Intel recently rallied some partners and made announcements about support for Open Fibre Channel over Ethernet (FCoE), so I wanted to take a few moments to be open about Open FCoE and Emulex’s position on the subject. As we know in our business, “free” only addresses acquisition issues, not the real total cost of ownership (TCO) or return on investment (ROI), and as we learned from watching the adoption of Enterprise Linux, “free” does not mean that there is no revenue. Emulex is already leading the market with our Universal LAN on Motherboard (ULOM) that has “included” FCoE and is shipping with HP today. Our ULOM has “included” hardware FCoE as part of their base G7 blades and we will see more of these solutions as we move into next-generation server solutions. So, if we take the price issue off the table, then the real questions are:

1. What is the best implementation for your consolidation plans: hardware FCoE Offload vs. Software Initiator?
2. Which solution will lower your operational expenses (OPEX) for management, power, cooling and cabling?
3. Are you buying new, faster CPUs to maximize consolidation by running more virtual machines (VMs)/CPU or doing it to run software I/O drivers on the CPU that prevent maximum consolidation?
4. Who are we going to call for 24/7 support?

But first, let’s take a look at the core messages of the Open FCoE announcement:

1. Innovation vs. Infrastructure – The real theme Intel is trying to say is that since the Open FCoE driver is “free” of acquisition cost, it provides IT managers with the opportunity to spend more on innovation. Adding network convergence to data centers via unified fabrics is innovation and also a change in infrastructure. Now we are just arguing about the acquisition cost of FCoE vs. the TCO of “included” FCoE. We all know that in the open source world, you can pose questions and many times get help and new code to address issues, but who supports it now, is it interoperable and what happens when the inevitable vendor finger pointing occurs? Innovation needs to be balanced with proven technology and enterprise-class support to make it real innovation.
2. Interoperability – Many can argue whether Fibre Channel should be as complex as it is, and that it should be more like Ethernet and just plug and play, but that is not the case. Vendors like Emulex have spent more than a decade building a deep and proven set of interoperability tools, exception cases and error handling that can’t be replicated overnight. Free is not the primary criterion for enterprises — it is performance, reliability, availability and support.
3. Unified Fabrics Are Good – We could not agree more. We have been a champion for network convergence and are a leader in delivering solutions, interoperability and ease of deployment to make network convergence a proven and deployable solution for the enterprise. In fact, Emulex was first with second-generation 3-protocol offload CNAs, first with virtual Network Interface Card (vNIC) technology for IBM, Cisco and HP converged fabrics, and the first to deliver third-generation converged ULOMs with “included” FCoE.
4. Open FCoE Is Free – The general gist of the Open FCoE release is that since Open FCoE is “free,” then it is better for IT and will enable them to spend more on innovation. If our industry has learned anything, “free” acquisition costs do not always equal the best value, the lowest TCO or the best ROI. Emulex is not saying the acquisition cost is not important, because it is. What we are saying is that the proven Fibre Channel stacks from Emulex and other vendors have more than a decade of performance and error handling optimizations and fully integrated management tools that make our hardware-based FCoE the best technology to implement to lower operational costs. As one customer said at our financial analyst day, “We have had Emulex’s FCoE running for 11 months and have had no problem with it.” Without a doubt, one of IT managers’ biggest concerns is too much cost spent on maintenance and not enough spent on innovation–our own CIO said this same thing in our cross-functional IT meeting last week–but we don’t think Open FCoE solves this problem today. To dig deeper on the cost deltas of Open FCoE vs. “included,” check out our Convergenomics™ Calculator to see how much convergence saves. The calculator allows you to change the price of the CNAs so you can see the real delta between 10Gb Ethernet (10GbE) NICs with Open FCoE and our UCNAs.

