CEO Blog

I'd like to make you aware of IDC's recently published research: Virtualization 2.0: Opportunities and Challenges for Next Generation Datacenter Networks.  Among other important guidance, IDC's spotlight illustrates how enterprises using virtualization technology to improve business operations and mobility can leverage the capabilities provided by widespread adoption of 10 Gigabit Ethernet as a single, converged network fabric.

Perhaps the greatest challenge that organizations will face when moving to what IDC calls "Virtualization 2.0" will be networking related. As IDC's John Humphreys writes, "With the initial uses of virtualization technology in which multiple images are consolidated into a single server, Virtualization 2.0 will require the consolidation of network traffic and will also increase the need for more bandwidth to the server, both of which will be possible as enterprises make the move to converge and consolidate data, storage and inter-process traffic on 10Gb Ethernet networks."

What IDC calls the "Virtualization Effect" is accelerating the enterprise migration to 10Gb Ethernet. And as Humphreys points out, with virtual machines moving from host-to-host in Virtualization 2.0, it is critical that the network identity, quality of service and security policies associated with virtual machine move along with the VMs, IDC observes that a virtual machine aware and virtual mobility ready network requires a highly flexible and open 10Gb Ethernet network topology.

And this is what BLADE is working to refine and deliver with coming enhancements to our network virtualization software offerings. As I mentioned in my previous blog posting, BLADE is aggressively working on advanced network virtualization solutions to equip data center networks for virtual machine environments. Our advances in open network virtualization, combined with the higher speeds, lower latency and emerging standards-based lossless capabilities of 10Gb Ethernet, will enable organizations to create converged, single-fabric, multi-connection networks that offer the flexibility, speed and expandability required by virtual machine environments.

Unlike Cisco's propriety VN-Link for the Nexus 1000V, BLADE's solution will work and interoperate with all major virtualization platforms without forcing any software upgrades or firmware updates. BLADE is working to equip our customers' 10 Gigabit Ethernet network infrastructures to ensure that the "Virtualization Effect"  our customers experience in their data centers does indeed deliver the requisite combination of high performance, low latency, and virtual machine awareness they require to fully harness the tremendous potential of Virtualization 2.0.

If you would like to dive deeper into Virtualization 2.0, send an email to bladenews@bladenetwork.net to subscribe to our informational newsletter, BLADENews' February 2009 issue will focus on this topic.

Labels: Rackonomics RackSwitch BLADE Network Technologies Vikram Mehta

 

 

I particularly savor it when our friends on the analyst and editorial sides of the industry agree with us at BLADE that advanced data center technologies such as blade servers and10 Gigabit Ethernet are ready to bring the high productivity of High Performance Computing (HPC) to the mainstream enterprise.

Debra Goldfarb and Michael Feldman at Tabor Research and its media outlet HPCwire.com, are spot on in the realization that what we know as high performance computing on Wall Street, science and academia is coming to Main Street, will be known as High Productivity Computing. That stands to reason, especially in today’s frugal economy.

As Michael astutely observed a while back , the lines between traditional HPC and Main Street computing are increasingly blurred. Today, server OEMs can build commodity-based blade server and rack-level systems with all the characteristics associated with HPC and sell their wares into a vast range of markets -- financial, entertainment, oil & gas, biotechnology, manufacturing, and others. The same holds true for HPC storage and network component vendors, such as BLADE’s own bladed and top-of-rack data center-class switches.

We BLADErs agree with Debra and Michael that HPC coming to Main Street sounds like good news -- and it is even if it creates some initial confusion in the industry. We applaud their evangelism of the term “High Productivity Computing� instead of “High Performance Computing� to describe the new Main Street HPC.

Debra correctly asserts that due to the democratization of HPC, the market is rapidly maturing. Instead of being focused on flops or even flops/dollar, users -- practical creatures that they are -- are demanding return on investment (ROI). So rather than fixating only on the server hardware, the environment must be looked at from what a holistic perspective that includes the network and storage interconnects, the storage hardware, servers, applications et al.

