Server virtualization meets grid computing

GRIDtoday published an interesting interview with Kate Keahey, an Argonne National Laboratory scientist working on the Globus Toolkit and other aspects of Grid technology, about how server virtualization can serve distributed computing purposes:


Gt: Virtualization and distributed computing seem to permeate everything in IT today. Tell us about some of the ways virtualization is converging with distributing computing and how Grid technology fits in.

KEAHEY: I think of virtualization as a vehicle to realize the dream of Grid computing — obtaining on-demand computational resources from distributed sources in the same simple and intuitive way we get electricity today. Today, in order to run a job on the grid a user has to identify a set of platforms capable of running that job by virtue of having the right installation of operating system, libraries, tools, and the right configuration of environment variables, etc. In practice, this means that the choice of platforms will either be limited to a very narrow set, or the job first has to be made compatible with an environment supported by a large resource provider, such as TeraGrid. For some applications this is a significant hurdle. Furthermore, even if you do manage to identify such an environment, it is hard to guarantee that the resource will be available when needed, for as long as needed, and that the user will gets his or her fair share of that resource.

Virtualization introduces a layer of abstraction that turns the question around from “let’s see what resources are available and figure out if we can adapt our problem to use them” to “here is an environment I need to solve my problem — I want to have it deployed on the grid as described.” For a user this is a much simpler question. The issue is whether we can implement the middleware that will map such virtual workspace onto physical resources. One way to implement it would be to provide an automated environment installation on a remote node.

But what really gives this idea a boost is using virtual machine technology to represent such a workspace. This makes the environment easy to describe (you just install it), easy to transport, fast to deploy and, thanks to recent research, very efficient. Best of all, virtual machine management tools nowadays allow you to enforce the resource quantum assigned to a specific virtual machine very accurately — so you could for example test or demo your application in a virtual cluster making sparing use of resources, and redeploy the virtual cluster on a much more powerful resource for production runs. This is another powerful idea behind virtualization: the environment is no longer permanently tied to a specific amount of resource but rather this resource quantum can be adjusted on-demand.

Similarly, we can define virtual storage and implemented using distributed storage facilities, or overlay networks implemented on top of networking infrastructure. We can compose those constructs to put together whole “virtual grids” and test their operation before requesting serious resource allocations. There are many exciting ongoing research efforts in this area and some of them will be represented at the VTDC workshop.

Further down the road, if the idea of running virtual machines becomes ubiquitous, we may find other ways of leveraging the fact that we can have more than one isolated “hardware device” on a physical resource. We could use it to host physical devices requiring isolation for security reasons. We could carry around pluggable virtualized environments the way we carry laptops today. We could rely on migration to a greater extent to provide uninterrupted services. All those potential applications will come more clearly in focus once we see how widespread the appeal of virtual machines will prove in practice….

Read the whole interview at source.

I already covered this topic in January 2006, with my old Virtualization is the first step of a long walk called Grid Computing.

Opsware announces Virtualization Director

Quoting from the Opsware official announcement:

Opsware Inc., the leading provider of Data Center Automation software, today announced the introduction of Opsware Virtualization Director, the industry?s first integrated solution that seamlessly manages the complete lifecycle of both physical and virtual servers to enable management on an enterprise scale.

Opsware Virtualization Director provides comprehensive virtual server lifecycle management to accelerate enterprise deployments by seamlessly managing across physical servers and environments virtualized with technologies such as VMware, Microsoft Virtual Server, Sun Solaris 10, XenSource and others. Initially, Opsware Virtualization Director will support VMware and Solaris 10; support for other environments is expected to be added in the future. Opsware Virtualization Director?s integrated capabilities include:

  • Creation
    provision, activate and delete virtual servers on a large scale basis
  • Control
    start, stop, suspend and resume virtual servers as required
  • Visualization
    discover and view dependencies between applications, servers, network devices and storage
  • Tracking
    track relationships between virtual machines and their hosts and the applications they support; perform change impact assessments between applications, servers, network devices and storage
  • Standardized management
    seamless management across virtual and physical servers and networks encompassing the entire lifecycle, including patching, configuration, software deployment, audit and compliance, reporting, and best practices and policy-enforcement for the entire environment.

The Opsware Virtualization Director will be generally available as part of Opsware System 6 in the first quarter of calendar 2007…

Scalent too may be soon involved in Virtualization Director evolutions since the company entered the Opsware Technology Alliance Partner Program as charter member.

