LiquidIT

By: Vik Desai

Now that the marketing dust is beginning to settle, we can take a closer look at Acadia, the new JV from Cisco, EMC and VMware – analyzing the move and what it really is/means. There seem to be 3 major objectives to the strategy:

1. Accelerate mainstreaming of the unified computing market & adoption of hardware from the respective players, taking advantage of the relative weaknesses of HP, IBM and Dell who don’t have a credible response.
2. Align the otherwise independent Integrators (Accenture, TATA, etc) against EDS (HP), Perot (Dell) and IBM.
3. Create a vehicle (the JV) to recognize, separate and share revenues that come from Acadia.

I’ll address each of these points in separate blogs, starting with the first point in this posting, and then summarize the series with the last.

Point one is as obvious as you might guess. Cisco UCS is so darn complex to design, implement and support that the channel is rejecting it. No channel support means revenues won’t scale quickly. Since Cisco doesn’t have time to wait for the channel to ‘get it’, Acadia is going to step in and do all the designing, implementing and servicing of UCS sales. This will effectively make value added resellers more like sales agents. The launch offerings seem to embrace the notion that the channel needs easier solutions too. Did you notice that the difference between all of those Vblock configurations was essentially the size of the associated EMC storage?

Resellers thrive on the professional services associated with hardware sales, so I’m pretty darn sure that they’re not going to see the Acadia offerings as very friendly to their business. Apparently, having taken on HP and IBM for servers isn’t enough for Cisco. Now they want to take on all of the resellers too!

Our approach at Liquid Computing is to deliver products that are based on OPEN architectures, can FLEXIBLY accommodate virtualized or bare metal environments, and are EASY to implement by end-users and the channel. Liquid Elements is a great example: Intel servers and NetApp storage tied together by a few cables and our software fabric & switch.

By: Mike Kemp

There is a lot of buzz around unified computing since Cisco’s announcement, and it has been interesting to watch HP and others respond. Some analysts are heralding this as a new wave of innovation; and it certainly is, at least with respect to the computing industry incumbents. I have always believed that unified computing is the right architecture for the data center and will drive advances in virtualization, asset utilization, TCO reduction, applications and services.

However, while the term unified computing is new, the architecture behind it is not. What is unified computing? Essentially it is a collection of functions needed to deliver data center compute services (storage, networking, servers etc.), integrated at various levels, and managed as a single system with a high availability control plane to automatically configure resources and perform fault management. When I was in the Telecom industry the engineers were building unified systems with compute, storage and network over 30 years ago. It seems to me that the older mainframes meet those criteria as well.  I used the same unified architectural approach at Liquid Computing for our LiquidIQ product which has been shipping since Q3 of 2006. I personally think unified computing, driven by multiple vendors, will eventually be the predominant architecture in data centers.  What surprises me is why this move wasn’t made many years ago.

By: Mike Kemp

 I have been following the evolution of SAN for many years and observed with interest the heated debates of iSCSI versus fiber channel. More recently, Fiber channel over Ethernet (FCoE) joined the mix and looks promising as a way to achieve cost savings through converged networking.   Of course any new technology requires major standardization work.  To that end, the TII standards group is driving the FCoE standard while the IEEE data center bridging task group is driving the evolution of the 802.1 bridge specification to enable ethernet to support converged networks (including FCoE). It’s not an exact science to determine which new standardization initiatives become a commercial success, but FCoE has a good chance since several large vendors including Cisco have taken a leading role in driving its adoption. When Cisco announced its UCS product in March, it claimed it would fully exploit FCoE and, although this was ahead of standards closure, it was nevertheless a major step forward for commercial acceptance of FCoE. However, a recent blog post revealed that the company’s implementation of FCoE is not quite as advanced as anticipated. In Cisco’s UCS, FCoE protocols only operate through the switch and the southbound interfaces (to the servers); the northbound interface (to the SAN) requires native fiber channel and an associated fiber channel switch.  Clearly, more work is needed before widespread end to end standards based volume FCoE deployments becomes a reality.

By: Vikram Desai

It took a little while for calmer minds to prevail, however, the past few weeks have finally witnessed a return to common sense amongst reporters, analysts and prospective users of unified computing solutions. It’s as if the magical marketing dust sprinkled by the big players finally wore off, and everyone began to see the purported solutions for what they really are: full of warts and not quite baked.  In fact, the only standards-based unified computing solution in production today as verified by announced commercial customers doesn’t even come from HP or Cisco but rather from Liquid Computing.  This reality has left the two big players to battle one another for second place, and the mudslinging has started. Let’s take a look at some interesting recent events.

