One of the more interesting companies at VMworld this year is a small start up called Xsigo.  In my role as an analyst, I see loads of start ups and there are very few that have technology that is differentiated enough for me to consider them “game changing.”  Riverbed was one, Xangati another and most recently SeaMicro are a few and Xsigo is another.

At VMworld this week, Xsigo released an Ethernet-based virtual I/O solution which will allow virtual I/O to move from a niche technology to a main stream technology that can be used to scale virtualization and cloud environments.  For those who aren’t hard core, data center geeks, I’ll explain the significance of this and I’ll start with a definition of I/O and virtual I/O.

I/O is the term used for the flow of data to and from a device attached to a network.  When you save a file to the network, that’s  I/O.  Xsigo and other virtual I/O vendors deal with server I/O, that is information sent to and from a server.  Virtual I/O is a term used to describe the “splitting” a single physical I/O connection into multiple virtual I/O connections.  For example, if I had a server and I needed four network connections (NICs) and three storage connections (HBAs) without virtual I/O, I would need several physical I/O cards in the back of a server.  Generally new I/O cards are added when new VMs are added or connecting to different storage systems, etc., so it’s not uncommon to have many I/O connections in one server.  Virtual I/O would allow me to deploy a single physical I/O connector and then create seven virtual ones on the server itself.  If I need an eighth, I just create yet another virtual one.

Although this sounds simple, traditional virtual I/O requires the opening up of a server, the possible removal of I/O cards, installation of new cards, running new cables and then some stuff to be installed on the server.  Anyone that’s ever worked in a data center (and I spent the first 10+ years of my career in a data center – that’s why I’m such a fun guy!) will tell you that if requires opening up of servers and running new cables, it isn’t practical enough to deploy, no matter how big the cost savings.  That’s why virtual I/O technology has primarily been limited to new server implementations.

What Xsigo launched this week is that ability to create the virtual I/O connections over a standard Ethernet connection.  Almost every server comes with an Ethernet connector integrated on the mother board so the connector is already there.  No more cards to install, no more proprietary cables and no opening of servers but all the benefits of virtual I/O without the installation mess.  Another thing data center managers will tell you is that the more than can be done over Ethernet, the better.  Ethernet is simple to work with, it’s low cost, highly resilient and has become the preferred connectivity method in data centers.  Ethernet based virtual I/O brings the benefits of virtual I/O without the installation headaches and that’s the reason I think it’s got game changing potential.

A10’s claim to fame has been that it can offer the performance of the market leaders such as F5 and Brocade but at a lower price making many wonder how they were able to get as price aggressive as they have been and still maintain traditional industry margins.  It appears now though that A10 may have taken some short cuts in product development.

Back in April, both F5 and Brocade filed lawsuits against A10.  F5, who should have sued over the similar naming of A10 versus the market leader F5, filed a lawsuit over two patent infringements.  Brocade, launched its own suit that claims A10 violated 7 patents.  Earlier this month Brocade extended the suit and filed a patent infringement and trade secret misappropriation case against A10 and three A10 employees that were former Foundry employees (Foundry was acquired by Brocade, hence the suit comes from Brocade now).

The case alleges that A10s AX series infringes on nine US patents.  It also claims that A10 CEO Lee Chen, CTO Raj Jalan and software manager Ron Szeto all misappropriated trade secrets including the source code from ServerIron, the Brocade ADC.

On Wednesday July 21st, the company announced its June quarter earnings and absolutely crushed it; and raised guidance for the 5th quarter in a row.  The company posted revenue of $230.5 million, well ahead of Street expectations.  According to Yahoo, Street consensus was $218 million but I had heard that the whisper number from the Street was $223 to $225 and F5 beat that number as well and then gave guidance of 5-7% sequential growth.  Not bad for a company that everyone thought Cisco would run over a couple of years ago.

