VMworld2009

You are currently browsing articles tagged VMworld2009.

Cleaning Out Some VMworld Info

Monday, September 7, 2009 in Virtualization by slowe | No comments

VMworld is always a very busy time of year. Press releases, product announcements, new product or technology releases, companies emerging from stealth mode—it all happens around VMworld. Now that I’m back home again from VMworld, I thought I’d clear out my Inbox from all the various VMworld-related news items. Perhaps you’ll find something interesting or useful here!

VMware and HP Unveil Solution to Simplify Datacenter Management
World’s Leading Service Providers Build New Cloud Services on the VMware Platform
Altor VF 3.0 Meets VMware’s VMsafe Certification Requirements (no hyperlink available)
VMware Submits VMware vCloud API Specification to the Distributed Management Task Force (DMTF) — First Ever Submission of Key Cloud Interface
VMware Leads in Virtual Desktops With VMware View(TM) — Simplifying Desktop Management, Lowering Costs and Enriching User Experience
Cisco and VMware Validated Architecture for Long Distance VMotion
VMworld 2009 Hello Freedom videos (this one is funny)
VMworld 2009 Virtual Infrastructure Design – Lab Manager vPODS Enable Conference Cloud
VMworld 2009: VMware, Cisco, and EMC Super-Session (SS5240)
VMworld 2009: Best Practices for Multipathing in VI3.5 and vSphere (TA2467)
VMworld: Is it a scalability issue to run drivers in the Hyper-V parent partition? (Answer: No)

That’s it this time around. I promise that some original content is coming soon…

Tags: , , ,

TA2384 – Deploying the Nexus 1000V

Wednesday, September 2, 2009 in Networking, Virtualization by slowe | 17 comments

There is no Internet connectivity in this session, so I’ll have to publish this after the session has concluded.

The Cisco Nexus 1000V is, of course, a Layer 2 distributed virtual switch for VMware vSphere built on Cisco NX-OS (the same operating system that drives the physical Nexus switches). It’s compatible with all switching platforms, meaning that it doesn’t require physical Nexus switches upstream in order to work. The Nexus 1000V brings policy-based VM connectivity, network and security property mobility, and a non-disruptive operational model.

The Nexus 1000V has two components: the Virtual Supervisor Module (VSM). Interestingly enough, the slide shows that the VSM can be a virtual or physical instance of NX-OS; there has been no formal announcement of which I know that has discussed using a physical instance of NX-OS as the VSM for the Nexus 1000V. The second component is the Virtual Ethernet Module (VEM), which is a per-host switching module that resides on each ESX/ESXi host. A VSM can support up to 64 VEMs in a distributed logical switch model, meaning that all VEMs are centrally managed by the VSM. Each VEM appears as a remote line card to the VSM.

The VEM is deployed using vCenter Update Manager (VUM) and supports both ESX and ESXi. The Nexus 1000V supports both 1Gbps and 10Gbps Ethernet uplinks and works with all types of servers (everything on the HCL) and upstream switches.

The Nexus 1000V supports a feature called virtual port channel host mode (vPC-HM). This feature allows the Nexus 1000V to use two uplinks (NICs in the server) connected to two different physical switches and treat them as a single logical uplink. This does not require any upstream switch support. Multiple instances of vPC-HM can be used; for example, you could use four Gigabit Ethernet uplinks, two to each physical switches, could be used to create two different vPC-HM uplinks for redundancy and separation of traffic.

For upstream switches that support VSS or VBS, you can configure the Nexus 1000V to use all uplinks as a single logical uplink. This requires upstream switch support but provides more bandwidth across all upstream switches. Of course, users can also create multiple port channels to upstream switches for traffic separation. There are lots of flexiblity in how the Nexus 1000V can be connected to the existing network infrastructure.

These network designs can be extrapolated to six NICs (uplinks), eight NICs, and more.

One interesting statement from the presenter was that Layer 8 (the Human layer) can create more problems than Layers 1 through 7.

Next, the presenter went through the use and configuration of the Cisco Nexus 1000V in DMZ environments. Key features for this use case include private VLANs (private VLANs can span both physical and virtual systems). Network professionals can also use access-conrol lists (ACLs) and remote port mirroring (ERSPAN) improve visibility and control over the virtual networking environment.

At this point, I left the session because it was clear that this session was more about educating users on the features of the Nexus 1000V and not about best practices on how to deploy the Nexus 1000V.

Tags: , , , , , ,

Some Additional VMworld Vendor Meetings

Wednesday, September 2, 2009 in Storage, Virtualization by slowe | 2 comments

Earlier I posted some notes on meetings I’d had with Virsto and Xangati. In this post I’d like to discuss some additional meetings I’ve had with Virtual Instruments and Tranxition.

Virtual Instruments

Virtual Instruments makes a solution that is intended to help troubleshoot and optimize storage environments. I had the opportunity to grab some coffee with them this morning and hear about what they’re doing and how they’re doing it. As a company carved out of Finisar and taken private, their goal is to help drive higher levels of virtualization by providing more visibility into the storage fabric.

