncing and credit card/loyalty card transactions.

> retail POS generates a large amount of transactional data, and you need low latency for database syncing and credit card/loyalty card transactions.

I've always been curious about what sort of infrastructure the average big-box retail store actually needs. My local Walmart can't have anything but a wireless uplink, unless they actually installed 5+ miles of fiber or something when they built the store.

This Quora answer gives a little insight in to the needs of a modern store: http://www.quora.com/What-is-the-new-instantaneous-payment-s...

Also these stores basically couldn't exist without robust inventory management systems. There is probably more technology in the average big box than we realize.

I was told that every time an item is scanned at the register at any Walmart store, it updates an inventory database in near real-time (I assume there is some sort of message queue with that kind of volume and distribution). Then their system automatically places replenishment orders with their vendors when the inventory of that item goes below a certain level. They are a well-oiled machine, focused on making sure that they almost never lose any sales due to being out-of-stock, nor waste money on carrying excess inventory in their warehouses. Mismatching inventory to sales will drive a retailer out of business very quickly.

This is also part of the reason why they are able to push "everyday low prices" instead of discounts. Discounting is what you do with the excess inventory when you messed up and ordered too much of something.

The system is Retail Link and there's quite an ecosystem around it.

Working for a regional grocery retailer from the mid-90's to the mid-2000's, I can assure you that not nearly enough consideration was given to bandwidth needs when designing and building new stores. I remember, as late as 2002, having an entire store run off of something like a 512kbps circuit.

We literally shipped register (paper) journal tapes back to the warehouse via truck mail.

Time/date/UPC. That should fit over a 9600 modem line, and that's before you add compression.

More like time/date, UPC, quantity, item description, price, coupons, sale discounts, sales tax, rebate info, clerk info, customer info, payment info, and more. Times dozens of items per receipt, times dozens of registers. All continually operating 12+ hours a day. All needing low-latency access to master databases, credit card servers, and banks.

Trust me, retail is big. Like "Big Data" big. There is more data there than you think. ;)

Sears also runs (and used to run many more that have since shut down) an auto repair shop -- Sears Auto Center. Auto shops will have some pretty hefty power requirements.

Well, my guess would be that they might have similar requirement power-wise. Keep in mind that those store need power for fixtures, refrigeration / freezing ... Someone knowledgeable around to confirm or refute ?

It probably would need some development but to me it's not as senseless as it seems.

The closed stores tend to be old, which means their original power requirements were based on what today would be grossly inefficient, and they are in locations with high levels of utility infrastructure and declining demand due to both the closure of other buildings and the increasing efficiency of those that remain.

Retail is actually fairly intensive from a power perspective, lighting levels are high to show off goods. The capacity of cooling systems is also high compared to office uses because the design case is a mob of shoppers [edit: think August afternoon during the tax free back to school weekend]. Unlike lighting, people generate a lot of latent heat, and unlike office workers they are moving. Also unlike office workers, feeling a little hot means they will take their business elsewhere.

Now, that's not to say that the infrastructure of a retail space is going to compare to a heavy industrial use, such as a steel mill or juice plant. But those aren't what Sears owns. And in any event, industrial sites don't offer the opportunity for redevelopment of the parking lot with smaller boxes as a means of diversifying the underlying asset. The sites aren't bad, the way in which they are developed is.

"Unlike lighting, people generate a lot of latent heat, and unlike office workers they are moving. Also unlike office workers, feeling a little hot means they will take their business elsewhere."

Excellent point. I remember when I put an HVAC system in many years ago (in a plant) the calculation was something like 2000btu per person in heat generated in terms of sizing.

I was curious so I just checked (couldn't back then) to see if that figure is anywhere near correct but can't seem to find the answer.

This seems to indicate that 600 watts is about 2k btu per hour it wouldn't surprise me if a human gave off 600 watts of heat:

http://www.rapidtables.com/convert/power/Watt_to_BTU.htm

Actually this more or less shows it is ballpark:

http://answers.yahoo.com/question/index?qid=20081206141450AA...

it wouldn't surprise me if a human gave off 600 watts of heat

Much closer to 100W sustained. I can hit 600-700W, very briefly, on a rowing machine. Wouldn't care to keep that up for more than a minute or two.

I second the "100W sustained" number. I did the math once with the assumption: 2000 kilocalories/day diet, 4.1852J/calorie, 86400 seconds/day, all input calories are eventually given off as heat.

2000 * 1000 * 4.1852 / 86400 = 97 watts

That is the problem, there is NOT enough power on site to run a datacenter. The article quoted one location with 127ksqft and 5MW. 5MW is tiny for that footprint. To give you an idea, you need to roll at least 300-400w/sqft of power/cooling to be viable, calculate in UPS and other losses, and 5MW powers around 5000sqft of datacenter space which is only a couple hundred racks of servers and storage. They're probably going to need 40-100M to retrofit that space, so I'd say the math is not on their side...