This view of High Productivity Computing for Main Street is dependent upon the balanced integration of all system and subsystem-level components. And, as Debra rightfully concludes, this notion of productivity is aligned with ROI, which is what BLADE’s approach to Rackonomics is all about, and which you can read about in detail in my earlier postings.

While the conventional HPC industry will still look to the Top500 list for the fastest computers on the planet, Main Street is looking to get their HPC work done and is simply interested in getting the most computing bang for the least amount of bucks. And that’s what BLADE and our partners such as HP, IBM, NEC and now Verari Systems, are all about.

We’d be delighted to have you visit BLADE at Supercomputing '08, booth 322, 17-20 November, in Austin, Texas. BLADE’s VP of Product Strategy and Management, Dan Tuchler, will be giving a talk on “10G Ethernet – Ready for Major Breakthrough in HPC Environments in 2009,� on Wednesday, 18 November at 10:30am.

Labels: Rackonomics RackSwitch BLADE Network Technologies Vikram Mehta

 

 

It has been said that in today’s data center, virtualization changes everything. In fact, it’s how virtualization actually changes every IT “thing� that has data center mangers perplexed. The core proposition of virtualization breaks what has historically been inviolable bonds, such as the relationship between operating systems and their underlying hardware platforms. And whenever data centers shatter such enduring ties, new and sometimes unanticipated “things� happen.

These “things� are becoming increasingly pervasive in today’s data centers. In fact, IDC analyst John Humphreys observes in “The Future of Virtualization: Leveraging Mobility to Move Beyond Consolidation� , that recent IDC surveys found that over 50% of all customers are employing server virtualization for production applications, including mission-critical applications such as supply chain management and enterprise resource planning. Within the next 12 months, the survey found that these same users expect nearly half of their applications will be hosted on a virtualized server.

For example, with the introduction of virtual machine mobility through exciting new technologies such as VMware’s VMotion comes the promise of great “things.� These include greater server utilization and flexible, no-downtime migration, along with new opportunities for data center consolidation, flash crowd control and virtual disaster recovery. However, VMotion also brings other “things,� such as a feeling of uncertainty surrounding the boundaries of VLANs, the inadequacy of network security policies for virtual machine environments, and Quality of Service for network access and application performance.

The underpinning of data center virtualization is a move away from managing devices and toward managing information. The virtual environment promises to free IT architects to focus less on hardware and more on who accesses what information and how they gain this access. However, it’s in understanding all those changes to every IT “thing,� including the underlying network, that is today’s hidden secret to successfully managing virtualization on an enterprise scale.

Yes, virtualization does change everything. But the hidden secret to successfully managing virtualization is revealed from discovering all the unanticipated “things� that come with virtualization. And, once you understand how the hidden secret can impact and even jeopardize your enterprise and its vital data, it then becomes much easier to establish IT practices for managing how virtualization actually changes every IT “thing� from the server, to the OS to the underlying network.

BLADE is working on advanced network virtualization solutions to equip data center networks for VMs. Unlike Cisco’s VN-Link for the Nexus 1000V, BLADE’s solution will work and interoperate with all major virtualization platforms – without forcing any software upgrades or firmware updates. As a VMware Technology Alliance Partner, BLADE will be delivering additional enhancements to improve our joint customers' virtualization experience in the critical areas of networking and virtual machine migration. Look for more on this to come.

Labels: Rackonomics RackSwitch BLADE Network Technologies Vikram Mehta

 

 

Today’s trends of convergence, converged networking and cloud computing mean different things to different people. To consumers, convergence means unified billing and customer service for combinations of phone, Internet access and TV on a single wire. Converged networking implies that these services come on a single wire. And, cloud computing alludes to Google Apps, Web 2.0 and VideoOnDemand.

To enterprises, convergence signifies the vital importance of always-on, high-speed access to secure telecommunications networks. In fact, enterprise IT departments name their number-one priority when building a new data center as access to telecommunications networks.

To the enterprise, converged networking points to the promise of delivering data, storage and inter-process communications across a single unified fabric – a single wire for all enterprise applications and services, including VoIP, enterprise apps, VideoOnDemand and secure access to enterprise data. Today, convergence on converged data center networks has become viable primarily due to the emergence of affordable 10 Gigabit Ethernet on mainstream servers and affordably priced blade and top-of-rack 10Gb switches, such as BLADE’s new RackSwitch and embedded blade server switches for HP, IBM and NEC.