Clearly, this message will really only resonate with larger customers, and that is their target market: multiple hundreds of terabytes into the single petabyte range. At this scale, providing visibility into the thousands of virtual machines across hundreds of ESX/ESXi hosts attached to hundreds of Fibre Channel ports is almost impossible. Virtual Instruments tackles this with a multi-prong approach:

  • First, they use a SAN tap to plug into the Fibre Channel fabric and mirror traffic information to a collection device for analysis. If you’re a networking person, you can think of this as using a SPAN port to mirror traffic. This is done on the storage side to reduce the scale due to fan in-fan out ratios.
  • Second, they gather SNMP information from the Fibre Channel switches. This enables visibility at the switch level.
  • Third and finally, Virtual Instruments collects information from VMware vCenter Server. This information provides the final piece necessary to correlate per-host and per-VM traffic to the information being gathered by the fabric taps and the switch monitoring.

What this allows Virtual Instruments to do is to feed information back to vCenter Server to enable I/O-based recommendations for VM movement. It also enables visibility into path utilization so that multipathing information can be configured for optimal performance. Finally, more detailed storage information is exposed that enables organizations to more effectively place VM storage on Tier 1, Tier 2, or Tier 3 according to its storage needs. In some cases, in fact, money saved on buying additional Tier 1 storage can more than pay for an implementation of Virtual Instruments.

Overall, this is very interesting soltuion, albeit limited in scope to larger environments. If this describes your organization, though, it may definitely be worth a closer look.

Tranxition

Tranxition makes software to do “personality virtualization.” Apparently they’ve been around since 1998 and are just now becoming more visible, creating a partner program, and starting to expand coverage. Their key product is Adaptive Persona, which some have said can be called “Softricity for user personality data”. The product seems to work a lot like ThinApp in that it creates a virtual file system and virtual Registry that captures all user personality data. This user personality data, which can reside either inside or outside the traditional user profile file system structure, is then continuously streamed back to a central server. When a user logs off, whatever data has not been synchronized to the server is then copied up to the server, and the local system is scrubbed of user personality data. Then, when that same user logs on to a different system, Tranxition streams down only those portions of the user personality that are needed at that moment. All other data is fetched “on demand”. This helps speed up the logon process by decoupling the size of the profile from the time required to log on.

Overall, I was fairly impressed with the product. They seem to have done a reasonably good job of taking the principles behind application virtualization and applied them to user personality management. If anyone has any additional feedback on Tranxition (vendors, please disclose yourselves!), I’d love to hear it in the comments.

Tags: , , , ,

Notes on Some VMworld Vendor Meetings

Wednesday, September 2, 2009 in Storage, Virtualization by slowe | 1 comment

I’ve had the opportunity to speak with a few different vendors over the last couple of days here in San Francisco at VMworld 2009. Here are some notes on my meetings.

Virsto

My first meeting of the week was with Virsto, a early storage startup (they just closed Series A funding in the last few weeks). Virsto is led by some long-time storage professionals from companies such as StorageTek, Veritas, and others.

Virsto is unique, to me, in that they have an interesting view of the storage component. I met with Alex, one of the founders, and he used a term that I found quite illustrative and useful: “the I/O blender”. This is the term he applied to the effect that the hypervisor has on I/O as it moves from the virtual server to the physical server to the storage layer. If you think about it, it makes sense: I/O from each virtual server has to be multiplexed onto the same HBAs as the I/O from every other virtual server. The end result is, of course, that the storage array ends up having to deal with small, random I/O workloads instead of large, sequential workloads. This impairs performance.

The Virsto solution combines a software portion that is currently architected only for Microsoft Hyper-V. Virsto’s software component illustrates both the strength and the weakness of Hyper-V’s indirect I/O model. It’s a strength in that it’s very easy to write a filter driver to run in the management partition to modify VM I/O; the weakness is that it’s really easy to write a filter driver to run in the management partition to modify VM I/O. I’m being partially facetious here, but I hope you get the point. In any case, what Virsto’s software layer does is help undo the I/O blender effect by working in conjunction with a storage staging layer. Typically this would be some sort of high-speed local storage, such as an SSD. As a result of the software working in conjunction with the hardware, Virsto can “re-assemble” I/O into workloads that are better suited for performance at the physical layer and thus undo the I/O blender effect.

Virsto’s solution also allows for some forms of storage virtualization, in that different types of underlying block storage can be combined and managed by Virsto. Virsto’s solution also offers snapshots (checkpoints), the ability to split data streams for replication, and better support for disk-to-disk (D2D) backups via their snapshots.

My biggest concern with Virsto is that they are competing in a space with lots of very large, very well-funded organizations that are laser-focused on making their storage work extremely well with VMware vSphere and virtual environments, including Hyper-V. Think of NetApp integration with Hyper-V, or EMC integration with Hyper-V (remember that Virsto supports only Hyper-V at this time). These companies have lots of development talent, lots of money, and an established presence. I fear it will be difficult for Virsto to really gain a foothold in that space.