You all know how passionate I am when it comes to the issue of security. By removing the walls that separate servers, storage and switching you open the entire system up to a new generation of security concerns that cannot be resolved using traditional methods.  HP seems to agree, since they quoted me and Yankee Group analyst Zeus Kerravala as experts in their recent article (”One Giant Switch”), which exposes the holes in Cisco’s UCS product. Of course, the blogosphere’s reaction is that HP’s solution is no less problematic or complex since it’s based on old technology.

Next, I really have to salute Dan Kusnetsky of The 451 Group and ZDNet for going the extra mile for his blog to try and find a real, live production Cisco UCS customer to speak with.  You may recall that John Chambers proudly rolled out Savvis as an initial UCS beta customer. Well, that’s who Dan went to.  The final paragraph of the interview­­­ sums it all up:  “You’ll note that Bryan (Doerr)  always spoke about Cisco’s UCS as something for the future. This is because the first few configurations have only recently been announced and are not in use in enough places for a datacenter’s manager to have full confidence in them. Cisco’s competitors, on the other hand, have years of experience in the field”.

Bravo Dan!

iftyk5qc3p

Posted By: Vik Desai

Today in his Virtually Speaking blog on ZDnet, Dan Kuznetzky posted the following : 

“The rumor is that Cisco is going to create a joint venture with VMware and possibly EMC to spin off the unit that creates the UCS systems. Fun, no?” (see the full post here) 

I wouldn’t be at all surprised to see some form of consolidation take place over the short term.  Why?  Because both virtualization and unified computing solutions aim to drive down the time and cost of operating data centers; they simply address different aspects.  Virtualization seeks to help you do more with assigned physical IT resources (i.e. resources that have been pre-configured and set aside for use by the virtualization layer) whereas a properly designed unified computing solution enables the ability to dynamically reallocate and reconfigure the underlying physical IT resources either automatically or on command through software control in order to support the needs of virtualized and/or bare metal applications.  Apart, each delivers dramatic benefits.  Together, the benefits multiply.  So, the real question is: WHEN will partnerships amongst the handful of virtualization and unified computing vendors take place, and what form will these partnerships take?  For sure Cisco and HP will seek to lock you into a proprietary solution.  For end-users sakes, best to hope that at there’s one solution that stands on its merits and doesn’t require vendor lock-in.  Historically that’s been the job of a new market entrant.

By: Vikram Desai

In 2005 when I was CEO of web application security vendor Kavado, Cisco revealed a security vulnerability in their operating system (IOS) that could allow a hacker to take control of any product running IOS where web management was turned on.  The attacker didn’t even need direct access to the server; just the IP of the target device.  For example, using a now well known technique called cross site scripting, a hacker could inject random code into the Cisco device via its web interface, and then reset the password to gain full administrative privileges.  That means that the hacker controls that device, and soon your entire network.  Today, four years later, Cisco revealed that this vulnerability still exists. Apparently security isn’t that important to Cisco as a switching company – besides, the firewall will protect you, right? Oh yeah, if it’s a Cisco firewall it runs IOS too.

 Now imagine that you’ve bet the farm on Cisco and have deployed their Unified Computing Solution – controlled by (yes, you guessed it) – IOS.  That’s right, not only have you exposed all of your networking devices, but now because the manager for the entire system that controls visibility and access to your Cisco servers and FCoE storage is on an IOS powered switch, you’ve lost control over your entire data center.  A hacker could potentially start re-directing corporate data somewhere else, or minimally wreak havoc by turning applications on and off.

 Time to pack up and go home, because your days of gainful employment are over! 

 You don’t have to put all of your eggs into one vendor’s basket, and there’s no call to compromise on security. There are providers that deliver a complete and secure standards based solution without locking you into proprietary schemes.  In fact, companies like Virtuoso have proven out the dramatic and immediate game changing benefits of unified computing solutions.  However, one must be certain that the selected unified computing solution is complete.  It should employ a secure method that physically isolates the command and control structure managing the compute, networking and storage elements from the operating system(s), virtualized environment(s) and associated applications. This requires an entirely different “from the ground-up” architecture versus the patch work approach or bundles of existing parts and professional services offered by the big switch and server vendors. Beware, buying “big” doesn’t mean buying smart.

