Rational Survivability

Simon Crosby and I have been going 'round a bit lately arguing the premise of where, why, when, how and how much security should be invested by either embedding it in the virtualization platform itself or being addressed by third parties.

Simon's last sentence in his latest riposte titled "Hoff is Still Confused" was interesting:

Well, how the hell am I supposed to argue with that!? ;)  OK, now I am confused! Simon's taken the high road and thus I shall try to do so, too.  I wrote a ton more in response, but I'm not sure anybody cares. 😉

All told, I think we're both aiming at a similar goal in spite of our disparate approaches: achieving a more secure virtualized environment.

But seriously, I don't think that I'm confused about Citrix's position on this matter, I just fundamentally disagree with it.

I feel strongly that Simon and I really are on different sides of a religious issue but without a more reasonable platform for discussion, I'm not sure how we'll intelligently discuss this more coherently without all the back and forth.  Perhaps a cage match in sumo suits!?

I appreciate Simon clarifying his position and reaching out to ensure we are on the same page.  We're not, but the book's not closed yet. 

So we agree to disagree, and I respect Simon for his willingness to debate the issue.

/Hoff

Hello.

I wonder if you might help me.

I operate an e-commerce Internet-based business that processes and stores cardholder data.

I need a QSA to assess my infrastructure and operations for PCI/DSS compliance.

Oh, I forgot to mention.  All my infrastructure is in the cloud.  It's all virtualized.  It runs on Amazon's EC2.  All my data is hosted outside of my direct stewardship.  I don't own anything.

Since the cloud hides all the infrastructure and moving parts from me, I don't know if I meet any of the following PCI requirements:


I don't know if there are firewalls. I don't know about the cloud-vendor's passwords, AV, access control/monitoring, vulnerability management or security processes.

A friend told me about section 12.8, but it doesn't really apply because the "service" provider just provides me cycles and storage, I run the apps I build but I don't see any of the underlying infrastructure.

Also, I have no portability for BCP/DR because my AMI only runs on the Amazon cloud, nowhere else.  I don't know who/how backups are done outside of my manifest.

I'm sure we could ask though, right?

Seems logical…

However, this is the quandary we're facing with virtualization and cloud computing.  In terms of the companies that hire these PCI compliance experts, the assessment methodology/requirements are predicated upon a "standard" that continues to be out of touch with the economic and technological world around it.  That's not the experts' fault, they're scoring you against a set of requirements that are black and white. 

As companies try and leverage technology to be more secure, to transfer risk, to focus on the things that matter most and reduce costs — if you believe the marketing — It's really a no-win situation.

The PCI Security Standards Council doesn't even have a SIG for virtualization and yet we see the crushing onslaught of virtualization with no guidance and this tidal wave has been rushing at us for at least 3-5 years.   If you believe the uptake of cloud computing, we're blindly hurdling over the challenges that virtualized internally-owned infrastructure brings and careening headlong down a path to cloud computing that leaves us in non-compliance.

The definition of what a "service provider" means and how they interact with the cardholder data companies are supposed to protect needs to be redefined.

It's time the PCI Council steps up and gets in front of the ball and not crushed by it such that the companies that would do the right thing — if they knew what that meant — aren't punished by an out-of-touch set of standards.

Hat-tip to David Marshall for the pointer.

In what can only be described as the natural evolution of Xen's security architecture, news comes of a Xen community project to integrate a VM Introspection API and accompanying security functionality into Xen.  Information is quite sparse, but I hope to get more information from the project leader, Stephen Spector, shortly. (*Update: Comments from Stephen below)

This draws naturally obvious parallels to VMware's VMsafe/vNetwork API's which will yield significant differentiation and ease in integrating security capabilities with VMware infrastructure when solutions turn up starting in Q1'09.

From the Xen Introspection Project wiki:

The
purpose of the Xen Introspection Project is to design an API for
performing VM introspection and implement the necessary functionality
into Xen. It is anticipated that the project will include the following
activities (in loose order): (1) identification of specific
services/functions that introspection should support, (2) discussion of
how that functionality could be achieved under the Xen architecture,
(3) prioritization of functionality and activities, (4) API definition,
and (5) implementation.

It is important to note that this is not the first VMI project for Xen.  There is also the Georgia Tech XenAccess project lead by Bryan Payne which is a library which allows a privileged domain to gain access to the runtime state of another domain.  XenAccess focuses (initially) on memory introspection but is adaptable to disk I/O also:

I wonder if we'll see XenAccess fold into the VMI Xen project?