Xangati

Xangati (pronounced “zan-gotti”) is an application performance solution. I had a spirited discussion with the Xangati folks about what differentiates them versus other solutions like AppSpeed, BlueStripe, etc., in that Xangati relies upon network traffic information to measure application performance. In Xangati’s case, they rely upon NetFlow (or the various vendor-specific implementations of NetFlow). At first, I found this a bit limiting because I wasn’t aware that NetFlow v5 was supported in ESX 3.5 on vNetwork Standard Switches (I know, this is probably something everyone knows). But it indeed is (see here); the real question is whether it will continue to be supported on vNetwork Standard Switches on vSphere. In any case, Xangati insists that using NetFlow to gather network information is very different (and yields different results) than performing packet analysis. I must admit that I don’t fully see the difference; perhaps a network guru can explain it?

Having stated all that, what Xangati does it pretty interesting. It requires that you enable NetFlow throughout the environment—on both virtual and physical switches and other physical network equipment—and then allows you to see end-to-end network usage on an application-by-application basis. From there, Xangati allows the organization to take “network recordings” of the network behavior and then replay that recording later. Different views can be created for different roles within the organization, allowing IT pros to see only the information they need or want to see.

I’ll go back to my earlier statements and say that while Xangati offers some unique functionality—such as the ability for an end-user to initiate a network recording and submit that as a “Visual Trouble Ticket” to the help desk—I’m still at a loss to explain how, in the end, they are different from AppSpeed, BlueStripe, and others who provide end-to-end application performance and correlation. Yes, they use a different way (perhaps a superior way) of gathering information, but what the customer ultimately wants to know is this: “Which part is slow?” It seems that there are other, more well-established solutions already on the market that are trying to address this. Whether or not Xangati is successful in the space is yet to be seen, but I do wish them the best of luck!

I also met with Virtual Instruments and had a great discussion with them, but as I’m running a bit short on time I’ll have to do their write-up later on. Check back here later for the write-up on Virtual Instruments.

Tags: , , ,

VMworld 2009 Day 2 Keynote

Wednesday, September 2, 2009 in Virtualization by slowe | 1 comment

The day 2 keynote starts with another entertaining video.

After the video concludes, Steve Herrod, VMware’s CTO, takes the stage. He echoes that virtualization is a “tectonic shift” in the IT industry. Herrod again shows the three pillars—VMware vSphere, VMware View, and VMware vCenter. Herrod’s initial focus is on VMware View, due to the increased emphasis on desktop virtualization and the focus VMware is providing in that arena.

VMware vSphere provides the foundation for desktop virtualization and is the best platform for desktop virtualization due to commonality, security, availability, and efficiency. VMware’s advances in this area are coupled with advances by Intel (Xeon 5500) and storage (high-speed block caching, for example). VMware View builds on this foundation to enable faster and more efficient image provisioning, image updating, and policy enforcement. One piece that helps with this is the effort that VMware has placed into decoupling the OS, applications, and user profiles.

VMware is announcing this morning that they will be OEM’ing RTO Virtual Profiles to assist in decoupling user profile/user personality data from the underlying OS image and the applications. Unfortunately, Herrod did not provide any additional detail on exactly how RTO’s product will be integrated (or if it will even be integrated).

User experience is another key point. To that end, VMware has been focusing on the user experience to enable the “productive desktop” across the WAN, the “PC-like desktop” across the LAN, and the “rich portable desktop” using VMware CVP (Client Virtualization Platform).

To help with user experience on the LAN and WAN, VMware has been working closely with Teradici on PC over IP (PCoIP). VMware’s implementation of PCoIP is software-only, but fully supports hardware acceleration. This enables the use of the same protocol from task workers to knowledge workers to designers. PCoIP will be included in the next version of VMware View and will be shipping later this year.

Employee-owned IT (EOIT) is another area that VMware is trying to enable. This can be accomplished in a couple of different ways. One way is using hosted virtualization with a product like VMware Workstation, VMware ACE, or VMware Fusion. VMware ACE has policy support (and I think that Workstation does as well), but Fusion does not have policy support, so it sounds like VMware is planning on extending policy support across all hosted virtualization platforms.

Of course, CVP is the primary focus for VMware, where they are leveraging their relationship with Intel and using the Intel vPro technology to perform desktop virtualization with a local hypervisor but managed centrally via VMware View. This leads into a demo of CVP. The demo shows Windows 7 running on a local Type 1 hypervisor, and then shows a session using a thin client on a local connection using PCoIP. The demo wraps up with a demonstration of using Wyse PocketCloud on an iPhone.

Regarding VMware’s mobile strategy, Herrod begins to discuss the various ways in which VMware is enabling the use of mobile devices. vCenter Mobile Administrator is one product that VMware is providing, plus the VMware View-approved iPhone clients like PocketCloud, and—of course—virtualizing the phones themselves. This leads into a discussion of VMware Mobile Virtualization Platform (MVP).

VMware MVP is about device freedom as well as application freedom. Visa takes the stage and, rightfully so, Herrod asks why Visa is on the stage for mobile virtualization. His focus is about enabling financial services on mobile platforms without compromising security. MVP helps cut complexity to enable that functionality. Herrod then introduces Srinivas Krishnamurthy to show an MVP demo.