So, to get to the bottom line, here are Emulex’s key thoughts on Open FCoE vs. “Included” FCoE:

1. Emulex Is a Champion for Network Convergence and Unified Fabrics – Some folks like to say if you don’t immediately support Open FCoE, you are against network convergence, unified fabrics and cost savings; needless to say, this thought is off-base. Emulex has been a champion of network convergence and is a leader in host and target connectivity for unified fabrics. Open FCoE is a potential new tool for accomplishing this task. We do feel our current implementation and future roadmap of FCoE solutions offer a better value, an optimized deployment experience and lower TCO/ROI for IT managers because of our proven Fibre Channel stack, full hardware off-load, industry-leading interoperability, the ability to provide direct engineering support for the drivers and the list goes on.
2. “Included” FCoE to Lower Networking Costs – Again, Intel and others are trying to make the FCoE market about the acquisition cost of free vs. “included” on FCoE because that is the only place they can perceive an advantage. Emulex is currently shipping “included” 10GbE FCoE with full hardware offload on the latest HP G7 server blades as a LOM, and expect to see more of this innovation for FCoE as we go forward in the market. As I said above, the key in selecting FCoE will not be about the base acquisition cost, but about TCO. It will be about 24/7 support, proven interoperability, standard drivers, proven partnerships and reliability of IT tools and OEM support.
3. CPU vs. CPU Offload – One other point Intel tries to use as a competitive advantage is that with Open FCoE, it can scale with the CPU vs. being limited by the hardware on the 10GbE card. This is exactly what you would expect from Intel: run more on the CPU, and then you have to buy more CPUs. Emulex knows that IT wants to reduce the number of CPUs to lower total capital cost, total OPEX and total management cost. That is what our vEngine technology does for IT managers–our ULOM and UCNAs enable up to 20% more VMs per CPU. We make the entire data center more efficient across servers, networks, storage and virtualization. Intel just gives IT managers another reason to buy more CPUs.
4. Does Emulex Support Open FCoE Today? Will They in the Future? – Like every company, we are driven by our customers’ needs and demands, and we have not been flooded with demand to support this driver on our hardware. No, we do not support Open FCoE on our cards today. Yes, Open FCoE drivers could be run on our hardware. Do we expect to in the future? Maybe. That will be based on released OS support, OEM requirements and input from our IT customer community. As market maturity and customer demand requires, we might adjust our Open FCoE support as needed. We have invested in an industry-leading technology with our “included” FCoE and it has been selected by nine of the 10 top OEMs because our implementation of FCoE is hardware, offers better performance (IOPS, latency and bandwidth), better error handling (fully offloaded from the CPU) and better virtualization support (vNIC, SR-IOV, vCenter Plug-in).
5. Who You Gonna Call? – As they asked in Ghostbusters, “Who you gonna call?” if something goes wrong with an Open FCoE driver. This is why companies like Red Hat exist in the Linux space, even though Linux is “free.” IT managers’ knew partners that they needed partners that provided support and consistent release of drivers that IT can count on and get support for in the middle of the night. Fibre Channel and FCoE live at the core of the data center and IT needs 24/7 support, standard drivers, proven partnerships and reliability of enterprise-class IT management tools from partners that are committed and proven in providing this type of support. Emulex delivers this for our FCoE solutions. All IT managers know that even if solutions are “free” to acquire, they must have the right support model to be used in the enterprise. If they can’t be supported and they don’t come with one throat to choke, as the saying goes, then their adoption will be limited.
6. Does Open FCoE = iSCSI Software Initiators? – We do not think Open FCoE will be a direct repeat of iSCSI software initiators in the market. Emulex’s position is that Open FCoE ultimately will look more like enterprise Linux than iSCSI with the keys to adoption and vendor choice being reliability, acquisition cost, 24/7 support, standardized drivers, proven partnerships and interoperability. Based on what we have seen IT do with other “free” enterprise technologies, even if the acquisition costs are “free” or “included,” they willing to pay for support, interoperability, reliability and consistency in deployment.

In the end, Emulex and Intel will fight it out to be the top players in the 10GbE market, and we think our blending of 10GbE with full hardware IP, iSCSI, FCoE and software management services and tools will place us as the leaders in the market.