Just last month, I discussed the ramifications of security in a unified computing world with Chris Preimesberger of eWeek.  Reportedly, Cisco’s response to Chris when he asked them about the glaring lack of a security strategy was: “…in a UCS deployment, customers are expected to use their own existing server, storage and management security vendor-not one provided by Cisco itself.”

You may have noticed that relationships amongst IT vendors are really stirred up. Companies who were formerly “close collaborators” are now quietly aligning themselves for direct battle. Some attribute this to Cisco’s entry into the server space while others believe that it’s due to a larger need for vendors to expand their addressable markets into adjoining areas in pursuit of continued growth. I believe that both of these are actually symptoms of a more fundamental and quite logical next step in the evolution to the dynamic data center.

IT vendors have been pitching the concept of “dynamic” or “agile” data center for years, but delivering on the vision on piece meal basis. In almost every instance, solutions essentially introduced software based control over a very specific sub-set of data center operations. Enough to make a targeted difference but not enough to solve the overall problem. For example, Layer 4-7 application switches (i.e. load balancers) delivered the ability to quickly re-route traffic destined for one server group located within a specific data center to another server group that could be located in a different data center – all based on based on server availability. But when you think about it, this approach relies upon the need to have alternate completely pre-configured data center resources (well beyond the receiving servers) ready and waiting to be called into action. So, the switching of the inbound traffic is dynamic but really worthless to you unless you have the rest of the underlying IT infrastructure waiting for it. Can you imagine how many dollars are wasted annually (real estate, power, professional services, equipment, etc) on these partial solutions? Very inefficient, expensive and truly manual.

The ideal approach is one in which all underlying physical IT data center resources and their interdependencies can be manipulated via software as building blocks, and the system itself can autonomically provision and/or reconfigure itself in response to corporate policy and real-time business needs. No more overbuilding, lots of corporate savings and most importantly dramatic increases in customer satisfaction.

The recognized power of this approach is the real underlying cause for all the vendor stir. If you dig a bit further, you’ll discover a lot of buzz centered on Liquid Computing. Liquid has this powerful software based approach available today while everyone else is just starting to think about. That’s why I say, add Liquid and stir.

I’ve been on the road across the US and Europe for the past several weeks briefing major analyst groups (Yankee, IDC, 451, Gartner) and certain leading edge enterprise data center operators about the benefits of deploying a properly designed unified computing system. It turns out that several of these folks actually read this blog (gasp), and I was asked on more than one occasion to delve further into my opinions of the serious flaws and lack of execution that I see vacant in Cisco’s vision and HP’s releases. I (humbly) accommodated, but more importantly, seized the opportunity when I could to offer a demonstration of a live production UCS system so that I could show how these major oversights are the product of systems that “look like” UCS and in other cases fell short and could lead to security issues. It was priceless to see the facial reactions as they saw how through powerful software control, very complex data center environments (switching, computing, storage and all associated physical and logical connections) could be deployed at once – in minutes – and visualized in seconds. How virtualized environments and associated applications could be interchanged with entirely different bare metal environments and configurations at the touch of a software button, and how the entire standards based ecosystem was designed to defend itself from attack and self-heal. You could actually see the lights turn on in their heads as they realized that solutions marketed to them from the big infrastructure players were missing major pieces of the puzzle (you know, like being able to configure and provision storage). That’s because in some cases what’s being marketed is the perception of a unified computing solution and in other cases the offering isn’t yet complete or mature. I suppose that explains why Cisco alone has visited the Liquid Computing website from its corporate locations more than 70 times since May!

So, what defines a mature unified computing system? A flexible, dynamic and secure system that brings together standards based (compute, network and storage) capabilities under the software driven command and control of a single architecture. Guess what, I’ve just loosely described the fly-by-wire system of a modern jet, where standard parts are controlled by software that makes decisions far faster and more accurately – without error – than a human being can. In fact, a human being cannot actually fly a jet without the assistance of software control, and it’s the software control that delivers all of the benefits. Without it, the sum of all the other parts is no more or less than an ordinary plane.

By Vikram Desai

There’s been a lot of speculation as to why Cisco decided it had to enter the server market space and why they completely ignored bare metal or “non-virtualized” applications (which according to Yankee Group represent 80% of all enterprise data center applications). To many, the move made no sense because the margins associated with compute gear would severely dilute those of the lucrative switching business. Others believed that they entered the server business as a reaction to a war HP started when it entered the switching business with Pro-Curve. Well, I suppose that could be true if the real world were more like “Survivor”, but it’s not. Companies of Cisco’s caliber seek paths to market dominance at every opportunity. However, when they discover that fate is challenging their domination, they tend to make big moves – reacting with a grand new vision in an attempt to “change the course of a mighty river”.

