Industry


Ads by TechWords

See your link here


Jeff Boles's picture
Jeff Boles

Virtual Frontiers

Security and the cloudy cloud: A revolution for the infrastructure?

A reader recently dropped a comment on my other post about security and control issues in the cloud. For the record, that other post was about defining the cloud (I define cloud stuff as being about the infrastructure and services underneath any hosted environment - not the hosted services themselves).

I agree with his concerns, but from my vantage point, the industry is on it, and we're going to see some seriously revolutionary capabilities pop up for the cloud that address security and much more. For his part, the reader referenced Mumbai (implying the cloud is about offshore), but the cloud isn't necessarily about offshore, and in fact I think cloud models are hindered by offshore. But successfully engaging the cloud is about "what else does it need, and what do we get that justifies our participation in it."

Addressing security issues is absolutely a critical component, and is one of these areas in which we should be asking ourselves the question, "what else do we get?" In the hosted virtual infrastructure as one aspect of the cloud (what I call virtual private data centers, involving the full range of infrastructure vs just compute) the possibilities are pretty interesting.

You see some folks like Reflex Security coming up with solutions that could have far reaching capabilities in a virtual infrastructure or cloud context, and potentially even provide better security and isolation capabilities there than you can achieve in your own infrastructure. On other fronts, you see some really far reaching capabilities for deep information classification that really creates a new generation of governance, risk, and compliance management - kind of a combination of data loss prevention, and deep awareness of data content that can allow you to do some proactive control not just at the packet level, but around information, who's using it, etc. Nobody is quite to the ideal yet, but I think of folks like Varonis, Abrevity, Njini, SmApper, etc. You shouldn't miss the fact that these very aspects have a potential litigation and regulation angle as well.

The thing about the cloud is that all of these technologies could be deployed as software services within a virtual context, and encompass everything going on in your cloud environment. Now you have an environment that you know has no physical stragglers or departmental servers or unknown physical attack vectors (assuming you've done your due diligence at the physical layer).

Moreover, with a defense-in-depth type approach in the cloud, you may get real benefit from economies of scale, and be able to harness many more layers of security than you ever could on your own. You might get real sophisticated stuff by the service provider's infrastructure to start with, and then start layering in your own stuff in a virtual environment.

There's a whole slew of vendors that can enable different capabilities at the software layer in the cloud - maybe Akorri enables better SLA management, maybe Surgient or someone enables better automated infrastructure adaptation and resource utilization, etc. There's really no end to the possibilities here - look at the comment Carter George from Ocarina left on an earlier de-dupe post. Sure, there's interesting opportunity for compression in Polyserve, what we call primary capacity optimization, but think about the opportunities around deploying that same technology or de-dupe throughout the virtual infrastructure and then granularly managing its application.

Now look at some recent technology that IBRIX is rolling called Cirrus, which is a total example of what I've defined as cloud-based Storage, file storage, in the InfoStor piece. Even without a virtual infrastructure, they have a backend architecture where you could plug-in classification services, policy engines, etc. that act upon stored files, and might be triggered by various API events. Pretty killer stuff, even just for collaborative file storage.

The thing is, while these services enable some killer service provider possibilities, they are revolutionary at the enterprise level. Any large enterprise should be turning some of their gears around whether this has application in their environment. Can you get more capabilities in an internal service provider model that delivers either file storage or entire virtual private data center hosting (security, scalability, automation, better service alignment, clear SLAs, etc.)?

Interesting implications, but still, today, who's harnessing it? At the strict service provider cloud level, there's lots of room for innovation and leadership, and few solutions today. Mine is a long term view, and I'm waiting for the players to step up and enable some sophisticated capabilities in this idealized version of the location abstracted, remotely hosted cloud. You see it starting to happen. IBRIX is maybe the first throw down around truly cloud-based technology from an out of the box solution, but trust me, there's a lot more to come.

Beth Pariseau wondered will the recession drive users to the cloud? It won't be just the recession, but it may be a matter of the recession intersecting with the economies to be reaped from the cloud combined with new capabilities.

What People Are Saying

http://www.airnk.com

If there is something I want to get, I think tiffany is the first, and then, maybe the louis vuitton handbags. If those are my luxury, I want to say some sneaks are not so luxurious. Like jordan shoes, it’s really not so freakish for me. And you know, I like playing basketball, and I need a pair of good quality basketball shoes, it must be a good brand. Maybe the nike products is my favorite, so a pair of nike shoes is necessarily.

Distributed Trust Management is perfect for the cloud

For seemingly new problems, there are known solutions which can be applied. For security in the cloud a refreshing approach would be a Distributed Trust Management model. This is something like Keynote (Matt Blaze, etal., AT&T research) described nine years ago. Decentralized trust management carrying just enough contextual policy would be better than the status quo of today, i.e. authenticating against directories like Active Directory, OpenID, Windows Live and mapping userids to static ACLs. Now is the time to think differently about cloud security instead of adding security as an afterthought when it's all built and too late to change. Windows Azure to date has failed to innovate in cloud security.

Great point - How about trust as a layer of abstraction

Hi Richard,

Thanks for the great comments. I totally agree, you have to have some trust and security model that crosses boundaries regardless of whether those boundaries are file systems, operating systems, an individual, an organization, a service provider, or something else. Moreover, a new approach needs to make access semantics lightweight, more atomic and associated with meaningful levels of content such as file objects. You see some of this emerging in IBRIX's Cirrus, about to be announced, that creates a single namespace layer of abstraction on top of file systems, be they one or many, and it really could work regardless of the file systems, semantics, ACLs, etc. underneath as long as there is a good mapping in place. Then Cirrus essentially turns to metadata to impose isolation, structure, and security against file objects stored in that namespace. Now getting the industry to coalesce around a broader vision that takes IBRIX's approach to the next level, across the cloud, is the next step.

Jeff Boles
Sr. Analyst
Taneja Group
www.tanejagroup.com

Great post

The challenge that many companies will have in trying to enforce existing policy on their new cloud storage deployments is that much of their policy today is based around limitations such as which data sits on which servers, etc. Take the physical limitations away and things like access control are going to have to be figured out all over again from top to bottom. Whereas most access control is based on servers or directories for example - it should be based on each Electronically Stored Information object individually and will be if new tools are used.

This is going to be no less significant for existing companies than their transition to network computing 10-15 years ago.

Exactly! The Cloud can be all about data control.

Jonathan,

Thanks for the comments - and that's absolutely part of my point. While the cloud is going to be a significant, if not revolutionary transition in enterprise computing, I think it will actually help us tackle some of our issues around security and control today. Since in the cloud model you are now working with whole environments (a virtual private data center) that are better connected than today's silos, and more granular services (whether they be in the form of virtual appliances or just lighterweight applications) you may now have the ability to pop up governance, risk, and compliance type solutions as needed, on-demand, and deliver better data management and control.

What's more, is that if you're an enterprise today with anybody doing development on an EC2/S3 cloud or similar (and you may have this going on and not even know it), then you better be thinking along these lines already. You may have close to production data in that cloud, in the form of structured or unstructured data (and maybe even creating more) and you need an ability to see data across all those systems and control it. It might be fine there for an interim, but you better be sure it is being used in the right way, is being controlled against the risk of compromise and/or loss, and is end of lifed when it needs to be.

Security guys, time to start running a little faster to keep those couple of steps ahead.

Jeff Boles
Sr. Analyst
Taneja Group
www.tanejagroup.com