Cloud computing, which Forrester Research defines as a “pool of abstracted, highly scalable and managed compute infrastructure capable of hosting end-customer applications and billed by consumption,� promises exciting new applications such as 3D technology that allows you to have conference calls with others and place them on the left right, front or rear. Until you have heard a conference call in 3D stereo, you have not heard a conference call. In addition, the technology allows the addition of overtones like adding a metallic sound to a speaker or group of speakers on a conference call. This is great if you are looking to find a way to discern what group a person belongs to -- the London office for example.

However, there are problems in the cloud. James Staten of Forrester Research notes that growing pains will plague cloud computing in its early stages. Staten projects that outages lengths will decrease from the eight-hour incident that took down Amazon’s S3 cloud storage service and the congestion that plagued Apple’s MobileMe service across several days beginning on July 18, ostensibly because of the failure of an email server. For this recent news about cloud computing outages, see Systems Management News With outages of hours and days – enterprise acceptance of cloud computing will clearly require the most reliable, available and high-bandwidth access to telecommunications networks underpinned by an exceptionally robust data center infrastructure.

To BLADE and its customers – literally half of the Fortune 500, convergence, converged networking and cloud computing are practical matters. Bottom line, data center networks must work right every time and all the time, around the clock and around the globe. The wave of the future is convergence on 10 Gigabit Ethernet networks, converged single-fabric networks using iSCSI or the rapidly emerging Fibre Channel over Ethernet (FCoE) technologies, and ever more productivity-enhancing and interactive applications and services in the cloud. And that wave is driving the Rackonomics imperative for non-stop, high-performance, low-latency and energy-efficient data center networks that can be scale out and replicated on a massive scale. To the enterprise, that’s what convergence, cloud computing and converged networking are all about.

NetEvents.tv has posted a new documentary about the future of convergence, converged networking and cloud computing entitled “100 Gig Ethernet - Why, How and When?� in which I am interviewed, along with other industry executives. The documentary is hosted by journalist, Manek Dubash. To view the video, go to:

Labels: Rackonomics RackSwitch BLADE Network Technologies Vikram Mehta

 

 

Today, what IDC calls the “Data Center of the Future,� is ramping up to full throttle with 10 Gigabit Ethernet, the unifying network standard. Data centers equipped with 10 Gigabit Ethernet can enjoy a range of compelling benefits including investment protection, product flexibility, performance, and energy efficiency backed by a deep, multi-vendor ecosystem to deliver 10Gb NICs, interconnects, blade and rack-level switches, as well as conventional fixed and chassis-based switches. What’s more, one of the issues that until now has impeded the adoption of 10Gb was price, which is now becoming less of an issue as per port prices have declined below $500 per port, the price point at which the volume economics of Ethernet are expected to accelerate the move to 10Gb NICs, edge and aggregation switches.

BLADE and IDC recently teamed up with HP and IBM to air two webinars: “10G Ethernet: Overcoming Network Overload with Rackonomics—An Innovative Approach to Scaling Out Data Center Networks� and “10G Ethernet: The Future of Scalable Networks for Blade Server Virtualization." These webinars offer insights into factors that are introducing new loads and requirements for low latency to the network. For example, IDC’s Cindy Borovick explains the impact on the network from “Dynamic Datacenters� that are equipped to manage workloads based on business priorities, provide virtual I/O and virtual network services and deliver SOA-based applications.

Cindy’s “Essential Guidance� is fivefold:
1. The “Datacenter of the Future“ will require performance, availability and scale.
2. Virtualization changes network requirements.
3. Asset consolidation drives the need for high bandwidth and low latency.
4. Ethernet is on a path to be the unifying network standard in the datacenter.
5. 10 Gigabit Ethernet will be the foundation of the future datacenter network.