The virtualization.info Virtualization Industry Roadmap has been updated accordingly.

XenSource launches ISVs / IHVs Partner Program

After launching a Channel Partner Program in August, XenSource is now working on alliances with software and hardware vendors worldwide.

Quoting from the XenSource official announcement:

XenSource, Inc., the leader in infrastructure virtualization solutions based on the open source Xen? hypervisor, today introduced a partner program that will leverage relationships with independent software vendors (ISVs) and independent hardware vendors (IHVs) to develop high performance products for virtualized environments based on Xen technology.

The new program will offer companies three level of participation, each providing a level of integration that best suits the member company:

  • Technology Partners
    for companies interested in increasing awareness of the virtualization support provided by their products through the XenSource website partner pages, which will highlight solutions by category.
  • Premier Partners
    for companies seeking a more integrated marketing strategy that includes joint PR and branding; shared collateral such as white papers, case studies, and solution descriptions; and participation at XenSource user conferences. Additionally, XenSource will invest in co-marketing initiatives including channel partnerships and interoperability and compliance testing.
  • Strategic Partners
    for companies committed to XenSource technology, this level of partnership is by invitation only and includes long-range, fundamental engineering interactions; development and optimization activities; and co-marketing initiatives including channel tie-ins.

All partners will have early access to technical specifications, and participate in the XenEnterprise Beta programs. There is no fee for charter members who join by December 1, 2006. For more information or to apply for the program, go to www.xensource.com/partners

Charter members include:

  • IHVs
    AMD, Azul Systems, Brocade, Emulex, Intel, Isilon Systems, QLogic, SuperMicro, Tehuti Networks and Verari Systems
  • ISVs
    Availigent, Avinti, Avocent, Cassatt, Citrix, CohesiveFT, ConVirt, Embotics, Enigmatec, Enomaly, Evident Software, Hyperic, Ingres, Klir Technologies, LeoStream, Marathon Technologies, Microsoft Corp., Mountain View Data, PlateSpin, Platform, Propero, Provment, Qlusters, rPath, SignaCert, Sphera, SteelEye Technology, ToutVirtual, Transitive, uXcomm, Verari Systems Software, Virtual Appliances, VMLogix, Voltaire, and Zmanda

Several of these charter members announced efforts to support XenEnterprise in their current products. Among them:

Emulex announces HBA management software for Microsoft Virtual Server 2005

Quoting from the Emulex official announcement:

Emulex Corporation today announced VMPilot, a new management software application for Microsoft Virtual Server 2005 environments. The Emulex VMPilot is designed for simplified creation and migration of virtual machines attached to storage area networks (SANs). Employing Emulex’s 4Gb/s Fibre Channel host bus adapters (HBAs) with its industry-leading LightPulse Virtual HBA technology, VMPilot is designed to allow administrators to seamlessly deploy and migrate SAN-based virtual machines, while leveraging their existing SAN management tools and best practices.

A simple wizard creates virtual machines with SAN connectivity using virtual HBA ports that work like physical HBAs.

VMPilot also easily migrates virtual machines to alternate physical servers while maintaining their SAN attachment, saving storage managers from the time and effort of reconfiguring storage and fabrics or copying files. In addition, VMPilot allows storage managers to isolate applications and comply with service level agreements. VMPilot is in beta release and is scheduled to be available during the first quarter of 2007…

LynuxWorks unveils LynxSecure Separation Kernel on Intel VT

Quoting from the LynuxWorks official announcement:

Raising the bar for embedded software security and safety once again, LynuxWorks, Inc., today announced LynuxWorks’ LynxSecure, the first embedded separation kernel with multiple independent levels of security (MILS), on Intel’s Virtualization Technology.

LynxSecure supports Symmetric MultiProcessing (SMP) and 32-bit/64-bit addressing for high-end scalability. LynxSecure will be certifiable to both Common Criteria EAL 7 and DO-178B level A, and its separation technology puts it in a unique place to attain these certifications…

Whitepaper: Altiris SVS 2.0 Technology Audit

The Butler Group research firm released a 7-pages technical analysis paper about Altiris Software Virtualization Solution (SVS) 2.0, providing this conclusion:

Altiris SVS is a well developed solution to combating ‘DLL hell’ in the Windows environment, however, it also provides an excellent tool for ensuring that desktop management is optimised and support costs minimised. Offering organisations a different approach to managing applications on the desktop, Altiris SVS represents a significantly different perspective to that offered by hardware virtualisation solutions. Application virtualisation lacks the resource balancing capabilities offered with hardware virtualisation, but Altiris has worked closely with VMWare so its SVS product operates in a physical or virtual hardware environment where it benefits from the resource balancing capabilities offered by products such as VMWare Infrastructure 3. This capability makes Altiris SVS a versatile solution that provides organisations with management tools to perform desktop operating system migrations and desktop application management simpler and more cost effective.