One of the questions I’ve been getting, particularly from investors, is what’s fueling the growth in this market and I think there are a few different trends:

• The enterprise demand for ADC (application delivery controllers) has never been stronger. All of the technology initiatives that ADCs support are red hot right now.  Exchange upgrades, Share Point implementations, data center consolidation, virtualization, etc. are all big drivers for ADCs and these are red hot right now so F5’s been going along for the ride.
• New market opportunities in both traditional wireline telcos as well as mobile operators. The mobile tsunami that’s coming will definitely benefit F5 and telco now represents almost 25% of F5 revenue.
• There’s a definite lack of competition in the ADC market. F5’s biggest competitor is Cisco but the ACE product from Cisco is miles behind F5 in features and functions.  Customers that buy ACE buy it because they’re a big Cisco shop and they buy everything Cisco.  Almost every company I talk to that actually does even a little bit of testing settles on F5 over Cisco.  F5’s strongest technology competitor is the Netscaler product from Citrix but the majority of the Citrix channel sells Netscaler to support Citrix software and not for broader purposes such as Exchange and Share Point.  There a few smaller competitors that give F5 some challenges in specific areas such as Radware in telco and A10 in Japan but no one that challenges them cross regions and cross industries
• The DevCentral community. F5 has a simple scripting tool known as iRules that customers can use to create a bunch of custom features.  In conjunction with that F5 hosts a community of iRules administrators and developers called DevCentral which may be the best example that I’ve seen of how a Web 2.0 tools can be used to drive product loyalty.  Over the years I’ve said that DevCentral is F5’s secret sauce and has created loyalty to the product that only gets stronger as the community gets bigger.

So I’ve shared the good but there are a few areas of concern:

• Of the companies $230 million in revenue, 218 was the core ADC products. Over the years F5 has made several acquisitions to move into new markets such as Acopia (storage virtualization), SWAN (Wan Optimization) and uRoam (SSL VPN) and none of these have been significant revenue contributors.  ARX (formerly Acopia) did do about $6 million in revenue, up about 25% year over year but considering it paid $210 million for it almost three years ago, it’s fair to say it’s a disappointment.  If F5 can ever get one of these other products to fire only just a few cylinders, that could help create yet another growth initiative for F5 and allow it to move into more of a suite sale than single product.
• The Cisco factor. Many of the industry watchers I know still look at Cisco as F5’s biggest threat.  There have been some rumors that Cisco had been doing some due diligence and possibly acquiring A10 (which does compete well against F5 for raw performance).  In theory moving A10’s strong product into Cisco’s channel could cause F5 some problems, but frankly I’m not buying it.  As strong as Cisco is in networking, application networking seems to be an area Cisco just can’t get a handle on.  The acquisition of FineGround was supposed to be the end of F5 but that never played out and I really don’t think Cisco + A10 will be either.
• Maintaining the growth story becomes harder over time. F5 is expected to do roughly $850 million in revenue for its FY10 which means investors are expecting big things to justify its $6 billion market cap.  Maintaining the growth trajectory comes down to strong execution but it also needs the technology trends that are fueling the growth to continue.   This is one of the biggest reasons that the acquisitions need to start bearing some fruit soon.

Overall though I think the tailwinds that are pushing F5 will be maintained for the foreseeable future and I’ll maintain my stance that I think they will be among the top performers in the networking sector.  Cloud computing, virtualization, mobile applications and the other initiatives driving growth just have too much momentum behind them right now.

The announcement included a 72 port 10GbE 1U switch and a 24 port 10GbE blade switch. Delivering 72 ports in a single rack space is an impressive feat and is currently industry leading. This isn’t the first time Oracle has shipped an innovative port dense switch as Sun rolled out a 3456 port InfiniBand switch back in early 2008. To get 72 ports in 1U, QSFP connectors are used to aggregate four 10GbE links into a single 40 Gbps port. Although QSFP is a newer connector, it is starting to see some traction for use in InfiniBand and 10GbE products. By designing the blade and the top of rack switch together, they achieved a significant reduction in cables with only three uplink cables (rather than 10) required to build a non-blocking fabric. Cabling is frequently a major source of cost, complexity and trouble, so this is something that will be appreciated in many data centers.