This reads like a realestate person proposing this, not a datacenter person...

Rooftoop solar is good for a mere 8 Watts / sqft [1] [2] when the sun's shining, and it's not a constant 8 Watts output over the day even if the weather is clear all day.

So repurposing a Sears building's rooftop area for solar power would not cover even the minimum requirements you cite (300-400 W/sqft -- x 24hrs/day) as necessary for a datacenter re-purpose.

But then there is that big, unused parking lot. If it's got, say, 30x the rooftop's area and can be fenced in, maybe then a ground-mounted solar system might supply some significant fraction of daytime energy.

------

[1] http://www.ehvacdesign.com/solar_system_design.htm

[2] http://www.solar-estimate.org/?page=solar-calculations

Don't forget Rackspace turned an abandoned mall into their HQ

http://www.bizjournals.com/sanantonio/news/2013/03/21/stormi...

And if anybody is looking for more photos, http://www.rackspace.com/blog/get-an-inside-look-at-the-rack...

Interesting move -- feels like this is a natural response to the not-so-sudden realization by "normal" folks that you can get a BIG multiple on $/sqft as a data center over a more traditional tenant.

A parking lot is better than an empty lot. A warehouse is better than a parking lot. And a data center is better than a warehouse :)

The good news for us consumers is that as the supply steadily ticks up, pricing will come down (at least for space - who's to say where bandwidth costs go).

>A parking lot is better than an empty lot.

Paving paradise, &c ;)

It bothers me that the biggest issue I have with this article is that "blue light specials" were K-Mart, not Sears.

I guess the author thought he could get away with it since Sears and K-Mart are the same company now.

Same company now.

Not quite retail, but Amazon's main Data Centre in Dublin is the former main warehouse used by one of Ireland's largest retailers. It's gone from being full of beans and nappies to servers full of kitten gifs.

http://www.datacenterknowledge.com/archives/2011/02/09/amazo...

Cooling costs ! Buildings whose whole structure doesn't take data center cooling requirements into account are at a significant cost disadvantage over modern ones built specifically with cooling optimization in mind.

But you have to include the cost of building the entire building and real estate. That's a sunk cost for Sears, so they don't have to worry about that.

The buildings do not have to be full to bring in more revenue than when sitting empty. Once they pay for operating expenses, everything else is profit. In many cases, I suspect the buildings are paid for, so the value of the real-estate is a sunk cost.

These aren't going to be on Google's scale. This is Ace Hardware not Home Depot. Optimization scales accordingly.

Now, all they need to do is to turn all the ghostmalls into hackerspaces and we are on to something.

I think that's Makers: http://craphound.com/makers/

One of the few fiction books I've read in recent years that's bent my mind.

Where do you think I stole the idea from? Great book.

It's a good enough book to make the association concrete.

Agreed.

Anyone else reading this comment: http://craphound.com/makers/donate/ (download the book for free, donate a copy to a school)

First thought that came to mind when I saw the OP. Wonder what the ROI is on using Sears retail as a data center vs a co-working/incubator space.

Dealing with the politics, aspy personalities, and payment issues with a small space is hard enough... I can't imagine it magnified to shopping mall proportions!

I'm not sure I understand this. Retail real estate is all about increasing footfall, not just the size of it. Hence why you see the airport bombarded with shopping nowadays. Unless there are particular Sears locations that don't see benefit from adjacent footfalling (just made that word up) retail outlets, then it makes sense. But then it wouldn't make sense for Sears to be there in the first place if the footfall opportunity wasn't there or couldn't be created there.

Times change, I've seen this locally. In the US Midwest, various metro areas have lost population over the last couple decades. As a result, they don't need as much retail space to support the current population. Something has to give. The Sears closest to me is attached to a condemned mall. Sears stays open because they own the building though. The reason they're there is because originally, the mall was the place to be for retail.

I'm not surprised at all, given the Sears real estate portfolio. In Indianapolis, a company converted an old mall into a data center, and it seems to be working pretty well [1]. Retail buildings probably have a decent connection to the power grid, so it would really make sense to make use of the open spaces.

[1] http://www.datacenterknowledge.com/archives/2012/10/30/lifel...

the erosion of shopping malls continues.

warrants noting that malls require anchor tenants to work. It's the convenience of to being able to go to one location and get clothes, tools, make-up, music, videos, books, etc, that makes malls work. I know people have been predicting the demise of the shopping mall since their introduction but this will be interesting to watch.

I realize they have a building with lots of empty square footage, but how do they get enough power, bandwidth, cooling, etc?