Butler Group considers that Altiris SVS meets the requirements of managing the desktop applications extremely well; we particularly like the simple filter driver approach to virtualising the applications as this removes the need for organisations to change the infrastructure to accommodate the move to a virtual world. Butler Group believes the weaknesses in the product, especially the lack of support for thin client, server applications, and operating system patch management, are all of secondary importance. Altiris has stated that many of these will be supported by the next major release of SVS scheduled for mid 2007.

It’s very short and seems more a small review (excluding title page, summary page and contact page it’s just a 4-pages brief overview) than a technical assessment. But Altiris claims this is an independent analysis so someone may want to take a look.

Download it here.

VMworld 2006 round-up

VMworld 2006 has been the biggest edition so far: VMware welcomed in Los Angeles Convention Center nearly 7000 attendees.

virtualization.info was present by gratious invitation and had opportunity to meet industry key figures from VMware, Microsoft, XenSource, Virtual Iron, SWsoft, PlateSpin, vizioncore, Surgient, VMLogix, Double Take, EqualLogic, Zeus Technology, IBM, HP and many others.

Below you’ll find a short report highlighting VMworld 2006 main points of interest.

The explicit message of this year has been: software as a service (SaaS).

VMware launched the Virtual Appliances Certification Program and the Virtual Appliances Marketplace (VAM) where ISVs can sell their virtual appliances, offering customers a reliable and controlled environment for their applications.

The entire VMware top management spend words endorsing this approach and several sessions have been focused on how to manage a datacenter populated with virtual appliances.

Even company products are going to reflect such effort: Workstation 6 will introduce for example a new Virtual Appliance View to let ISVs describe environment features at every system boot.

With virtual appliances VMware is trying to approach the software as a service (SaaS) philosophy in a different way than application virtualization: porting (and in a near future streaming, I bet) a whole environment instead of just applications.

This approach may be more successful than application virtualization since it doesn’t care about underlying operating system, allowing companies to use any kind of product, regardless of which platform the software has been written for. But for sure it is more resouces demanding and raises a whole new set of maintainance issues.

The implicit message has been: focus on Enterprises.

VMworld 2006 demonstrated current company little attention for SMBs:

  • few sessions were about Server and no mentions have been done about future of this product
  • upcoming Lab Manager lost support for hosted platforms (both VMware Server and Microsoft Virtual Server)
  • upcoming VMmark has some hardware requirements that several small realities will not be able to satisfy

Only Virtual Appliances initiatives, Marketplace and Certification Program, have been a good news for some small companies.

VMware is definitively spending a lot of its resources for enterprise-grade products and features and SMB could feel forgotten, but Raghy Raghuram, Vice President for Datacenter and Desktop Platform Products, assured new efforts in SMB segment for the coming year.

The overall theme has been competition.

The VMworld exhibitors floor showed booths from 89 exhibitors, plus an amount of other virtualization vendors representatives among crowd.

  • Microsoft, XenSource, Virtual Iron and SWsoft were all presents competing with VMware on plaftorm offering.
  • PlateSpin and Acronis were present competing with VMware on P2V offering.
  • Surgient and VMLogix were present competing with VMware on lab management offering.

Nonetheless VMware allowed all of them to show their products and insert marketing material inside welcome bags.

Unofficial reports mention PlateSpin and vizioncore as the most popular booth, while some big players like Citrix have been barely noted.

It was ironic how EMC Corporation, actually owning VMware, had so low popularity among attendees against competitor like HP and IBM, both present with respective booths.

After 3 years integration between two companies is still near zero and from a marketing point of view VMware does nothing to push EMC brand.

The overall impression has been of fierce competition even if not every company would admit it.

And competition has been greatly welcomed by attendees which have been hungry to discover alternatives on the market.

Among several notable sessions it’s worth to mention:

General sessions

Diane Greene (President and co-Founder) first and Dr. Mendel Rosenblum (Chief Scientist and co-Founder) then introduces big themes of this year’s edition in two general sessions.