But there are a couple of questions that remain unanswered with this announcement. A critical piece that is still more vision than reality is Oracle’s ability to deliver the software required to manage virtualized networking infrastructure. Oracle has a strong management and virtualization portfolio with a combination of products from previous acquisitions as well as x86/SPARC hypervisors and virtualization tools picked up from Sun. As well, they are busy integrating Sun’s xVM Ops Center software with Oracle Enterprise Manager. This combination has the potential to deliver a complete solution for managing a complex virtualized data center, but it’s still just potential and only time will tell how Oracle’s network products and management software fit together in Oracle’s overall strategy.

The other big question is: What are Oracles long term goals in networking? Right now, they have an industry leading (but not market leading) product that will primarily be used as part of an Oracle implementation. That’s probably good for maybe 1/4 percent of market share at best. But the trend of integrated computing and networking is here. Brocade bought Foundry, Cisco launched its own server, HP acquired 3Com, IBM and Dell are partnering with Juniper and Brocade and the list goes on, so it shouldn’t be a total shock that Oracle is making a play.

In my opinion, with the shift if buying centers being in flux because of virtualization and cloud computing, this is the perfect time for Oracle to jump into this and they should go out and build a portfolio through acqusition. They could easily afford Brocade or, if they wanted to go smaller, Oracle could buy Force10, Arista or Voltaire.

We’ve seen a lot of activity in the data center over the past two years and now We’ve seen a lot of activity in the data center over the past two years and now Oracle is throwing its hat in the ring. Is it serious or just background noise? Time will tell.

The Cius is a tablet computer, there’s no doubt about that, but that’s probably the only commonality with the iPad.  The Cius is meant to dock into the base of a phone and makes a video phone.  The device can then be undocked and carried around as tablet.

The Cius is designed to be a communications tool.  It’s optimized for video, social networking and other collaboration tools.  The target audience will be teachers, nurses, doctors, field service workers and other transient campus type workers.  In many cases this might wind up being a shared device.  The device isn’t meant to be a personal entertainment tool, e-reader or a device to use productivity applications like word processing and spreadsheets.  The Cius is also meant to be sold to companies and then supported through the IT department.  In most cases it will be an augmentative device to the things being used, not a replacement (except for maybe the phone).  Lastly, the Cius is built to have security integrated into it.  For markets like education and healthcare, security is a must.

Conversely, the iPad is consumer device first.  The iPads are brought into organizations through the users and, in many cases, self supported by the community.  It’s a music player, video player, personal productivity tool and is meant to be a replacement for another device that a worker would carry.  Security is an afterthought to iPad.  Part of the appeal is the coolness factor.  The Cius is a corporate class device, the iPad is a consumer class device.

For Cisco, the fact it’s an Android based device does signal some winds of change at the company.  Instead of building its own operating system and interface, it chose to work with Android where there is already strong developer support.  Had Cisco chosen to go down this path for its IP phones, maybe we would have seen a number of applications for the phones that would have driven additional value to the VoIP industry.  This should also help Cisco build up its own fledgling Cisco Developer Network (CDN) community, by being able to reach out to the Android developers.

Looking ahead, I see Cisco continuing to deliver pro-sumer type devices that can raise Cisco’s overall brand awareness with corporate workers and consumers, not just IT pros.  It looks like it will be a great corporate device–complementary, not competitive, to iPads and other tablets.

The following day, Brocade Communications held its annual technology day amid similar hype (at least in the tech world) where many people (myself included) thought the actual content would fall well below the expectations of hype. Tech day and the related announcements was a breath of fresh air for the beleaguered Brocade as the company has been under fire from Wall Street and the media for the erratic performance of the SAN and networking divisions over the past couple of quarters.

At the event, Brocade announced its vision for how they see the data center evolving and the role the network will play as well as how Brocade plans to solve the problem. The vision, a data center that is fully virtualized where the network plays in key role in this transition, is similar to the data center vision outlined by many of Brocade’s competitors such as Cisco, Juniper and Force10.