The MVP demo primarily focuses on the Visa Mobile app, but Srinivas then exposes the relationship of this demo to MVP: the application he’s showing is a Google Android application running on a Windows Mobile phone. This enables users to use whatever application they want regardless of the platform.

Herrod now shifts the focus to VMware vSphere. (He promises not to demonstrate VMware Fault Tolerance again.) Once again Herrod brings up the idea of the software mainframe. VMotion is a key component, a key enabler, of the giant computer (aka the software mainframe aka the cloud). This is the sixth anniversary of VMotion, and VMotion now has maturity, breadth, and automated use. Herrod says VMware estimates a new VMotion migration occurs every 2 seconds. (Where does that calculation come from?)

VMotion was extended to Storage VMotion in late 2006, and with vSphere it was extended again to Network VMotion, and extended again with long distance VMotion via some additional partners and vendors. Herrod sees VMotion as a major part of VMware’s drive to greater efficiency (one of three marketing pillars around vSphere: efficiency, control, and choice). This is because VMotion is a foundational technology for VMware DRS. He uses the recent study about DRS and improved performance as an example. VMware DRS is being extended (no timeframe yet) to include I/O. This includes the ability to assign shares and IOPS values to individual VMs.

VMotion and VMware DRS are combined yet again to form a foundation for VMware DPM, which enables greater power efficiency.

Herrod next moves from efficiency to control, and what VMware is doing is this area. His first topic in this space is VMware AppSpeed, which helps organizations provide a level of control over application performance. Next Herrod moves into a discussion of vApp, which is a logical collection of one or more VMs described using OVF. This will help enable the “IT service” policy descriptor. This means embedding SLA definitions into the OVF standard and enabling vSphere to act upon those definitions appropriately.

The VMsafe APIs are another point of control and are now officially available with VMware vSphere (released back in May). Embedding security policies into the OVF description and integrating it with vApp is another method of extending control in a policy-based way.

Next Herrod moves into a demonstration of VMware ConfigControl. This is the first time ConfigControl has been demonstrated. The demo was a bit limited, but considering that it won’t be shipping until early next year that’s not too terribly surprising.

Herrod finally moves on to the third vSphere marketing tenet: choice. Choice includes choice of hardware and choice of where to run applications. Herrod also sees choice as including self-service IT, which leads to a discussion of Lab Manager. (It sounds to me like Lab Manager is a major central focus point for VMware as enabling self-service IT.)

The last focus point in Herrod’s keynote is cloud computing, i.e., VMware vCloud. (I suspect we’ll see more information on vCloud Express and the official announcements of the vCloud APIs.) Herrod calls out the use case of Site Recovery Manager to connect two internal data centers as a (limited) form of cloud connectivity that customers are using today.

Long-distance VMotion is another level of cloud connectivity that VMware and associated partners are trying to tackle. There are lots of challenges to be addressed here. Why is long-distance VMotion so important or interesting? It could be follow the sun computing, disaster avoidance, etc. In the partner realm, Cisco is one partner that is working on long-distance VMotion, but there are still network identity challenges. F5 is using BigIP to abstract network identity to help address some of the challenges around long-distance VMotion.

Moving on to the VMware vCloud APIs, Herrod talks about a few applications leveraging the API. He also reiterates that the vCloud API has been submitted to the DMTF for consideration as a multi-vendor standard. This will help foster choice.

At the beginning of the keynote, Herrod laid out three initiatives: View, vSphere, and vCloud. Now he adds a fourth one: vApps. This is more related to the SpringSource acquisition, not to the vApp functionality within vSphere (as far as I can tell). Herrod begins to explain how the SpringSource acquisition allows VMware to move “up the stack” from Infrastructure as a Service (IaaS) into Platform as a Service (PaaS) and provide more products to support Software as a Service (SaaS). Adding SpringSource to the VMware mix gives VMware coverage in 2/3 of the cloud definitions (IaaS and PaaS covered).

This move also allows VMware to optimize the IaaS and PaaS layers to provide even greater performance, mobility, and management for SaaS vendors, developers, and providers. Adrian Coyler, CTO of SpringSource, now takes the stage for a demonstration.

The demonstration shows a Java application being deployed to an external cloud using CloudFoundry, a company that SpringSource themselves acquired just in the last month or so. This demonstration was actually shown yesterday during Maritz’ vCloud event.

Herrod wrapped up the keynote with a summary of the key takeaways.

UPDATE: Here are some links to additional coverage of the Day 2 keynote:

VMworld Keynote – Day 2
Live blogging the VMworld 2009 Day 2 “technical” keynote
Live from VMworld 2009: Day 2
Thoughts on the VMworld Day 2 Keynote

As with the Day 1 keynote, I also encourage readers to check out the Planet V12n feed for some great coverage.

Tags: , ,

BC3209 and DV2484

Tuesday, September 1, 2009 in Virtualization by slowe | 2 comments

I had on my schedule for today two sessions, BC3209 and DV2484. BC3209 was titled “Creating the Fastest Possible Backups Using VMware Consolidated Backup – A Design Blueprint”; DV2484 was titled “Server and Storage Sizing for VMware View”. You might be wondering why I haven’t published summaries of either of these two sessions.