If you’re like me, you’ll come away from these webinars as time well spent with some fresh insights into why 10 Gigabit Ethernet is the best choice for the dynamic and highly virtualized “Data Centers of the Future.�

Once again Ethernet is the de facto mainstream network topology with 10 Gigabit Ethernet today’s best choice for massive scale-out enterprise networks, emerging converged networking/storage networks and virtually everywhere and anywhere a server or bank of servers requires high bandwidth, low latency, advanced energy efficiency and the affordability of Ethernet’s volume economics. Even High Performance Computing (HPC) clusters, which have been the domain of Infiniband and other proprietary server I/O interconnects, show growing use of Ethernet – see the TOP500 List of the world’s most powerful supercomputers.

The technology blogosphere is paying attention to the coming widespread migration to 10 Gigabit Ethernet as the mainstream standard. I encourage you to visit Stephen Brown’s informative blog about all things 10GbE.


Also, it’s worthwhile to read what Chris Mellor for Blocks and Files in the U.K has to say in his view about BLADE’s groundbreaking solutions for implementing the lossless 10Gb Ethernet required for converged Fibre Channel over Ethernet (FCoE) storage networks. “No need to wait,� writes Mellor, because BLADE can implement a loss-less, low-latency FC0E network now.�

At BLADE, our embedded 1Gb/10Gb Ethernet switches and new 10Gb top-of-rack data center-class RackSwitch are the flagships of our product line and the mainstays of our Rackonomics concept for provisioning massive-scale out networks at the rack level. BLADE 10Gb switches are being deployed across our customer base in more than two dozen market segments, including finance, automotive, defense, and academia to name a few.

For example, a leading academic computer center one of the largest supercomputing and networking centers in Eastern Europe, recently installed BLADE’s 10 Gigabit Ethernet switches an HPC solution using IBM BladeCenter. After evaluating other vendors and fabrics such as Infiniband and Myrinet, the computer center chose BLADE’s 10GB solution primarily because Ethernet is a well-known technology that is easier and more affordable to deploy, configure, install and manage. Using 10Gb Ethernet interconnects within the IBM BladeCenter chassis also delivers significant price-performance advantages along with high bandwidth and low latency. You can download BLADE’s Solution Brief about 10 Gigabit Ethernet High Performance Computing Clusters on IBM BladeCenter at

Just as 10 Gigabit Ethernet is now growing through widespread deployment in the data center, the discussion has not shifted to consider tomorrow’s even higher-speed interconnects – namely 40Gb and 100Gb Ethernet. Coming in September 2008 to a Web browser near you, I will be appearing on NetEvents.tv to speak about the future of Ethernet – specifically “100Gb Ethernet - why, how and when?� Stay tuned to NetEvents.tv’s Telco Channel. For now, hold on to your hats because 10 Gigabit Ethernet is ready to help your dynamic data center ramp up to full throttle.

Labels: 10G Ethernet, 10Gb Ethernet, data center, Rackonomics RackSwitch BLADE Network Technologies Vikram Mehta

 

 

The data center of the future will comprise replicated racks, rows and even containers with massive processing, I/O and storage capabilities.

We’ve coined the term Rackonomics to describe this datacenter of the future that comprises replicated racks of servers, storage and network elements, organized into rows and even containers.

Rackonomics can drastically reduce facilities and support costs and drive down total cost of ownership. It can also improve energy efficiency through high-density packaging and unique cooling designs that do not require the construction of a brand new data center facility. It is the way of the future.

Google pioneered the container data center concept with the intent to distribute Google data centers to every network peering point in self-contained shipping containers. Today, massive scale-out containers can house thousands of processors, tens of petabytes of disk storage, and hundreds of Gigabits of I/O, and can be dropped-off overnight by a tractor-trailer rig. Google’s idea is to place one of these containers anywhere Google owns access to fiber, primarily at one of the 300 major Internet peering points, in effect transforming the Internet into a massive processing and storage grid.

At BLADE, we are designing and building the top-of-rack and blade-based data center-class networking products that make the ideal, standardized networking components for the rack, row, and container-based data center. And we’re partnering with companies like IBM, whose iDataPlex platform is amongst the most innovative solutions for building out the Rackonomic datacenter.