The surprise of the event, and where Brocade exceeded expectations, was in the product detail.  Under a new marketing umbrella called “Brocade One” the company launched its Virtual Cluster Switching (VCS) architecture and related products. This announcement has been one of the missing pieces in the network landscape puzzle, which we pointed out in a previous blog (see Virtualization threatens the status quo in enterprise networking).

VCS promotes a scalable, “flat” network fabric through a trill based clustering technology. VCS will be delivered via a set of Ethernet switches that are driven by a Brocade operating system which takes the Fibre Channel OS and Foundry’s OS and converges them onto a common platform. This architecture and product is the first fruit born from the acquisition of Foundry by Brocade. The VCS products will be delivered now through mid 2011.

The fabric strategy is similar to the fabric launch that Juniper had earlier this month with a couple of differences. Juniper’s Stratus fabric promotes the concept of a single control plane to make every packet no more than one hop away from its destination. However, Brocade’s fabric is storage aware where that was alluded to by Juniper but no details were given. Additionally, Juniper’s Stratus products will not be available until mid to late 2011 given Brocade an early mover advantage over Juniper.

Brocade also demonstrated Virtual Compute Blocks, which will simplify data center management by tying the network and compute stacks together, similar to Cisco’s Unified Computing System but Brocade’s solution works with any storage and server platforms.

This announcement doesn’t solve some of the short term problems Brocade has been having balancing the execution of its SAN and network business but Brocade One and VCS puts the company in a strong long term position. Over the next 12 months the data center networking industry will continue to become more and more competitive.

To date Cisco, Juniper, Force10, Arista, HP and now Brocade have all outlined unique strategies to solve the challenges that virtualization and cloud computing cause in the data center. The lone vendor left to show its hand is Nortel (Avaya now) or it will miss out on the early adopters, which is were many best practices are developed.

Overall, Brocade’s Tech Day did what it needed to do. It exceeded my expectations and many of the people I talked to at the event and left me incrementally positive on its chances in this impending data center war.

Now that Strasburg has won his second game, let’s see if Brocade can keep the momentum going.

• Simplification of access control through centralized management.  The BIG-IP Edge Gateway will allow companies to implement access control and encryption policies to secure the data as it travels to the cloud.  (http://www.f5.com/pdf/white-papers/securing-the-cloud-wp.pdf).  This can be done today but with great difficulty as there hasn’t been a central point of control.  This has the potential to remove one of the bigger barriers to cloud computing.

• Securing and improving the experience for the mobile worker.  The F5 BIG-IP Edge Client will now cache the Windows log on credentials of the user the first time the information is entered.  This will create a common access method for users whether the application resides on premise or in the cloud.  For many users, this can make a big difference.  Many workers will shy away from using cloud-based applications since the access experience is foreign to the normal work environment.  This should bring a level of consistency to the end user experience.

Obviously F5 isn’t solving all of the security problems with cloud computing but these enhancements will allow companies to at least get started with cloud while maintaining their own security policies rather than having to adopt the policies of the cloud provider.  Over time, the cloud provider’s security may prove to be on par or greater than that of the enterprise, but this needs to be proven out.  The F5 solutions allow companies to kick start their cloud strategy while continuing to investigate the capabilities of the cloud provider.

For F5, this is an important step in transforming itself.  The company has experienced phenomenal growth over the past year (its stock has gone from under $30 to almost $70 right now) with much of the growth being fueled from the enterprise data center. Vendors need to continue to provide such solutions as organizations move along the adoption curve with virtual editions of more solutions to ensure consistent control and management capabilities as architectures broaden to encompass more cloud in more complex ways.

When I talk to data center managers and VM administrators about this trend though, one of the big inhibitors is the network itself and its inability to meet the demands of a fully virtualized data center.  In essence, the network will be the backplane of the virtual computer and the network needs to deliver traffic as fast as possible with minimum latency and packet loss.  Todays traditional layer 3, multi tiered data center networks were never meant for this kind of traffic causing organizations to have to re-evaluate their networks needs.  The vendors, in turn, have responded accordingly, developing new solutions to handle this kind of traffic.