Astute readers will also remember my post titled "The Ghost of Future's Past: VirtSec Innovation Circa 2002" in which I reviewed work done by Mendel Rosenblum and Tal Garfinkel (both of VMware fame) on the LiveWire project which outlined VMI for isolation and intrusion detection:

What's old is new again.

Given my position advocating VMI and the need for inclusion of this capacity in all virtualization platforms versus that of Simon Crosby, Citrix's (XenSource) CTO in our debate on the matter, I'll be interested to see how this project develops and if Citrix contributes. 

Microsoft desperately needs a similar capability in Hyper-V if they are to be successful in ensuring security beyond VMM integrity in their platform and if I were a betting man, despite their proclivity for open-closedness, I'd say we'll see something to this effect soon.

I look forward to more information and charting the successful evolution of both the Xen Introspection Project and XenAccess.

/Hoff

Update: I reached out to Stephen Spector and he was kind enough to respond to a couple of points raised in this blog (paraphrased from a larger email):

Bryan Payne from Georgia tech will be participating in the project and there is some other work going on at the University of Alaska at Fairbanks. The leader for the project is Stephen Brueckner from NYC-AT.

As for participation, Citrix has people already committed and I have 14 people who have asked to take part.

Sounds like the project is off to a good start! 

Microsoft’s announcements today at OzzieFest (Microsoft’s PDC) include the unveiling of Windows Azure.

The Azure “services platform” is described as:

Holy buzzword bingo, batman!

Look, when I’m forced into vendor lock-in in order to host my applications and I am confined to one vendor’s datacenters without portability, that’s not ” the cloud” and it’s not an “open architecture,” it’s marketing-speak for “we’re now your ASP/XaaS service provider of choice.”

Azure doesn’t run “in the cloud.” It’s a set of hosted services connected to the Internet. In this case the “cloud” is more like fog which encircles the islands of data inhabited by Dr. Moreau and his ghoulish API-infected creatures. (Ed: In full disclosure a year later this strategy makes a crap-load more sense. I simply didn’t get it at all back when I wrote this post)

Amazon has their hosting infrastructure and API’s/SDK’s, Microsoft has theirs. Google, too.

You might convince me there is such thing as THE cloud if there was ONE standardized API subscribed to by everyone who claims membership in the cloud. But there isn’t. Everyone is announcing their own little island with their own API, own “datacenter operating system,” etc.

I go back to my recent rant titled “Will You All Please Shut-Up About Securing THE Cloud…NO SUCH THING…” wherein I stated:

Again, I’m not suggesting that this model is not reasonable, warranted or worthwhile. I am a big believer in leveraging open architectures for the interoperable exchange of data as well as resiliency, scale and utility computing.

I’m simply suggesting that re-branding the word “Internet” and implementing ROT13 to arrive at “Cloud” is really confusing and intellectually dishonest.

It’s not FUD, it’s FOG.

/Hoff

Just to confuse you, as a lead-in on this topic, please first read my recent rant titled “Will You All Please Shut-Up About Securing THE Cloud…NO SUCH THING…”

Let’s say the grand vision comes to fruition where enterprises begin to outsource to a cloud operator the hosting of previously internal complex mission-critical enterprise applications.

After all, that’s what we’re being told is the Next Big Thing™

In this version of the universe, the enterprise no longer owns the operational elements involved in making the infrastructure tick — the lights blink, packets get delivered, data is served up and it costs less for what is advertised is the same if not better reliability, performance and resilience.

Oh yes, “Security” is magically provided as an integrated functional delivery of service.

Tastes great, less datacenter filling.

So, in a corner case example, what does a boundary condition like the out-of-cycle patch release of MS08-067 mean when your infrastructure and applications are no longer yours to manage and the ownership of the “stack” disintermediates you from being able to control how, when or even if vulnerability remediation anywhere in the stack (from the network on up to the app) is assessed, tested or deployed.

Your application is sitting atop an operating system and underlying infrastructure that is managed by the cloud operator.  This “datacenter OS” may not be virtualized or could actually be sitting atop a hypervisor which is integrated into the operating system (Xen, Hyper-V, KVM) or perhaps reliant upon a third party solution such as VMware.  The notion of cloud implies shared infrastructure and hosting platforms, although it does not imply virtualization.

A patch affecting any one of the infrastructure elements could cause a ripple effect on your hosted applications.  Without understanding the underlying infrastructure dependencies in this model, how does one assess risk and determine what any patch might do up or down the stack?  How does an enterprise that has no insight into the “black box” model of the cloud operator, setup a dev/test/staging environment that acceptably mimics the operating environment?

What happens when the underlying CloudOS gets patched (or needs to be) and blows your applications/VMs sky-high (in the PaaS/IaaS models?)