Rackonomic data centers have tremendous potential, for example, enabling Internet TV to scale to the same level as broadcast and cable TV with latency and system response as low as they can possibly be made for a networked application using Ethernet as its essential network backbone. We’re proud to be playing a leading role in an industry-wide effort to redefine how enterprises build out their datacenters.

Labels: Rackonomics RackSwitch BLADE Network Technologies Vikram Mehta

 

 

As we have been briefing various customers and industry analysts in conjunction with the recent launch of our new top-of-rack data center-class RackSwitch, it has become clear to me that there is a tremendous amount of confusion in the marketplace. This confusion seems to stem from the relative absence of holistic best practices surrounding how data centers can harness the tremendous benefits of serve/storage consolidation and virtualization without inadvertently driving hidden costs and pitfalls into the networking facilities infrastructure. Power, for example, is a huge issue, when companies
using collocation facilities can't supply their racks with more than a small KwA per rack, so end up deploying racks that are only half or one third full.

This is why I'm so passionate about the potential for "Rackonomics" – a new set of holistic rules for the data center that BLADE is defining and championing. IT departments can leverage Rackonomics to scale out their data center networks affordably and holistically at the rack level. In modern data center architecture, Rackonomics refers to the concept of rack-level provisioning--designing, deploying and replicating server/computer systems, data networks and storage area networks (SANs) rack by rack to:

• decrease the total cost of ownership of data center infrastructures
   
• reduce IT complexity
   
• enable incremental scalability.

One of the primary tenets of Rackonomics is that putting switches into blade enclosures and server/storage racks is far more economical than using external core switches. Our customers have told us that a "put it in the rack" strategy can enable them to save tens of thousands of dollars in switch hardware, deployment and energy costs for every external switch that they can avoid deploying. And, the network switch has the ability to viritualize the network connections associated with the server/storage elements located in the rack, thus vastly simplifying the datacenter network.

What's more, every time an IT manager can make a rack-level deployment decision, they can fully understand the holistic impact of that rack on other elements of their data center infrastructure. Through Rackonomics and its rack-level approach, as requirements to grow and scale out emerge, an IT manager can understand exactly the impact of what's being added in terms of compute capacity, latency and power and cooling - without over-provisioning the expensive core network, SANs, and power and cooling capabilities to keep pace.

The fundamental concept of Rackonomics is straightforward. By linking servers/storage into localized switches that reside inside blade and/or rack enclosures, IT buyers can save money by dramatically reducing acquisition costs and eliminate network complexity by virtualizing network resources at a rack level. Operational costs can also be reduced because IT administrators are managing fewer entities (racks instead of individual elements in a rack). Furthermore, cabling costs are reduced often by as much as 80%) and cooling requirements for network equipment can be cut in half because fewer power and space hungry core switches are required.

However, as the saying goes, "the proof is in the pudding." One of the biggest benefits of Rackonomics comes from the real-world understanding of how all the elements of the rack work together in a holistic way before making the big decision to scale out using a specific rack full of components. Rackonomics offers a new set of customer-driven rules of the data center. But, how can vendors and their customers play together by those rules?

At BLADE, we're committed to overcoming confusion about how best to implement server/storage consolidation and virtualization by helping to establish shared understandings between the vendor and user communities around holistic data center best practices. Our recent demonstration in Orlando, Florida at Storage Networking World of Fibre Channel over Ethernet (FCoE) with Emulex and NetApp is one such holistic example. FCoE, with its ability to converge data and storage networks, is heralded as the wave of the future, but only to the extent that vendors can communicate its viability and value and the end-user community can understand how to make it work in the real world. That is Rackonomics at work.

One final thought for the busy IT manager grappling with a myriad of issues….think of your datacenter as a collection of racks vs. individual server, storage and networking elements. Make a rack as the smallest unit of deployment in your environment and then replicate that rack. This will make deployment, management and scaling much simpler.

Resource Links:

• Virtualizing the Network Facilities (Interop 2008 Podcast I did with InfoWorld)
   
• ZDNet Blog entry on BLADE and Rackonomics from a conversation with BLADE VP of Strategy and Product Management, Dan Tuchler

Labels: Rackonomics RackSwitch BLADE Network Technologies Vikram Mehta