I skipped out on BC3209 to attend the vCloud event with Paul Maritz, which—unfortunately—turned out to be a complete non-event. So, that’s why you don’t see any coverage of that session.

As for DV2484, I did attend the first 30 minutes of that session, but didn’t find anything particularly useful in the session. At least, nothing that I hadn’t already seen or experienced before. The portion of the session that I attended simply discussed how to gather information about the physical desktop. They did provide some examples of information they’d gathered:

  • Average of 130MHz of CPU utilization
  • Average of 5 IOPS of disk transfers
  • Average of about 75MB of RAM used by a VM with Windows XP/Office installed (after TPS kicked in and reclaimed some RAM)

I found the IOPS figure rather interesting; I’d had some conversations earlier in the week during the PTAB meetings with another participant around VMware’s reference architecture for VMware View. In those reference architectures, VMware had only specified 7 IOPS for each desktop, and that PTAB participant felt that number was far too low and, in fact, had been bitten by trying to use that reference architecture in real-world implementations. This makes me question the 5 IOPS figure that was presented in DV2484. The real question is, if 5 IOPS isn’t realistic, what is realistic? I’d love to hear some feedback from readers on that question.

Tags: , ,

VMware vCloud Event with Paul Maritz

Tuesday, September 1, 2009 in Virtualization by slowe | No comments

This special event with Paul Maritz around VMware’s vCloud initiative took place back in the same general hall as the morning keynote. The subject of this special meeting is VMware’s cloud offerings.

VMware’s strategy has three legs: choice, application compatibility, and something else that I didn’t stay long enough to find out.

With regard to choice, VMware now has over 1,000 service providers that provide VMware-ready cloud services. Examples include services from AT&T and others. AT&T takes the stage first to discuss their enterprise-class cloud services. Particularly interesting is AT&T’s Synaptic Storage As a Service offering, which leverages EMC’s Atmos cloud storage offering. Unfortunately, AT&T is short on details about this specific service. After the AT&T presentation is over, VMware takes over again to provide a demonstration of workload federation.

Dino Cicciarelli takes the stage to run the demonstration of the workload federation. The demonstration will show workloads being moved between corporate data centers or from a corporate data center up to an external cloud service provider. The demonstration shows a SQL Server instance under load being migrated between two data centers about 35 miles apart. (There was some questions among attendees as to whether the demonstration was real or simulated; I’d be interested to know the real story behind the demonstration. Anyone want to take me up on that?)

Next on the stage is Savvis. Like AT&T, Savvis spends some time describing their company, their services, etc. Their announcement is their next-generation data center service, called “Project Spirit.” This is a virtual private data center (VPDC) offering with multi-tiered QoS capabilities. The Savvis discussion concludes with a demonstration by Savvis of Project Spirit. It’s pretty clear, though, by the way the demo played and how the Savvis representative was talking on the stage, that the “demo” was really nothing more than a Flash demo or screen movie. (I’m very disappointed.)

Verizon Business is the next player to take the stage and discuss Verizon’s products and solutions. Verizon’s solution is called Computing As a Service (CaaS). (Interesting side note, ePlus has been working with Verizon Business on the design and development of CaaS.) Verizon’s announcement for today is the expansion of their cloud platforms, adding new capacity in Amsterdam and an expanded presence in Europe.

VMware vCloud Express is another part of VMware’s efforts to support choice. Of course, vCloud Express was already discussed (and supposedly demonstrated) in the Day 1 keynote. Oh, look, another partner taking the stage: this time it’s Terremark. One very interesting piece that Terremark showed in their demonstration was PowerShell integration with vCloud Express; now that’s the kind of information they should be showing off here!

Moving away from choice to application compatibility, Paul Maritz again refers to the formal announcement of the vCloud API. The vCloud API is actually a series of APIs that are being/have been submitted to standards organizations (as I mentioned in the keynote coverage, I believe it was submitted to the DMTF). SpringSource takes the stage to talk about what they do and then perform a demo (a live demo?) of their products and technologies. The demo shows off SpringSource and CloudFoundry deploying applications to an external cloud.

I’m sure there are more announcements buried in this event somewhere, but I have to shut down coverage and move on. I’ll update this post later with links to coverage from other bloggers and press members.

Tags: , ,

BC1500 – Site Recovery Manager Best Practices

Tuesday, September 1, 2009 in Virtualization by slowe | 1 comment

I arrived at this session a few minutes late because the VMworld 2009 Day 1 keynote was running late. The early part of the session was pretty straightforward with information on the various SRM components: multiple vCenter Server instances, SRM servers, storage replication, SRAs, etc., so I won’t bother recreating this here. Most people are already familiar with the basics of SRM.

The technology preview of SRM (the beta of the next version) does support vCenter Server linked mode, which means that your two vCenter Server instances can run in a linked mode group. Very nice!

The next version of SRM will allow you to change the authentication option via a Repair installation. This is a big improvement from the current release, where this has to be selected at installation and can’t be changed afterward.