What I’ve noticed over the past year or so is that all of the vendors that play in the data center are choosing a different way of solving the problem, causing the first real challenge to the status quo in a long time.  With all due respect to the network vendors, there hasn’t been that much innovation in networking over the past decade.  We’ve had some increases in speed (FastEthernet to Gig to 10 Gig), the advent of power over Ethernet and a few new features such as horizontal stacking but, for the most part, most of the vendors settle into the same feature set and the only way you compete is on brand and market share (Cisco), being faster than Cisco (Force10, Brocade) or providing a better TCO than Cisco (HP/3Com).  But basically, the feature set of what you get is pretty much the same from vendor to vendor.  That’s one of the reasons why Cisco has the market share it does.  All things being equal make the safe choice, and that’s Cisco.

Monday’s announcement of Juniper’s ‘fabric’ strategy is a good example of how a vendor is choosing to solve the problem differently than the incumbent vendor.  For those that don’t understand what a network fabric is (since the term is throw around loosely by the vendor community), think of being able to extend a switch outside the box.  A switch itself is pretty simple.  It’s made up of a bunch of ports and a fabric that connects all the ports.  When traffic moves from one port to another, it makes a single hop to that other port because they’re all connect to one other.  All traffic in a switch is a single hop away.  If you could extend that concept outside the box, you would have a data center wide network that looks like a single switch.  All traffic is virtually only ever one hop away minimizing the latency in transmission and simplifying the architecture.  In a traditional network, the packets must go through several hops (one at each tier) with packet processing done at each hop.  Even though this is done at wire speed, this process can cause enough latency for the movement of the virtual workload to fail or at least perform poorly.  Juniper’s strategy here is to create an underlying network fabric that’s capable of being that backplane of the virtual data center.  It’s a great concept and should create some differentiation for Juniper in its quest to gain share.

Cisco’s solution to this problem is Unified Computing, where it built a platform that has computing and network technology integrated together.  When a virtual workload is passed from one server to the other, the UCS (Unified Computing System) handles all of the interplay between the virtual servers, physical servers and network.  I’ve talked to a number of customers that have used UCS in a lab environment and the feedback has been positive but the majority of deployments are in the labs today.  A 2010 moving into 2011 challenge for Cisco is to take many of these lab deployments and move them to production to prove it works in demanding environments.

Force10 launched its automation strategy at Interop where it flips the model of where Cisco and Juniper are.  Instead of making the data center network centric, it is choosing the keep it server centric and the network will act as a “slave” to the compute technology and will allow on the fly provisioning of QoS, VLAN assignments, ACLs, etc by the compute systems.  This solution does help promote the status quo in the data center by leaving it compute centric.

Arista launched its entire company around the concept of virtual computing delivery.  With this solution the network device operating system is virtualized and includes an integrated hypervisor, so the switch itself becomes an application deliver appliance.

Even Avaya through its Nortel acquisition has thrown its hat in this ring.  Many of the innovative features that Nortel’s data products have (such as split multi link trunking) will help with the movement of virtual computing traffic.  The company just needs to drop its vision and architecture in front of the infrastructure to bring more credibility to its solution.   Now, vision and architectural messaging hasn’t been a big strength of Nortel’s in the past but the Avaya management team is much stronger than Nortel’s and we’re expecting to see more from Avaya throughout the balance of the year.

The industry is still waiting for a solution from Brocade but it bought data center specialist, Foundry Networks, because of this trend and we’re expecting to see their vision and solution soon.

HP Procurve is the other vendor that’s missing from this dance right now.  HP launched Matrix last year but it seemed to be more of a stop gap solution.  Obviously HP has been busy integrating 3Com but it has the largest set of assets (computing, networking, storage, management software, etc) to build a credible solution.

So what does this all mean?  It means that for the first time in a long time, simply buying the status quo may not be the right decision.  Much of the decision may depend on whose compute and virtualization technology is used.  The ability to solve the virtualization problem will far outweigh brand, the number of certified engineers or market share as this is the computing foundation companies will use for the next decade.  Only time will tell whose solution works best but for the first time in a long time, we’ve got a fair fight on our hands.

