If you need proof that cloud deployment stories can touch off religious disputes, my recent report about @Walmartlabs deploying 100K cores of OpenStack to run the retail giant’s e-commerce operations is Exhibit A.
This is, by any measure, a massive private cloud, and some readers were incredulous that [company]Walmart[/company] would go this route instead of plying public cloud services. It’s the old build versus buy discussion all over again, with many of the participants weighing in on the “buy” side.
One reader, termed this decision “ridiculous,” pointing out that @walmartlabs has hired on 1,000 or so engineers over the past year — although no one said all those people were dedicated to building or maintaining the aforementioned OpenStack private cloud. Still the argument is, if you go with public cloud, you won’t need to bring that much expensive talent in house. Engineering talent is pricey, especially in Silicon Valley. @walmartlabs is headquartered in San Bruno, Calif.
His opinion is that a big retail outfit is far better off using “out of the box” public cloud capabilities for much of its work rather than reinventing the wheel (or building its own cloud.) For this camp, Walmart’s decision to build a customizable and flexible cloud with OpenStack makes no sense.
On the other hand, private cloud (and OpenStack) proponents noted joyously that Walmart’s work proves “private cloud deniers” wrong. (Does anyone else find that phrase disturbing? It brings to mind thought of climate change and holocaust deniers and seems to lack a sense of proportionality but back to the topic.)
Server Density CEO David Mytton, a buy sider, wrote about the Walmart private cloud here. Bottom line, he said Walmart is:
dedicating significant resources to building their own “private cloud” and although it’s true there is no specific vendor lock-in, they are locked into their own development. They’re competing in resources, talent and innovation against the public cloud providers (who have more resources to dedicate to engineering both product features and efficiency at scale).
Anybody but AWS?
Remember, given the competitive retail landscape, Walmart was hardly likely to run Amazon Web Services public cloud seeing as how Amazon.com is seen as Darth Vader by many of the rest of the retail universe. Target used Amazon.com (not AWS) for infrastructure but left the fold in 2011.
AWS would likely point out, if it were prone to comment on such things, that its cloud business is run as a separate entity than [company]Amazon.com[/company] — [company]Netflix[/company] is a huge customer after all and Amazon also runs Amazon instant video. But I’ve talked to other retailers who, off the record, will point to the political incorrectness of turning over key retail functions to Darth, er AWS.
Jeff Aden, co-founder of 2nd Watch, a systems integrator that works with customers to deploy AWS, said his company has several retail customers running on AWS, including Diane Von Furstenberg. Other AWS retail users include Gilt.com and Nordstrom Rack.
Mytton, conceded that AWS might be a tough sell for a big reseller to use, but why not throw in with [company]Google[/company] Cloud Platform or [company]Microsoft[/company] Azure? He points out that Ocado, the big British retailer is a Google cloud customer.
Last week I spoke with Sudhir Hasbe, director of software engineering BI and data services for Zulily, a members-only online fashion retailer that has fully embraced Google cloud services — BigQuery, Google Storage and Google Compute Engine. In this, Zulily is sort of a counter-narrative to the @Walmartlabs story.
Zulilly puts 9,000 new items on its site daily but wants to make sure it displays only the items that are relevant an potentially of interest to a given shopper. If you’re a woman who shops for herself and maybe a 6 year old boy, then she’ll see options for those demographics and not have to wade through the rest. “Search doesn’t work well in retail,” Hisbe said.
“For this we need the full big data platform so we can perform maximum data processing– what preferences do they have, what do they like. It also means when you have that much data, the whole supply chain side needs to consume it to make decisions,” he noted.
What’s nice about deploying Hadoop clusters on GCE, is that once the processing has run, the data is pushed into BigQuery where it’s available to all the business units and analysts, and the bill for Hadoop processing stops. The data is all stored in inexpensive Google Storage.
Anyway, feel free to comment on when and in what circumstances it makes sense to deploy public cloud or BYO private cloud. But please keep it polite.
Agree or not, Mark Cuban’s take on net neutrality is worth a listen
For those who missed, Mark Cuban visited the Structure Show last week to re-iterate/explain his thinking on net neutrality and why he thinks turning over internet governance to the FCC is a big mistake. Check it out below.
[soundcloud url=”https://api.soundcloud.com/tracks/193100656″ params=”color=ff5500&auto_play=false&hide_related=false&show_comments=true&show_user=true&show_reposts=false” width=”100%” height=”166″ iframe=”true” /]