Posted in Uncategorized | No Comments

January is always a busy time for the folks at Emulex; we are back for the new year and we have our sales kick-off and our earnings call, all in the first three weeks after we get back from the holidays. This year proved to be no different. However, the news this year was a bit better than last year, because Emulex not only is going strong and gaining share in the new 10Gb Ethernet (10GbE) network convergence markets, we are also gaining share in the Fibre Channel market.

IT Brand Pulse released an Industry brief this week about the ongoing market share battle between Emulex and QLogic in the Fibre Channel market. They said, “In their January, 2010 earnings calls, Emulex and QLogic made claims about their market share positions in Fibre Channel Host Bus Adapters (HBAs). IT Brand Pulse estimates that Emulex is correct: the company gained share over QLogic in each of the last three quarters and for all of 2010. We estimate that QLogic’s assertion is also true: the company remains the leader with 56.3% of market share.” Yes, both statements are true, but there is more to the story.

Last year, every major OEM (server and storage) had to make a 10GbE network convergence choice, and nine out of 10 selected Emulex. Some chose us exclusively, and some qualified multiple vendors, but they all found that the combination of Emulex’s enterprise-class 10GbE, full hardware offload for IP, iSCSI and Fibre Channel over Ethernet (FCoE) and our proven Fibre Channel stack a compelling solution for their new fabric-based computing networks. This new presence in accounts is helping Emulex drive a stronger I/O portfolio story for IT managers, providing a single I/O management framework and supporting the latest virtual Network Interface Card (NIC) technology for HP, IBM, Cisco, Fujitsu, EMC, HDS, NetApp, Dell and more.

According to IT Brand Pulse, “In 2009, Emulex gained the upper hand with the only converged network adapter offering the combination of solid LAN functionality — that serves 90% of the network requirements in the data center — plus iSCSI and FCoE offload. By 2010, IBM and HP integrated Emulex chips and adapters into their virtual networking schemes, and began leading with Emulex-based adapter solutions…the battle to watch is at HP where Emulex starts the year as the incumbent 10GbE LOM vendor for HP servers, and QLogic starts as the dominant Fibre Channel adapter vendor. Emulex will attempt to convince customers to purchase the same converged networking adapters as their LOM chips, while QLogic will attempt to leverage their strength in Fibre Channel adapters. The bottom line is this: Emulex has better NIC functionality, and many IT professionals prefer to use the same brand of LOM chips and NICs because it eliminates existing or potential interoperability problems. If Emulex exploits this advantage, their momentum will continue.”

This is the real I/O battle for 2011, each vendor in this space is seeking to leverage their previous strength, expanded technologies and new OEM wins to drive better solutions for IT managers. From the comments above, we are seeing a strong correlation between our growth in 10GbE and our growth in Fibre Channel, and the share data bears that out.

Posted in Uncategorized | No Comments

I’m on the way back from Interop 2010 in New York City. It was a refreshing show for me, since it was truly out of the storage realm and into the network world. I had a chance to sit down with Howard Marks and talk about network convergence deployment challenges, and Stu Miniman to talk about the leadership race in 10Gb Ethernet (10GbE). I also have to say, the two biggest topics at the show were the cloud and network optimized delivery.

Deploying 10GbE with Howard – One of the things I love about talking to Howard, is he actually plugs this stuff in and tries it; he does more than write about it. He plugs it in and finds out what makes it tick and tock, as the case may be. One of the things he shared was, if you are moving to 10GbE, make sure your CAT-6 cables are a different color than your CAT-5 cables. This simple issue can cause you to spend $10,000 troubleshooting a $10 problem.
Continue reading…

Posted in Uncategorized | No Comments

I was sharing a few tweets with @Stu from Wikibon and @Jonisick about the newly ratified 16Gb Fibre Channel standard. The question was raised, “Great, when will 16Gb be taking over?” Obviously, this is a multi-part answer. The pace of Fibre Channel transitions has definitely slowed down with the shift from 4Gb to 8Gb. From 1Gb to 2Gb and 2Gb to 4Gb, the change took about 24 months, primarily because the cost was the same from one transition to the next. However, with the transition from 4Gb to 8Gb, we saw a non-trivial cost bump for the 8Gb optics. This means that we are over two years out from the first 8Gb shipments (February 2008) and most Host Bus Adapter (HBA) companies have not reach 50% of 8Gb port shipments. According to Seamus Crehan of Crehan Research, we should see the crossover point between 4Gb and 8Gb sometime in the first half of CY2011.