How does one negotiate the process for determining when and how a patch is deployed?  Where does the cloud operator draw the line?   If the cloud fabric is democratized across constituent enterprise customers, however isolated, how does a cloud provider ensure consistent distributed service?  If an application can be dynamically provisioned anywhere in the fabric, consistency of the platform is critical.

I hate to get all “Star Trek II: The Wrath of Khan” on you, but as Spock said, “The needs of the many outweigh the needs of the few.”  How, when and if a provider might roll a patch has a broad impact across the entire customer base — as it has had in the hosting markets for years — but again the types of applications we are talking about here are far different than what we we’re used to today where the applications and the infrastructure are inextricably joined at the hip.

Hosting/SaaS providers today can scale because of one thing: standardization.  Certainly COTS applications can be easily built on standardized tiered models for compute, storage and networking, but again, we’re being told that enterprises will move all their applications to the cloud, and that includes bespoke creations.

If that’s not the case, and we end up with still having to host some apps internally and some apps in the cloud, we’ve gained nothing (from a cost reduction perspective) because we won’t be able to eliminate the infrastructure needed to support either.

Taking it one step further, what happens if there is standardization on the underlying Cloud platform (CloudOS?) and one provider “patches” or updates their Cloud offering but another does or cannot? If we ultimately talk about VM portability between providers running the “same” platform, what will this mean?  Will things break horribly or be instantiated in an insecure manner?

What about it?  Do you see cloud computing as just an extension of SaaS and hosting of today?  Do you see dramatically different issues arise based upon the types of information and applications that are being described in this model?  We’ve seen issues such as data ownership, privacy and portability bubble up, but these are much more basic operational questions.

This is obviously a loaded set of questions for which I have much to say — some of which is obvious — but I’d like to start a discussion, not a rant.

/Hoff

*This little ditty was inspired by a Twitter exchange with Bob Rudis who was complaining that Amazon’s EC2 service did not have the MS08-067 patch built into the AMI…Check out this forum entry from Amazon, however, as it’s rather apropos regarding the very subject of this blog…

Arista Networks is a company stocked with executives whose pedigrees read like the who's-who from the networking metaverse.  The CEO of Arista is none other than Jayshree Ullal, the former senior Vice President at Cisco responsible for their Data Center, Switching and Services and Andres von Bechtolsheim from Sun/Granite/Cisco serves as Chief Development Officer and Chairman.

I set about to understand what business Arista was in and what problems they aim to solve given their catchy (kitchy?) tagline of Cloud Networking™

Arista makes 10GE switches utilizing a Linux-based OS they call EOS which provides high-performance networking. 

The EOS features a "…multi-process state sharing architecture that completely
separates networking state from the processing itself. This enables
fault recovery and incremental software updates on a fine-grain process
basis without affecting the state of the system."

I read through the definition/criteria that describes Arista's Cloud Networking value proposition: scalability, low latency, guaranteed delivery, extensible management and self-healing resiliency.

These seem like a reasonable set of assertions but I don't see much of a difference between these requirements and the transformative requirements of internal enterprise networks today, especially with the adoption of virtualization and real time infrastructure. 

Pawing through their Cloud Networking Q&A, I was struck by the fact that the fundamental assumptions being made by Arista around the definition of Cloud Computing are very myopic and really seem to echo the immaturity of the definition of the "cloud" TODAY based upon the industry bellweathers being offered up as examples of leaders in the "cloud" space.

Let's take a look at a couple of points that make me scratch my head:

That's  a very narrow definition of cloud computing and seems to be rooted in examples of large, centrally-hosted providers today such as those quoted.  This definition seems to be at odds with other cloud computing providers such as 3tera and others who rely on distributed computing resources that may or may not be centrally located.

While I don't disagree that consolidation through server virtualization is not the same thing as cloud computing, the statement that "basically all cloud computing applications today run directly
on a physical server without the use of virtualization or Hypervisors" is simply untrue.

I don't see how that assertion has been formulated or substantiated.

I'm puzzled when I look at Arista's assertion that existing and emerging networking solutions from the likes of Cisco are not capable of providing these capabilities while they simultaneously seem to shrug off the convergence of storage and networking.  Perhaps they simply plan on supporting FCoE over 10GE to deal with this?

Further,  ignoring the (initial) tighter coupling of networkng with virtualization to become more virtualization-aware with the likes of what we see from the Cisco/VMware partnership delivering VN-Link and the Nexus 1000v, Ieaves me shaking my head in bewilderment.

Further, with the oft-cited example of Amazon's cloud model as a reference case for Arista, they seem to ignore the fact that EC2 is based upon Xen and is now offering both virtualized Linux and Windows VM support for their app. stack.