Data centers built by Google, Facebook, etc. seem to all be built from the ground up, and close to things like power plants.

Also, wouldn't the premium of leasing or owning "retail" style property make the cost of maintaining the building too costly? I don't think Sears needs a datacenter on main street, or within a strip-mall.

After RTFA I pulled this:

"Farney acknowledges that many of Sears’ mall-based retail locations aren’t viable for data center usage. “I don’t think the industry is yet ready for a mall-based data center,” he said. “That may take some time. The stand-alone location is optimal.”"

It seems they are considering it more for certain stand-alone locations.

The real-estate deed for any mall location almost certainly precludes such uses by covenant. If they didn't, Sears would never have bought the property in the first place.

And it woudl be harder to secure you would find it hard to add a moat even adding a proper double fence might be hard in most of these store locations.

They're not all built from the ground up: http://www.google.com/about/datacenters/inside/locations/ham...

The challenge I see here is power provisioning. With the space that a Sears store offers, you'd be hard pressed to put 10 to 50 MW of power into it. But perhaps if you already have the land covered smaller things are more desirable.

But will they be using all those CompuServe disks they never sold as hipster coasters? That's what I want to know.

Tomorrow:

Google buys SEARS, register as a new G4 provider.

Good use of the sites.

This is an opportunity where they could usher in a new glorious era of personal cloud computing that solves the problem of the government basically owning your hardware, software, and the usage of your software if the government catches you (or the software) so much as jaywalk.

Instead of selling floor space, they could subdivide the space like the "storage building" units. The customers could pay for a rack, and they would need a key to access it. Legally the hardware, software and data would be owned by the owner, and not the government. Anyone else would need a warrant to come in and rape your data, software, and hardware just like if the server was next to my bed.

Part of the reason why I have all my servers in places I own is because of this problem. This is a brief opportunity to explore cloud computing and give the communists and socialists who want everyone to share a common purse on the computing front a black eye.

My pain point is "I want boat loads of CPU cycles and Internet connection speeds, and I want physical legal ownership of the rack space, hardware, software, AND data, and I don't want to have to buy a building that I have to air condition, power, internet connect, secure, and pay taxes on.

There is a lot of odd precedence that makes renting small storage units not the same thing as larger units. Standalone Property > Multi Unit Property > Storage Space > Rented computers.

This is a brief opportunity to explore cloud computing and give the communists and socialists who want everyone to share a common purse on the computing front a black eye.

Wow, where did that come from?

Cloud computing is communist now? I could have sworn otherwise.

https://www.eff.org/deeplinks/2013/05/caleatwo

The government is going to get its backdoor into cloud computing platforms one way or the other, because of child porn, the terrorists, and corruption. The segmenting idea is a dumb one because we don't physically own the space, it isn't ours. People with wearable supercomputers on programmers who are able to understand the metal on up through the software and data will be only truly free citizens (free from government fiddling with and spying on everything they do).

There will be precious few free independent citizens in the coming centuries. Most humans will become as one, where one human's consciousness will blur and blend into everyone else's. I will be one of the islands, choosing: "thanks, but I prefer you not have data injection, and fiddling rights with my own mind".

The beginning of this blended and blurred consciousness and hive minds begins here, today, with cloud computing becoming owned by the government, and having your desktops, laptops, smartphones and interfaces with your own mind as thin-clients. Where you begin to think about a crime, and the others will say: "That is a bad idea" and stop you as you are doing the crime.

> Wearable supercomputers on programmers who are able to understand the metal on up through the software and data will be only truly free citizens (free from government fiddling with and spying on everything they do).

I like the assertion that the wearable supercomputers will be the only truly free citizens. It's probably not exactly what you meant... :)

I'm pretty sure that it's more of a window into the poster's psyche rather than an observation of fact (esp. based on the next post.) I guess right-wing totalitarian dictatorships don't want to read your mail?

Your comment is a bit of a window into yours - disparaging remarks made against the left wing don't imply support of the right wing.

Read their second post - that person is either very, let's say, eccentric, or under the influence of a psychedelic.

Yeah, I made that reply before I saw that second post. It doesn't really change much though - something tells me he's not a supporter of totalitarian right-wing rule either.

I'm just saying that there's not any more of a connection between communism and a surveillance state than, for example, Turkmenistan. You're reading more into it.

Hesitant to comment on this post, but:

"Instead of selling floor space, they could subdivide the space like the "storage building" units. The customers could pay for a rack, and they would need a key to access it."

Many DCs offer quarter racks that are under lock and key. You just bring your hardware and get your carrier neutral drop.

> Instead of selling floor space, they could subdivide the space like the "storage building" units. The customers could pay for a rack, and they would need a key to access it.

You know thats how colocation works right now, right?

Reply on news.ycombinator.com