VMware KB article 1008390 provides more information on how to setup the self-signed certificates and which subject names should be used for correct operation.

It’s important to remember that SRM doesn’t perform replication; the storage performs the replication. The SRA allows SRM to control the movement of that storage and how it is presented to the various hosts involved in the solution.

The SRA is written by the storage vendor and must be installed on both SRM servers. Multiple SRAs can be installed on the same SRM server to support multiple storage platforms. New adapters do not require SRM updates. The SRA then supports tasks such as setup, testing, and failover. So what goes wrong with SRAs?

  • Not all SRAs are created equal. They support the same functionality but the implementation might be different.
  • Some SRAs have external requirements, such as the Java JRE, or storage-specific components from the storage vendor. The key takeaway is to be sure to read the documentation from the storage vendor about the SRA. The SRM community forum can also be a good source of information.
  • SRAs expect a certain level of replicated storage configuration. This might include little quirks like ensuring that certain flags or attributes are set as expected by the SRA.

Some SRAs talk directly to the storage array, but some SRAs require connectivity to a control station or management station. Again, be sure to refer to the SRA documentation to be sure of the configuration requirements.

Recommendation: whatever you do at one site, duplicate that effort at the other site. This will help simplify things.

Some storage arrays require access to “gatekeeper LUNs”; when running SRM as a VM, you can present these LUNs to the SRM VMs as RDMs. Other storage implementations, like HP EVA or IBM SVC, need access to the management server itself and not to the array directly. HP recently released a guide to using SRM with HP storage that might be a good idea to read.

Other pitfalls during SRA setup:

  • Attempts to configure SRA at protected site will fail if SRA components are not installed at recovery site.
  • For active/passive setup, just perform configuration at the protected site, not the recovery site. Configuration at the recovery site is only required for failback or active/active setups.
  • Double-check path definitions; older SRAs used Perl or similar and caused problems with failing to update the path definitions. Get updated SRAs; this has all been fixed in newer versions of the SRAs.
  • Be sure the array configuration matches the pre-requisites.
  • SRM might need additional licensed features in order to perform a Test failover. (Key example: FlexClone required on a NetApp array in order to perform test failovers.)
  • Empty datastores won’t be detected by SRM even if it is replicated. This has been fixed in the next release.

Next the presenter shifts to a discussion of virtual machine networking.

SRM will work with stretched VLANs. To architect SRM to work with stretched VLANs and failover tests, you can use the bubble network, but that isn’t really practical. Instead, the presenter recommends the creation of test VLANs on the physical network to support failover testing.

Instead of stretched VLANs, you can also use disparate networks. In this case, of course, the VLANs change from the protected site to the recovery site. This impacts IP addressing, DNS, Active Directory, default gateways, DHCP, etc. You can use existing tools to fix these problems; you don’t have to use the SRM-supplied tools. SRM supplied tools include VM customization or the dr-ip-customizer.exe batch utility. Dynamic DNS updates are another option as well.

So what about virtual machine disks? What are the storage factors that come into play here?

  • Current version of SRM supports iSCSI and Fibre Channel, and supports both physical and virtual RDMs
  • VMs can span multiple datastores (but need to consider replication consistency)
  • SRM does not support VM snapshots
  • VMs cannot span arrays, even arrays of the same type

The presenter next moves into a section he called “Running,” but it was really more about how to walk through the site pairing, storage array configuration, and inventory mapping (more setup and configuration than operation, in my opinion).

A few notes about protection groups, placeholder VMs, and recovery plans:

  • SRM supports up to 150 protection groups. Each protection group may have multiple LUNs.
  • SRM supports up to 500 VMs. This should increase to 1000 VMs per site in the next release.
  • Need a datastore at the recovery site to hold VM placeholders. This datastore does not need to be replicated. Each placeholder is typically <1KB each, so there is not a lot of storage required.
  • Don’t delete the placeholders; this breaks the SRM configuration.
  • SRM supports multiple recovery plans, so you can use mutliple plans to support multiple recovery targets.
  • Recovery plans support permissions, so not everyone needs access to all plans.

In general, also be sure to keep in mind the “upstream” and “downstream” dependencies. SRM isn’t going to fix dependencies for Microsoft Exchange, for example.

I’m ending coverage here, because I have to leave to go to a special, super-secret press announcement with Paul Maritz. I’m not allowed to blog, tweet, or any such thing during the press announcement, so you’ll have to wait until after the event to hear anything from me.

Tags: , , ,

VMworld 2009 Day 1 Keynote

Tuesday, September 1, 2009 in Virtualization by slowe | 3 comments

(Note: You can also follow my Twitter updates here.)

I start out the Day 1 keynote at VMworld 2009 at the bloggers’ table, sitting with folks like Rodney Haywood, Mike Laverick, Alessandro Perilli, Rich Brambley, Thomas Bryant, Bob Plankers, and—of course—the ever-present John Troyer. There should be some good coverage of the event. Later today, I’ll update this post with links to the coverage from some of the other bloggers here.