To see them register here.

VMmark

VMmark is trying to solve the very complex question of how to reliably calculate application performances in a virtual environment.

It approaches benchmarking in an unusual way, discarding the traditional approach of measuring a single operation (like I/O write) performed on the operating system.
VMware proposes to verify virtual environment performances by loading a mixed set of workloads, made of some specific applications, installed on specific operating systems, inside specifically sized virtual machines. Calculate overall score and compare against a reference score provided by VMware itself.

In details each set of workloads includes 6 virtual machines loaded with:

  • Database server (Oracle Database Server) on Red Hat Enterprise Linux 4 [2 vCPU + 2GB vRAM]
  • Mail server (Microsoft Exchange Server) on Microsoft Windows 2003 [2 vCPU + 2GB vRAM]
  • Web server (Apache) on Red Hat Enterprise Linux 4 [2 vCPU + 512MB vRAM]
  • File Server (Dbench) on Red Hat Enterprise Linux 4 [1 vCPU + 256MB vRAM]
  • Java application server (SPECjbb) on Microsoft Windows 2003 [2 vCPU + 2GB vRAM]
  • idle server (empty) on Microsoft Windows 2003 [1 vCPU + 256MB vRAM]

These virtual requirements imply measuring performances on a host machine with at least 8GB RAM and 2 physical CPUs, while each connecting client (perfoming requests against all 6 virtual machines at the same time) will have to have at least 2GB RAM and 2 physical CPUs.

The overall impression has been of a very early work, where a lot of questions (like binary-translation vs para-virtualization vs hardware assisted virtualization approaches) has still to be addressed before reaching a commonly agreed benchmarking system.

VMmark is aimed to become a SPEC standard, but at the moment Microsoft, XenSource and Virtual Iron are not partecipating in the work and has to be seen if they will officially accept the proposal.

The approach VMware is using (different applications on different operating systems) also implies that VMmark can’t be used to compare server virtualization against OS partitioning (SWsoft Virtuozzo for example). While this is surely correct from a technical point of view, it doesn’t help customers which don’t care what’s behind GUI and just want a way to compare performances and VM/core ratios.

Lab Manager 2.4

Lab Manager 2.4 is going to leverage previous good work from Akimbi in the VMware ecosystem, providing notable features to handle development and QA challenges:

  • self-service provisioning of single virtual machines or complex multi-tier environments
  • fast deploying capability thanks to linked clones
  • capability to run multiple copies of the same virtual machine without caring of networking and hostname details (fencing)
  • capability to save VM state in image library and perform debugging on it
  • capability to limit use of physical resources with permissions and quotas

Final release of Lab Manager 2.4 is expected for end of this year.

The overall impression has been of a still early modification of previous Akimbi work, where VMware decided to completely focus just on a reliable integration with VMware Infrastructure 3.

While this choice is understandable, it’s unpleasant see support drop for VMware Server and Microsoft Virtual Server.

At the moment is not clear if these hosted platforms will be supported in future releases and James Phillips, formerly Akimbi Founder and CEO and now Senior Director of Virtual Software Lifecycle at VMware, talked about existing Akimbi customers granting them notables incentive to move to ESX Server.

Next generation ESX Server Storage architecture

VMware is redesigning actual monolithic storaging architecture for ESX Server, working a pluggable framework where 3rd party vendors will be able to inject their multi-path and load balancing drivers.

This approach will speed up support for newest hardware, reduce time to recertify drivers when ESX Server changes, take advantage of vendor-specific features.
Storage vendors will be able to count on a VMkernel Development Kit and will be allowed to ship ESX plug-ins from themselves.

The overall impression has been that VMware really aims to change IT industry mindset, pushing ESX as a true operating system.

Allowing partners to deliver optimized drivers, in time with their new hardware releases, will create strong indirect sales opportunities, where storage vendors will find much desirable selling ESX Server in bundle with their equipment.

Workstation 6.0

Workstation 6 is going to introduce several new features, enhancing the product in usability, scalability and integration with other VMware products.