It's unclear to me what problem they solve that distinguishes them from entrenched competitors/market leaders in the networking space unless the entire value proposition is really hinged on lower cost.  Further, I couldn't find much information on who funded (besides the angel round from von Bechtolsheim) Arista and I can't help but wonder if this is another Cisco "spin-in" that is actually underwritten by the Jolly Green Networking Giant.

If you've got any useful G2 on Arista (or you're from Arista and want to chat,) please do drop me a line…

/Hoff

I was reading an interesting article from James Maguire from Datamation that outlined various competitors in the virtualization platform space.

In the article, James referenced a Gartner slide that comparatively summarized hypervisor selection criteria including the maturity of features, pricing, management and ultimately security.  Unfortunately the presentation source of the slide was not cited, but check this out:


What I found very interesting was the security section which equated the security capability/maturity criteria of Oracle with that of VMware while at the same time demonstrating that the overall maturity/stability of Oracle was not has highly ranked. 

Since Oracle's hypervisor is based upon Xen and Citrix/XenSource is not ranked as high, it leaves me scratching my head.

Given that this chart references hypervisor selection to YE08, it more than likely does not take into consideration the coming vNetwork/VMsafe API's; it's unclear if this section is a measure of VMM "security" based upon published vulnerabilities, an assessment of overall architecture, the availability of security solutions in the ecosystem…

This is a very interesting assertion and I'd really like to get the entire document that describes how this was quantified and what it means.  Anyone know which report this came from?

/Hoff

I'm literally emulating a bobble head doll at this point.  In a fit of snarky confusion,  I'm simultaneously trying to nod-shake-shrug my oversize gourd to arrive at some commonsensical conclusion about this piece.  I can't, so my head just flops about like the headpiece on a 4-axis CNC machine.

Tarry Singh from the Avastu Blog spends his time as an independent analyst covering virtualization and cloud computing. His latest post regarding security left me scratching my head.

I had a bunch of folks ping me asking me for my interpretation of Tarry's latest work but I thought I'd turn it over to you lot since the more eyeballs the merrier.

Tarry's post is titled "Good News! Hackers Focus On Virtualization."

I read it.  I read it again.  I had something to drink.  I read half of it.

I think what Tarry's trying to say is that with more attention being paid to virtualization platforms by "hackers" that we ought to see increased pressure for more secure environments due to impending carnage from mounting exploits and regulators amassing mad virtualization audit skillz.  I could be wrong as it was really, really good wine.

Despite abusing the term "hackers," it's not an unreasonable assertion despite being dusty.  The rest of the post (or the wine) still leaves me a bit dizzy.

Pay attention now, I'll highlight the interesting bits in bold…

• This is a validation of the fact that Virtualization is going mainstream
• Security and Compliance will be core focus of all organizations
• Virtual Infrastructures are easier to battendown and secure due to its uniformity
  
• Regulators
  will increasingly ask for audits, where as in traditional environments
  (I've seen such audits by the like of KPMG etc) and always wondered
  like "wow–so are so prepared, dude, NOT!", Virtual environments
  suddenly enables auditors to ask the right questions and get or not get the expected results.
• Focus on security would mean that we will have to work harder to provide a secure and compliant platforms.

I'd be very interested to understand what a "secure and compliant practice" within the scope of a virtualized environment means, especially in light of some of the statements above. 

Tarry, you've got mail.

/Hoff

One of the things I'm very much looking forward to with the release of Cisco Nexus 1000v virtual switch for ESX is the release of performance figures for the solution.

In my Four Horsemen presentation I highlight with interest the fact that in the physical world today we rely on dedicated, highly-optimized multi-core COTS or ASIC/FPGA-powered appliances to deliver consistent security performance in the multi-Gb/s range. 

These appliances generally deliver a single function (such as firewall, IPS, etc.) at line rate and are relatively easy to benchmark in terms of discrete performance or even when in-line with one another.

When you take the approach of virtualizing and consolidating complex networking and security functions such as virtual switches and virtual (security) appliances on the same host competing for the same compute, memory and scheduling resources as the virtual machines you're trying to protect, it becomes much more difficult to forecast and preduct performance…assuming you can actually get the traffic directed through these virtual bumps in the proper (stateful) order.