The keynote session starts off with Todd Nielsen, VMware’s Chief Operating Officer (COO). He recounts the story of how he challenged the partners at Partner Exchange to win over one of the 40 customers in the Fortune 1000 for a free pass to the conference. It turns out that 10 of those 40 customers did switch to VMware, so there are now only 30 companies left in the Fortune 1000 that don’t use VMware software.

Mr. Nielsen then goes on to recognize the conference sponsors and the lab sponsors.

The VMworld 2009 conference has 12,488 attendees. That’s up quite a bit from the estimated attendance of only 8,000 attendees. It’s also funny how marketing and sales like’s to up those sorts of numbers (“It’s almost 15,000!”)

Mr. Nielsen goes on to talk about how VMware is driving cost savings (optimize financial energy), shift human energy, and save the Earth’s energy (via power reduction). Together these help VMWare “energize business through IT.” Spend more money on innovation, and less money on maintenance. This has been a key statement in VMware’s marketing over the last few months.

Now Paul Maritz takes the stage. He starts out by echoing Mr. Nielsen’s comments about needing to spend more on innovation instead of spending on maintenance. Mr. Maritz believes that business agility depends upon IT agility, and IT agility depends upon virtualization.

Mr. Maritz now moves the discussion to—what else?—the idea of cloud computing. He believes that virtualization is a key enabler, a key component, of cloud computing. It allows organizations to insert “cloud mojo” into their data centers. VMware vSphere is the product that provides the ability for organizations to transform their data centers and create private clouds. Once an organization has built their private cloud, VMware vCloud creates the bridge to external clouds.

The virtualization journey starts with CapEx savings and server consolidation, continues on to OpEx savings and internal/external cloud, and finishes with IT agility and business agility. According to Mr. Maritz, the release of VMware vSphere is a bigger release than any Windows release on which he worked while at Microsoft. (Is that a good thing?)

Next, Mr. Maritz reviews the vSphere architecture and discusses the new features and functionality that have been added to the product in this release. This includes DPM, new storage optimizations (more than 350,000 IOPS from a single server), new networking, etc. He reiterates: “There is now literally no application that cannot be considered a candidate for virtualization.”

So what is VMware doing with this great virtualization layer? VMware is introducing a family of new products that address specific usage scenarios. Mr. Maritz is referring to the vCenter family of products; not only vCenter Server, but Chargeback, AppSpeed, etc. His initial focus on vCenter is around the ability of integrate vCenter with other products via APIs. To that point, Mr. Maritz brings out an IBM representative to discuss how IBM is leveraging vCenter APIs to provide integration. This is illustrated with a demonstration of how IBM’s power meter, embedded in the IBM hardware, has been integrated with vCenter Server.

In the demonstration, IBM shows how their power meter is reporting power usage back to vCenter Server. This information is being reported not only for the physical hosts, but power is being reported for virtual machines as well. (I wonder upon which metrics IBM is basing virtual machine power usage?)

Following the demonstration, Mr. Maritz moves on to some new members of the vCenter family: capacity planning (CapacityIQ), configuration (ConfigControl), operations (Chargeback?), and continuity (SRM). Additionally, VMware is introducing products or technologies to create service profiles, service catalogs, enable self-service, and further extend chargeback functionality. In addition, VMware will be focusing on increasing app visibility.

This leads into another demonstration, this time with Bruce of VMware. Bruce is going to be demonstrating Lab Manager 4.0, which is enabling the self-service aspect of the private cloud. Next up Bruce and Mr. Maritz walk through a demonstration of vCenter Chargeback.

The discussion moves on now to vSphere Essentials, which provides “IT in a box” for smaller organizations. VMware is extending their push into the SMB market with VMware Go, a new service being introduced by VMware this week at the conference. VMware Go is a set of web-based services targeted at the SMB customers that helps automate the installation of VMware ESXi and helps engage partners and the virtualization community.

The next stage of the virtualization journey is bridging the connection to the external cloud. This leads into a discussion of VMware vCloud and vCloud Express. To help make this possible, VMware is adding a new construct called the “virtual datacenter”. The virtual datacenter allows organizations to amalgamate internal and external clouds for management, provisioning, resource allocation, etc. A key to making this work is having external providers ready to work with organizations and VMware to enable them to slide virtual datacenters from the internal cloud to the external cloud. (This sounds great and wonderful, but I think there are a lot of challenges yet to be addressed before this becomes reality.)

Mr. Maritz at least recognizes that organizations must have the ability to migrate out of the cloud as well as into the cloud. That’s better than some people I hear talking about cloud computing.

vCloud Express is targeted at providing the ability to quickly and cost-effectively turn up virtualized resources, targeting the sort of things that organizations can do today with Amazon. Bruce comes back out again for a demo of vCloud Express, created in conjunction with VMware’s partner Terremark. (vCloud Express looks interesting, although I’m not sure how excited companies will be about the ability for their employees to go create new virtual machines with an external provider for “only $20-30 per month”.)

Today VMware is formally announcing the vCloud API. This is something that VMware has been discussing for quite some time. In addition, this API is being submitted to standards organizations (the DMTF, I believe).