On the usability front we have to expect:

  • improved Host/Guest support (including Vista and para-virtualized Linux distributions with VMware VMI) and Drag & Drop capabiliies (cross-platform support coming)
  • new hardware support (including USB2 and multi-display)
  • new way to interact with virtual machines (background running VMs and VNC remote access)
  • new scripting capabilities (introducing the VIX API already used in Server 1.0)

but most of all new features for developers:

  • capability to invoke a virtual machine from Microsoft Visual Studio for code running and debugging
  • capability to replay everything happened in a virtual machine in any moment of its history (with pause, rewind and fast forward)

On the scalability front we’ll see even bigger improvements with:

  • no limit on overall allocated virtual RAM
  • virtual RAM limit per VM raised to 8GB
  • experimental support for 4-way vSMP
  • reduced limit to virtual PCI slots per VM

Integration with other VMware products is going to be notable as well:

Plus a lot more of other improvements.

Workstation 6 is actually in private beta (Friends & Family) and is expected to be opened to public before end of this year. No mention on final relase date planned.

The overall impression has been of a complex product becoming more and more focused on enterprise needs, better than Server in some scenarios.

I would not be surprised one day Server and Workstation would merge in a unique product, offering the web management console as an optional installation feature.

ACE 2.0

Unfortunately I missed this session (which anyway should be under NDA) so not much can be said.

One thing is already known: integration with Workstation 6 where ACE will be a superset of Workstation, allowing IT managers to easily upgrade to ACE Manager with the correct licensing code.

Another thing, unconfirmed, is possibility VMware will sign a deal with Cisco to use ACE 2.0 as part of Network Admission Control (NAC) endpoint security architecture.

(if you don’t have an idea of what endpoint security technologies are feel free to look at my blog about security tecnologies, Security Zero, where Cisco NAC and others have been covered several times)

Workstation for Mac OS (codename Fusion)

VMware didn’t release any session about upcoming Workstation for Mac OS but some details about it has been disclosed by Richard Garsthagen, Technical Marketing Manager at VMware, with a video, and Bob Roudebush, Director of Solutions Engineering at Double Take, with a review of current beta.

The big absentee of VMworld 2006 has been Server, as already said, with few sessions and no mentions about new features coming.

No space at all has been granted to the misterious VM Integrity instead, which should rely on virtual machines patch management at host level and which VMware is working on for sure, as this SearchServerVirtualization interview with Diane Greene confirms.

What has been announced during the conference timeframe:

Products and services from VMware:

Products and services from other virtualization vendors:

Alliances and partnerships

(both Virtualization Industry Radar and Virtualization Industry Roadmap have been updated accordingly)

A final word on the event itself

VMware did a very good job arranging a world-class event.

Conference material was high quality: welcome bag, notebook, conference guide, badge, general sessions directions, etc.

Logistics has been very good too: efficient transports to and from hotels, availability of press/analyst facitilies, availability of email stations, instructor-led and self-paced hand on labs, exam facilities, and free wireless access for all attendees.

Networking possibilities has been the best part: VMware arranged a big booth packed with a lot of company employees, with several special opportunities to meet the engineers.

Alumni, Core Customers, User Groups and Bloggers had also their own lounges where to meet each other and all attendees.

Nonetheless few things could be improved:

  • wireless connection has not been able to support the huge number of clients (I bet at list half of attendees had a notebook)
  • food was not the at the same level of everything else (both during conference and party)
  • quantity of sessions and networking opportunities absolutely require a longer conference to permit everybody to return on investment. Another day is mandatory.

At the end of the week VMworld has definitively been an event to attend. Highly recommended.

Release: InovaWave DXtreme 1.0

A new startup called InovaWave appeared at VMworld 2006: InovaWave.

Founded by two former Surgient executives, Terry Ferose (Vice President of Services) and Dave McCrory (Chief Scientist), InovaWave launched a new product called DXtreme, aimed to provide extreme performance boost for virtual machines in VMware and Microsoft virtual platforms.

The solution is currently available for Windows XP and 2003 and supports VMware Server and Workstation, and Microsoft Virtual Server.
InovaWave claims its product is able to achieve more than double performances, allowing a higher VM/core ratio, more transactions and a shorter boot time for every virtual machine.

DXtreme creates virtual disk channels for each virtual machine, with heuristic-based technologies.

The overhead it creates (from 128MB to 512MB of RAM, and from 5% to 15% of CPU power) is compensated by the big saving for all virtual machines.

Pricing for DXtreme starts at $149 per desktop and $575 per server.

A version of DXtreme for VMware ESX Server is expected to release in Q1 2007. Future releases will support Xen as well.

The virtualization.info Virtualization Industry Roadmap has been updated accordingly.

InovaWave has been included in the virtualization.info Virtualization Industry Radar.