Interop rolled around in May and saw Cisco release its “Clean Air” wireless portfolio.  Clean air allows the Cisco wireless solution to continually scan the “air quality” by looking for things (microwave, camera interference, etc) that can interfere with the WLAN operations.  The solution then automatically adjusts the configuration to ensure optimum performance.  Clean air is a significant release for Cisco as this is the best competitive feature Cisco has had outside of its integration with the wired network.

Interop also saw long time, but recently forgotten about network vendor Nortel release it’s own WLAN portfolio.  Of course this was released under the name Avaya, who acquired the Nortel enterprise business last year, but the technology was all Nortel and is optimized to run on a Nortel switched network.  Despite the rapid share loss Nortel saw over the past five years, the install base of Nortel switches is huge, about 10% of the overall install base by our estimates.

All of this activity has raised the question as to whether market leading, but pure play, Aruba Networks could maintain the momentum it’s had over the past year.  Last year at this time Aruba’s stock price was a little over $4 per share.  Since then the company has gone an incredible run, gained tremendous momentum and now has a stock price of over $12 per share.

On Monday, Aruba made its latest move by acquiring outdoor mesh networking vendor, Azalea Networks.  Although Azalea is headquartered in Silicon Valley, Azalea has a sizeable install base in China and brings with it a low cost engineering facility in Bejing creating a low cost R&D center that Aruba will be able to leverage as it develops new products.

This move opens up a huge opportunity for Aruba to extend the run it’s been on over the past year.  While the primary use of WLAN has been for indoor deployments, outdoor connectivity is the next frontier for this technology.  As organizations deploy applications such as industrial networks and smart grid, the use of outdoor mesh will accelerate. Azalea’s mesh technology combined with the traditional Aruba products will enable Aruba to go to market with a total solution for both indoor and outdoor WLAN.

So, while Aruba has the most current move in the WLAN market, we don’t expect it to be the last.   We’re expecting HP to leverage the WLAN products from the 3Com acquisition.  Additionally, valley rumors indicate that Juniper Networks will be releasing its WLAN solution over the summer months adding to an increasingly crowded market.

The desire to have workers connect from anywhere will continue to drive the WLAN market forward and we’re looking forward to a second half of the year that was as busy as the first.

Virtualization management is ultimately where these vendors will run their course in this market.  Everyday, dozens of elements managed in their frameworks are “falling off the grid” and popping up somewhere in VMware’s vCenter–which is rapidly evolving into the new dashboard for the new IT organization.  Yet, VMware is not sticking it to the big guys on their own, they are fostering a whole marketplace of start-ups allowing them to leverage their APIs and their virtual appliance construct.

One such company is Xangati, which briefed me earlier this week about their two new virtual appliances focused on virtualization management.   The power of the virtual appliance model is that it allows Xangati to have an enterprise-wide deployment within a couple hours (covering both the virtual and physical infrastructure) versus a framework which can take weeks and months to get started.  Even more importantly they are riding the wave of virtualization economics through being delivered as a virtual machine.  As a result they can as can others drive down the cost of management to a fraction of what it used to be–in the same way that VMs have done for servers.

What Xangati is capitalizing on is that there is a screaming need for management solutions that provide high value in the virtual world.  First of, you have to be virtual to properly manage the virtual.  Secondarily the value that is provided has to be focused directly at the virtualization teams first, since they are the ones feeling the growing pains with virtualization.  When there is a breakdown in virtual server or desktop performance the calls are channeled their way and not to IT operations.  And in turn they will use vCenter to first get their bearings on an issue, but they need more than just element management.  That is the void Xangati fills by focusing on how those elements are communicating and what they are doing at the time of a performance problem.  Why wouldn’t someone fill that visibility void at a mere $299 dollars a hypervisor?

That is what I like so much about this virtual appliance model, it that it takes the elongated business case out of the equation and allows the doers to buy what they need to by when they buy it.  It is after all this kind of model that has been driving IT over recent years–the organic growth of productivity solutions.  Will the incumbent management vendors go down this path?  Unlikely, but companies like VMware and Xangati will be happy to let them stick to their slow growth markets.