Another part of VMware’s “virtualization journey” involves desktop virtualization with VMware View. To help further discuss VMware View and how desktop virtualization is evolving, Mr. Maritz invites out Todd Dupree from HP. (While I love HP, this part is mostly a commercial. Bummer.)

Todd from HP does, at least, take a few minutes to discuss a new product from HP, Insight Control for VMware View. This is reasonably interesting and relatively well-integrated into vCenter Server and the vSphere Client, and provides hooks back into other HP products and management consoles.

PCoIP (PC over IP) is key component of ubiquitous desktop virtualization. To help illustrate that, Mr. Maritz next invites Chris Renter, of TELUS Communications in Canada. Mr. Renter logs into a virtual desktop using the next version of VMware View—presumably using PCoIP—to provide a short commercial about TELUS and how TELUS is using VMware View.

Mr. Maritz now moves into a discussion of the SpringSource acquisition, as a “final step” in the virtualization journey. Why is the SpringSource acquisition important? VMware believes that the use of lightweight frameworks and virtualization will result in “radical simplification” and will help eliminate redundant and complex layers of management. Mr. Maritz reiterates the commitment to keep the Spring framework as an open source project, and that the Spring framework will continue to support multiple platforms and Java services. What will VMware do with SpringSource? The idea is to find ways to slide the Spring framework onto VMware vSphere and enable ways for Spring to inform the hypervisor about application requirements and behaviors. To help show this off, Mr. Maritz invites Rod Johnson to take the stage and show off some of the SpringSource functionality.

My next session starts in 5 minutes, so I’m packing it up now and moving on. I’ll update this post later with links to other bloggers’ coverage of the keynote.

UPDATE: Here are some additional links with coverage of the Day 1 keynote:

Live from VMworld 2009 – Day 1
VMworld Day 1: Tuesday: The Keynote
Live from VMworld 2009: Day 1
VMworld 2009 keynote – day one
VMworld 2009 (San Francisco) Day 1: Key Note – A Summary
VMworld 2009 Keynote Day 1
Data Center Strategies: VMworld Day 1 Keynote A Few Thoughts

Readers should also refer to Planet V12n for more information as well.

Tags: , ,

VMworld Day 0 Wrap-Up and Links

Tuesday, September 1, 2009 in Virtualization by slowe | 4 comments

Today was a crazy day. I started the day with the second day of the Partner Technical Advisory Board (PTAB) meeting.

Of course, VMware released a bevy of press releases this morning, and being stuck in the PTAB meeting I haven’t really even had the chance to digest the news announcements. The announcements included:

Next-Generation VMware Partner Network Helps Partners Increase Virtualization Expertise to Expand Business Opportunities
VMware Ushers in a New Era of IT Management to Drive Business Agility with VMware vCenter™ Product Family
VMware Introduces VMware Go™—A New Service that Will Make Virtualization Even Easier for First Time Users
VMware Signs Agreement with Intel to Expand Reach of VMware vSphere™ 4 to SMB Customers
Broad Virtualization Ecosystem Rallies Behind VMware vCenter™ Product Family to Deliver Integrated Solutions for Simplified Datacenter Management
VMware Continues Its Reign as the Lowest Cost Per Application Leader with VMware vSphere 4″

That wrapped up about lunchtime, so I joined Crystal and Kathleen Troyer (John Troyer’s wife) for lunch at Amber, an Indian restaurant just behind the San Francisco Marriott. (Turns out Duncan Epping was eating lunch there, too.) After lunch was complete, the afternoon flew by; Crystal and I barely had time to squeeze in dinner at John’s Grill before heading over to the reception event.

The reception was good; I worked the vExpert booth with a number of fellow vExperts around, like Jason Boche, Rich Brambley, Eric Siebert, Steve Beaver, and a couple others. There wasn’t as much traffic to the booth as I had hoped there would be, but that’s OK.

When the reception ended, Crystal and I headed over to the Clift Hotel for an event in the Spanish Suite with EMC, VMware, and Cisco. I chatted with some Cisco and EMC folks (not so many VMware folks), and ran into Duncan (yet again). That was a great event but horribly crowded, so we left around 10 and headed back to our hotel.

Of course, the biggest news of the day was that my book, Mastering VMware vSphere 4, is sold out at the conference bookstore (or very nearly sold out). Unfortunately, there won’t be any way to get more before the conference ends. Sorry, guys, but take consolation in the fact that Amazon.com is still carrying the book in stock.

Here are some other blog entries about VMworld 2009 Day Zero:

VMworld 2009 – Day 0 Highlights
VMworld 2009 conference coverage from SearchServerVirtualization.com
3 Word Challenge – VMworld Reception and Monday Night Tweetup
VMworld Sunday and Monday – Monkeys Fly
Random VMworld 2009 Thoughts
VMworld 2009: Day 1

Tomorrow starts bright and early with a vendor meeting, and then we’ll quickly move into the Day 1 keynote, sessions, more vendor meetings, and more notes from the exhibit show floor. Stay tuned here for updates throughout the day tomorrow.

Thanks for reading!

Tags: , , , ,

« Older entries