Recapping Horsemen #2 (Pestilence,) VMware's recently published performance results (grain of NaCl taken) for ESX 3.5 between two linux virtual machines homed to the same virtual switch/VLAN/portgroup in a host shows throughput peaks of up to 2.5 Gb/s.  Certainly the performance at small packet rates are significantly less but let's pick the 64KB-64KB sampled result shown below for a use case:

Given the performance we see above (internal-to-internal) it will be interesting to see how the retooling/extension of the networking functions to accomodate 3rd party vSwitches, DVS, API's, etc. will affect performance and what overhead these functions impose on the overall system.  Specifically, it will be very interesting to see how VMware's vSwitch performance compares to Cisco's Nexus 1000v vSwitch in terms of "apples to apples" performance such as the test above.*

It will be even more interesting to see what happens when vNetwork API's (VMsafe) API calls are made in conjunction with vSwitch interaction, especially since the performance processing will include the tax of any third party fast path drivers and accompanying filters.  I wonder if specific benchmarking test standards will be designed for such comparison?

Remember, both VMware's and Cisco's switching "modules" are software — even if they're running in the vKernel, so capacity, scale and performance are a function of arbitrated access to hardware via the hypervisor and any hardware-assist present in the underlying CPU.

What about it, Omar?  You have any preliminary figures (comparable to those above) that you can share with us on the 1000v that give us a hint as to performance?

/Hoff

* Further, if we measure performance that benchmarks traffic including
physical NICs, it will be interesting to see what happens when we load
a machine up with multiple 10Gb/s Ethernet NICs at production loads
trafficked by the vSwitches.

How’d ya like this picture of “THE Cloud…”

This love affair with abusing the amorphous thing called “THE Cloud” is rapidly  approaching meteoric levels of asininity.  In an absolute fit of angst I make the following comments:

1. There is no singularity that can be described as “THE Cloud.” There are many clouds, they’re not federated, they don’t natively interoperate at the application layer and they’re all mostly proprietary in their platform and operation.  They’re also not all “public” and most don’t exchange data in any form. The notion that we’re all running out to put ALL our content and apps in some common repository on someone else’s infrastructure (or will) is bullshit.  Can we stop selling this lemon already? There will be lots of Clouds that we’ll spread much of our information and applications onto — some internal, some external, some public, some private….

   Yay!  More people have realized that outsourcing operations and reducing both OpEx and CapEx by using shared infrastructure makes sense.  They also seem to have just discovered it has some real thorny issues, too.  Welcome to the 90’s. Bully!Just like there are many types of real billowing humid masses (cumulonimbus, fibratus, undulatus, etc.) there are many instantiations of resource-based computing models that float about in use today — mobile.me, SalesForce.com, Clean Pipes from ISP’s, Google/Google Apps, Amazon EC2, WebEx — all “cloud” services.  The only thing they have in common is they speak a dialect called IP…

2. The current fad of butchering the term “Cloud Computing” to bring sexy back to the *aaS (anything as a service) model is embarrassing. More embarrassing is the fact that I agree with Larry Ellison wherein he stated:

   “The interesting thing about cloud computing is that we’ve redefined cloud computing to include everything that we already do. I can’t think of anything that isn’t cloud computing with all of these announcements.
   The computer industry is the only industry that is more fashion-driven than women’s fashion. Maybe I’m an idiot, but I have no idea what anyone is talking about. What is it? It’s complete gibberish. It’s
   insane. When is this idiocy going to stop?“A-Freaking-Men.

3. It ain’t all new, folks. Suggesting that this is a never-before-seen paradigm that we’ve not faced prior and requires entirely thinking as to privacy, trust models, security as a service layer and service levels mocks the fact that the *aaS model is something we’ve been grappling with for years and haven’t answered.  See #2.  I mean really.  I’ve personally been directly involved with cloud-models since the early 90’s.  Besides the fact that it’s become (again) an economically attractive and technologically viable option doesn’t make it new, it makes it convenient and marketable.  That said, we’re going to struggle with the operational and organizational issues and where theory meets practice on the battlefield.
4. Infrastructure Gorillas are clouding the issue by suggesting thier technology represents THE virtual datacenter OS. Microsoft, Citrix, VMware, Cisco.  They all say the same thing using different words.  Each of them claiming ownership as the platform/OS upon which “THE cloud” will operate.  Not one of them have a consistent model of securing their own vDCOS, so don’t start on how we’re going to secure “IT.”(Ed: In fairness just so nobody feels left out, I should also add that the IaaS (Infrastructure as a service)/integrator gorillas such as IBM and HP are also in the mix — each with their own flavor of service differentiation sprinkled on top.)

If you thought virtualization and its attendant buzzwords, issues and spin were egregious, this billowy mass of marketing hysteria is enough to make me…blog 😉

C’mon, people. Don’t give into the generalist hype.  Cloud computing is real.  “THE Cloud?”  Not so much.

/Hoff

(I don’t know what it was about this article that just set this little rant off, but well done Mr. Moyle)