Such is, I believe, the motivation behind their seemingly half baked, rushed to market before it was ready, unified computing story. Cisco recognizes virtualization as a major market threat and is willing to invest extraordinary time and resource in order to wrestle control away from fate. By their own account, Cisco had 500 engineers assigned to Nexus (UCS) and spent $250MM in cash on it over a 2 year period. That figure doesn’t include the cost of purchasing Nuovo (up to $678MM), whose technology is at the heart of UCS. Note that Ed Bugnion, one of the founders of Nuovo, was a co-founder of VMware, and of course Cisco actually owns a piece of VMware, so they know where that company is headed. Also note that whatever innovation there is in the UCS system, including their proprietary VN-Link technology, is designed to shift control of virtual networking back into Cisco’s physical network infrastructure and management systems. Well isn’t that convenient.

So what’s the concern over virtualization and why is Cisco willing to pay nearly $1B over it? It’s simple. Current and emerging virtualization technologies from VMware, Citrix, Oracle and Microsoft have shined a new light onto a different, more effective and better way to manage data center resources than Cisco has championed for the past 20 years, and the success of virtualization technologies is shifting control of the data center network away from Cisco. In the past, vendors who provided and managed switching and routing within a data center of any size had ultimate control. Now, more and more each quarter, as virtualization technologies evolve, control over WHAT rides across the network, HOW it rides across the network and WHEN it rides across the network is shifting away from the silo’d physically wired world into the domain of the virtual world without walls. Unless Cisco can control virtualization and make it part and parcel of the Cisco Powered Network, their bread and butter revenue is going to go away faster and faster every year – along with the fat margins.

This is the light that Cisco sees at the end of the tunnel, a light that they must transform from a freight train’s headlamp into the light of opportunity.

By Vikram Desai

Holy Cow!  Have you ever before seen such thick marketing hype encased around such little substance? In what should otherwise have been an exciting series of product announcements that celebrate a revolutionary way of creating Dynamic Data Centers, Cisco and HP have managed to roll out partial or cobbled together “new” solutions that fall far short of actually doing anything new to deliver unified computing today.

Last week, Cisco held a lengthy webcast sequel to its blockbuster Unified Computing show that was supposed to delve more deeply into the details of their UCS solution.  Between it and a tag along PowerPoint deck, they succeeded in comparing their blade’s performance to those of other computing blades, provided a very deceptive pricing comparison (because the vendors being compared to kept changing from page to page and it skipped over any associated recurring costs), and  – never addressed storage!  So much for unified computing – you know, compute/network/storage.  It sounds more like converged switching and compute (emphasis on the switching).  By the way, did you catch in the Q&A section where Cisco was asked how UCS stacked up against Liquid Computing’s LiquidIQ solution?  In short, Cisco claimed that Liquid Computing’s solution was based on Infiniband and that it required bridging to support  Ethernet. That would have been true for the company’s first product release – in 2007! Cisco needs to get its facts straight.

HP’s product, “BladeSystem Matrix and its Matrix Orchestration Environment”, sounds like an old movie favorite of mine recast as an off-Broadway show.  When a vendor resorts to such theatrical conventions, they’re probably trying to shift attention to the name rather than the substance.  Well, this appears to be the case here since BladeSystem (BS for short) is truly standard commercial piece parts that have been glued together at the management level rather than having been purposefully created to deliver seamless and unified compute, storage, and networking from the ground up.  Timothy Prickett Morgan of The Register called it “an amalgam of tools that HP has acquired and created from scratch to provision and manage server environments in a more autonomic fashion than has previously been possible” (http://www.theregister.co.uk/2009/04/20/hp_matrix_blades/). Of course, all HP parts are required.  The most amusing part of HP’s offering to me is the enormous price, combined with the bloated professional services required to assemble it all together.  Why is it so funny?  Well, because it makes perfect sense for a solution that’s comprised of pre-existing piece parts.  The Register reports that “a Matrix blade setup would cost in the same ballpark as a rack of servers with the same number of physical servers and storage”.  Now that’s special.

John Brodkin of Network World provided fact-based coverage of the recent announcements as well (http://www.networkworld.com/news/2009/042009-hp-bladesystem.html).  His complete coverage included mention of the only privately held unified computing company that’s shipping a complete solution today – Liquid Computing – whose solution supports both virtualized and bare metal applications while completely addressing compute, networking and storage needs.