Hey, You, Get Onto My Cloud!
Share this article
ADVERTISEMENT
(comments below)
Comments
Comments
9 responses to “Hey, You, Get Onto My Cloud!”
-
What could possibly go wrong?
-
I’ve spent nearly every waking moment of the last four years designing and building the cloud for one of the big players. Most people haven’t absorbed just how profound an impact cloud will have on IT. In my opinion, Amazon Web Services launched the cloud on March 13, 2006 when it first sold access to its object storage product (Simple Storage Service or “S3”) to a third party. In the twelve years since then AWS and the other major cloud providers have built compute engines, networks, runtimes, databases, application development tools, AI and machine learning platforms in their clouds. What started as a relatively slow speed storage capability 12 years ago has become the entire computing stack today. Almost all of the major innovations in IT are coming from the big cloud providers (AWS – Aurora, Google – Spanner, IBM – Watson). There are many reasons for this but one reason is that a cloud provider can keep some aspects of their infrastructure private. As an example, a cloud provider might be able to achieve a new level of quality of service at the network level inside the data center. However, that cloud provider might only allow its higher order services to access that capability. A database maker without a cloud can run in others’ clouds but can’t access private capabilities. Pretty soon the cloud provider is also the dominant database maker.
In 12 years, Amazon Web services will go from $0 revenue to (my guesstimate) $27B of cloud revenue in 2018.
Given all that money new entrants will emerge, right? Maybe. Building out the dozens of mega-datacenters required to build a global public cloud costs an eye popping, jaw dropping amount of money. As today’s big cloud providers continue to expand the cost of starting now and “catching up” becomes prohibitive for more and more companies – even giant companies.
Getting the second headquarters of one of the big cloud providers in your region might be worth a whole lot of concessions in the mid to long run.
-
How about a simple explanation of what the “cloud” is if not a massive dataserver that sells space to organizations.
Oh… and I have to say that I’m surprised that the Pentagon which
seldom has pure unclassified data separate from other data – can safely decide what to put on non-govt equipment and keep from having hackers get it…Remember all the hoo ha over that server in the basement? these are servers not even owned, operated or maintained by the govt…
by the way .. speaking of Amazon… check out the 5 Billion… yes that is a ” B” that Maryland is offering Amazon as “incentives” Will Virginia match it?
-
Cloud is hard to understand until you’ve been at it awhile. At first it seems like the same old, same old. Then you peel back the onion and see that it’s really different.
AWS bought a chip design company – Anapurna Labs. The send the AWS chip spect to Intel and Intel fabs a special version of the x(86) just for AWS. IBM designs their own chips as does Google. All three plus Facebook make their own servers. Dirt cheap. They fail a lot. It doesn’t matter because the software that monitors, manages and provisions that hardware is designed to fail over to working hardware very fast. Long MTBF has been replaced with short MTBF and very short MTTR. Nothing you can buy buy from a traditional vendor could do any of this. In cloud, it’s make your own everything. In a world with a lot of failures you can’t store state in the application. It has to be handed off to a database service that is also on commodity hardware but it uses a shared nothing, truly replicated architecture. When one node fails the data is also stored (usually) on two other nodes in two different availability zones or multi-zone regions. No monolithic Symmetrix-style monolithic redundancy, no RAID arrays. Dirt cheap JBOD where the software of the distributed database and the network provides resiliency.
All these failures mean that you can’t rely on people to fix things – they break too often. So, the software itself makes the fix. That only works because everything is standardized – open source operating systems (linux), open source virtualization (usually heavily modified Xen). Bye bye Windows, bye bye VMWare. Open source databases, sometimes expressed as proprietary forks of open source projects. Bye bye Oracle, bye bye DB2. Matrix networks running on Intel Red Rock Canyon chipsets embedded in fabric cards in the servers themselves. Bye bye Cisco. Hello 10 microsecond latency from anywhere to anywhere in a 6MW datacenter. Suddenly real time analytics and machine learning work, but only in a cloud provider’s datacenter. Bye bye customer datacenter.
Traditional behind the firewall infrastructure and the cloud are about as similar as Studebakers and self-driving cars. Yeah, they’re both types of automobiles but the similarity ends there.
-
-
“Every cloud provider who wants to compete for that business will have to beef up their presence in the Washington region.”
1. Why? The Cloud is everywhere. Why does the Pentagon care if the servers handling its data and storing its data are in Loudoun VA or Bangor ME or Columbus OH?
2. What is going to stop the Cloud providers from redirecting DOD data anywhere they choose? Even if DOD’s Cloud use begins locally, data centers relocate where the cheap electricity happens to be near the data cables.
3. “Where” is DOD anyway? DOD is military bases and offices and contractors and connections to federal and State agencies and research institutions all over the place. Why is the mid-Atlantic in general, or northern Virginia in particular, advantaged by DOD’s decision to move its work to the Cloud?
-
1. The cloud may be everywhere but the buyers are in Arlington, VA.
2. AWS GovCloud is dedicated to the US government and only the US government. It doesn’t matter where the datacenters are. It matters where the software developers who write the software that runs the cloud datacenters are.
3. They buyers are in The Pentagon, Ft Mead, NRO, CIA, etc. The days of base commanders having their own datacenters and making their own buying decisions are rapidly dying. There are certainly other areas – Air Force Cyber Command, San Antonio, Army Cyber Command, Augusta GA. But the big dog buyers still live in and around DC.
Why do enterprise tech companies pack into Silicon Valley despite the sky high prices, congestion, endless churn of engineers? Because they buy from each other. God only know how much Oracle (Redwood Shores) buys from Intel (Mountain View). That collaboration and those sales are facilitated by physical proximity. Ditto for DC with Bechtel, The Pentagon, Lockheed Grumman, etc.
-
-
This would seem yet another reason why Amazon’s 2nd HQ. will end up in the Washington Metro Area. At this point, it is hard to imagine any other ending to this story for many different reasons. But when you boil down all of his reasons and drivers in life, you are left with one central truth – Jeff Bezos cannot bear not to be in center of the action. And that is where Washington DC and its swamp is, ground 0.
-
In general , military data does not belong on Commercial servers. Certainly no weapon system data.. but even beyond that.. personnel data.. logistics , administration, etc… is all sensitive and can be “mined” to confirm or validate other data… even classified data.
I predict if the military actually starts using “clouds” owned and operated by others -not to govt standards – it will end badly.
Now.. I would not be surprised at all if the Military wants to create it’s own cloud and is willing to pay handsomely those organizations and people who have that skill and expertise… but I don’t see govt military and other sensitive data going onto unclassified commercial servers.
Now I know there are Yahoos in the govt itself who will think this is a wonderful idea but it’s not… it’s a bad idea…
-
Isn’t AWS GovCloud an example of a government-specific cloud?
-
Leave a Reply
You must be logged in to post a comment.