So what does this mean for 16Gb Fibre Channel? First, let’s look at where 16Gb is in the development cycle. We now have an approved standard and some early adopter companies, such as IBM, with whom we announced a 16Gb design win last fall before the standard was finalized.
Continue reading…

Posted in Uncategorized | No Comments

Okay, I really don’t have a tie-in for the Kate Plus 8 thing. I have never watched the show, but I read People Magazine. Sad, but true. However, it gave me a title with “8” in it, so I was happy. Also, I am sure some Fibre Channel customers feel a bit like Kate, jilted by their vendors who have fallen in love with network convergence and Fibre Channel over Ethernet (FCoE). While I have certainly got up on the network convergence soapbox, it does not mean that we have abandoned Fibre Channel. In fact, we continue to make significant investments in Fibre Channel, including our new 8Gb encryption Host Bus Adapters (HBAs), OneCommand Vision software for I/O management and our new 16Gb products coming to a Storage Area Network (SAN) near you next year.

In October, we announced the industry’s first 16Gb Fibre Channel HBA design win with IBM. While it won’t ship until CY2011, it shows our long-term commitment to Fibre Channel technology and innovation. Key new technologies such as PCI Express 3.0, Encryption, SR-IOV and MR-IOV will continue to drive Fibre Channel forward for years to come. Emulex is leading the market with solutions in these spaces and working with our ecosystem partners to deliver these solutions.
Continue reading…

Posted in Uncategorized | No Comments

“To offload or not to offload?” That is the question that many 10Gb Ethernet (10GbE) Network Interface Card (NIC) vendors are asking IT managers to answer.

So here is the story…Intel and other NIC-only vendors are asking IT managers to run software initiators for iSCSI and Fibre Channel over Ethernet (FCoE). Their claim is that with all of the new CPU cores on next-generation servers, dedicating a few CPU cores to I/O is not a big deal and they will even tell you that it “makes sense.” However, let’s take a step back for a minute. Before server virtualization, I might have gone along with software initiators because a single application did not max out the CPU cores, so it did not matter. Then server virtualization came along and we started running three to four virtual machines (VMs) per CPU cores and this is when things began to change. Now, according to a presentation delivered by IDC in November 2008, we are moving to 10 to 12 VMs per CPU core. What is driving all of this? Data center consolidation.

The goal of maximizing the number of VMs has become a top priority for IT managers in their consolidation efforts and using CPU cores for I/O seems to be at odds with their key goals.
Continue reading…

Posted in Uncategorized | No Comments

So, you have heard all of the vendors in this business pontificate about the benefits or detriments of network convergence, including yours truly. But today, I’m glad to talk about a report from the Taneja Group, where they actually spoke to some people who have done some early deployments of network convergence because…it just made sense. In this paper, we have the typical preamble of the problems network convergence solves, why IT should deploy it and all of the other things you would expect to hear. However, unlike most of the other papers shipped to us vendor types to date, this one has some real customer input. What you will hear a few recurring themes from the customers in their stories: less stuff saves money (CAPEX, OPEX), and less stuff means lower power and fewer cables. This is not rocket science; it is common sense. Ben Franklin would be proud.

As they said on Dragnet, the stories you are about to see are real, but the names have been changed to protect the innocent. As many of you know, legal departments hate to let their IT folks do testimonials, so we have described them, not named them. That being said, here are the profiles of the customers:

1. Major Server Vendor – This vendor has revenues over $50 billion and “data centers the size of football fields”, and they chose to deploy network convergence to reduce power and cooling with less cards and switches and to reduce cable chaos.
2. A Major Retail Organization – The retail organization serves the U.S. armed forces and their families across the globe. They deployed network convergence to get better security, and yes they still kept their Fibre Channel systems for storage. But they wanted a path toward convergence over the next five years as they begin to move to 10Gb Ethernet (10GbE). However, the overriding concern was that they picked a solution that was open and did not lock them into a single vendor.
3. A Small Regional Hospital – They are located in a major ski resort area and see significant spikes in demand for services during the winter sports season. They recently reduced 60 physical servers into 11 physical servers running ESX and network convergence. Given the file-oriented nature of their business, they need better network-attached storage (NAS) performance and want to buy less stuff to get it. A network convergence solution was just what the doctor ordered…sorry, I could not resist.

Continue reading…

Posted in Uncategorized | No Comments

Other than the Olympic Games, no other event brings the world together like the World Cup. It is a connection that spans almost every demographic. It crosses race, gender, politics and age groups, and connects all of us, even us Americans who are now getting a clue a about soccer…er, football. Okay, we are still working on the name thing. Right after I had this thought, I was on the Nike site thinking about ordering a U.S. team shirt when I saw this one and I had to order it. I laughed at first, but it showed how we are connecting with the rest of world and learning to merge and converge.

It is on the theme of merging and converging that I wanted to focus for this blog. In some ways, this reminds me of the management groups of IP and storage coming together. They have heard of each other, they share a data center, but they really don’t know the rules of each other’s games and they use the same words that mean very different things to one another. Each of these teams has built a specialized set of skills for playing their respective games and many overlap, but they are still very different games.
Continue reading…

Posted in Uncategorized | No Comments

Two of the key questions facing IT managers are whether or when to converge. We have posted many blogs on the pending move to network convergence and have been asked by many people, “Why do you think we should move to convergence?”, “Prove to me that it works” and “Show me the money (ROI and TCO)”. The bottom line being…why does this make sense? We posed that question to ESG’s Bob Laliberte and asked him to write up his thoughts on the subject, and he developed the following paper, “Network Convergence in the Data Center Makes Sense.”

In thinking about this question, Bob started by pointing out what the top IT priorities were for 2010. The table below shows what IT is thinking about and provides a basis for why network convergence in the data center makes sense. Look at the top items: “Increase use of server virtualization,” “Data center consolidation” and “Upgrade network infrastructure.” The move to network convergence can help them address three of the top four priorities with increased used of virtual I/O solutions, such as Cisco UCS, HP Flex-10/Virtual Connect and IBM’s Virtual Fabric Solution, which will upgrade their networking infrastructure to 10Gb Ethernet (10GbE), increase the flexibility of host connectivity with virtual network interface cards (NICs) and Universal Converged Network Adapters (UCNAs), and the list goes on. Continue reading…

Posted in Uncategorized | No Comments

As I write this blog, it is the day after Cinco de Mayo, and I spent many hours in front of a blender, thinking about Jimmy Buffet and how I can change my latitude, when Steve Duplessie came to my mind. I am sure he will be as disturbed as I was at that revelation. However, the point of this somewhat foggy flashback is the term he coined, “The I/O Blender,” referring to what happens when you take a bunch of virtual machines (VMs) with varying levels of I/O demands and dump them onto a single physical server, swirl them together and hope that it all comes out smooth and delicious, versus a clumpy science project.

Avoiding this kind of mixology is exactly what Emulex’s OneCommand Vision was designed to help you avoid. Vision is a unique product with some unique origins. Unlike most products that are designed to fill a market need, this started out in life as a testing tool for our Design Verification Test (DVT) engineers. When server virtualization came along, we needed a way to see what individual performance of each VM was while testing N_Port ID Virtualization (NPIV). Then people asked if it could be updated to test latency between the VM and adapter, and then the VM and switches, and then VM and target, and you get the idea. As we were working with OEMs and ecosystem partners, we used the tools to help optimize performance and find issues, and before you know it, the proverbial star was born, and it became the OneCommand Vision product. Continue reading…

Posted in Uncategorized | No Comments