Stay tuned for the liveblog of the keynote by Dr. Stephen Herrod, VMware CTO, live from VMworld Europe 2009 in Cannes, France.
Stephen Herrod was one of the original developers of ESX predecessor SimOS while at Stanford University (His homepage from the time is still online.). He worked at Transmeta, where he worked on their “code morphing” technology (somewhat related to the binary translation VMware is donig). He has been with VMware for a few years now, managing the ESX group.
Expectations for today’s talk are a bit unclear. Will the focus be on the desktop (VMware View, vClient, …), or will it be another “cloudy” day? Given the fact that there aren’t lots of sessions on vSphere (the next generation of the ESX product), expectations in this area are a bit low. This being said, vSphere 4.0 should still be released in the first half of 2009. (Although, given enough beers, some VMware employees are telling us august/september is a more realistic timeframe.)
Let’s find out…
8:44 The room is filling up for the keynote by Stephen Herrod. Keynote starts at 9.00 AM.
9:10 The introduction video (same as yesterday) is rolling.. Getting ready for the keynote. Dr. Steve Herrod seems fashionably late today.
9:11 Maurizio Carli, general manager EMEA is master of ceremonies again, welcoming the audience.
9:14 Small reminder by Carli that this afternoon there will be a session called “VMware Unplugged”, a Q&A session with CEO Paul Maritz, CTO Steve Herrod, Maurizio Carli and the COO Tod Nielsen.
9:14 Stephen Herrod takes the stage.
9:14 A reminder of the three initiatives – VDC-OS, vCloud and vClient. Today those concepts will be demoed in reality.
9:15 Lots of development done in EMEA. Several hundred engineers here.
9:15 The talk is called “The future of virtualization”. Subtitle: “Technical stuff”.
9:16 Herrod is walking through the different blocks of vSphere. First up: vCompute. ESX scales higher than ever before, to cope with today’s platforms and VM demands. Up to 8 vCPUs, 100K IOPS/sec.
9:18 Internal benchmark of Oracle 11g on RHEL running on next generation ESX on a development 8-core Xeon.
The numbers:
less than 15% overhead or 8vCPU VM, 24.000 total DB transactions per second. Near-perfect scalability from 1 to 8 vCPUs. 250 MB/sec disk I/O.
9:18 This is the performance of a 2002 Sun Fire 15k, then costing hundreds of thousands of dollars.
9:19 Not every application scales as well as Oracle. A reminder that breaking up hardware platforms can give better-than-native performance. Exchange supports about 8K mailboxes natively on a platform, breaking this up with ESX gives 16K mailbox support on the same hardware.
9:21 Record SPECweb2005 performance, also with VMware ESX Server 3.5. 16 Gbit/second throughput. Would server 3 billion page views per day. (By comparison, eBay servers about 1 billion pages in a day).
9:21 Next block: vStorage: aggregating and optimizing disk usage. One example: Thin Provisioning (on VMFS/VMDK level).
9:22 VMware has a software implementation, but if the hardware array supports it natively (read: with higher performance), it plugs in to ESX and the thin provisioning is done by the hardware.
9:23 vNetwork: distributed virtual switches (which everybody’s heard of by now), enabling third party virtual switches. (Cisco Nexus 1000V)
9:25 Important because the ownership of the network can stay with the networking team (same management tools, IOS, …) and new functionalities can be added.
9:25 vCompute, vStorage and vNetwork are the “bottom layer” that make up the “giant computer” – the software mainframe.
9:25 DRS and related technologies take care of distributing the workloads across the physical building blocks.
9:26 Distributed Power Management, today an experimental feature in ESX 3.5, makes this giant computer power thrifty.
9:27 DPM can throttle CPU speed, and turn off hosts that aren’t needed. When bursting, hosts can be powered on as needed.
9:27 DPM is one of the built-in features that make the “giant computer” as self-managing as possible.
9:28 The “top layer” of vSphere is related to management, policies and SLAs.
9:28 One example: attach an SLA policies to vApps (containers of virtual appliances that make up tiered applications), vSphere will allocate the right resources (and chargeback for the usage).
9:30 Maximizing uptime stays important as well. There are features for planned and unplanned downtime: VMotion, Storage VMotion, Maintenance mode, HA, SRM and of course Fault Tolerance (nee Continuous Availability). Nothing new here.
9:31 A reminder of what FT is all about: for some VMs HA isn’t enough, a reboot introduces unacceptable downtime.
9:32 FT runs a “shadow VM” on a separate server, and cuts over to the machine in case the first one goes down. As this is done in software, it can be configured on a per-VM basis, and it uses “off the shelf” hardware. (as opposed to specialized cluster equipment)
9:33 FT works together with HA: if one of the two copies dies, a new shadow VM is booted up, keeping the protection intact.
9:34 Next topic: Security – VMsafe APIs help protect workloads without needing to install agents in VMs. VMware’s partners have been working with this APIs for a year now (VMsafe was announced last year at VMworld Europe 2008), a lot of products should ship together with vSphere.
9:35 The security settings are a part of the vApp policies, the follow the VM as it moves between clouds.
9:36 Another vSecurity-based feature are vShiled Zones. This technology came to VMware with the acquisition of Bluelane. vShield zones encapsulate and firewall VMs, regardless of where they are running.
9:37 To manage all this nice stuff we’re used to VirtualCenter. This will become vCenter, as a central management hub.
9:39 As vCenter becomes an increasingly important part of a virtual infrastructure, VMware introduces vCenter Server Heartbeat. A passive vCenter server can run in the background and take over in case the first one goes down.
9:39 This monitoring/heartbeat mechanism keeps logging, allowing rollback of misconfigurations.
9:42 VMware infrastructure environments become bigger and bigger. vCenter has several limits – 200 hosts and 2000 VMs today, 300 hosts and 3000 VMs tomorrow. Multiple vCenter servers can be linked together to overcome this boundaries. Up to 10 vCenter can be linked in linked mode.
To keep things manageable, the Virtual Infrastructure client now includes a search mode.
First demo on stage of the new VI/vCenter combo, focusing on the search interface.
Anyone who’s ever used a web browser will know how to use this.
9:42 Advanced search capabilities to refine search results.
9:43 Possible business model: ads in the vCenter search results.
9:44 Automation is not a new concept for VMware: Guided consolidation, Update manager, orchestrator have been around for some time now. vSphere takes this to a next level.
One of the major features of vSphere is Host Profiles.
9:45 Host Profiles attach configuration templates to ESX hosts. Hosts can be configured to a certain standard setting with a simple click, and can be monitored for changes. If a change on a single host breaks the template compliancy, Host Profiles can remediate this and fix it automatically.
9:47 vCenter will be shipped as a Windows binary like it is today, but also as a Linux Virtual Appliance.
9:47 Beta available today.
9:47 Applause from the audience. (Have I mentioned already there are less suits in the audience today?)
9:49 Shout-out to Twitter from Herrod.
9:49 Next major chapter: vCloud.
9:51 Standard APIs for VDC-OS management and federation. (Linking “clouds” across datacenters.)
vCloud manages security, network, storage and monitoring.
9:51 First example of federating clouds: Site Recovery Manager
9:51 Storage is replicated on array-base, vCenter is replicated via IP.
9:53 VMware tries to figure out “long distance VMotion” for live migration. Today this requires very exotic setups and the right network environments.
Challenges for long distance VMotion:
Moving the memory
Moving the disk images
maintaining the network connections.
This is an ecosystem challenge (storage replication, dedup, wan optimization, stretching VLANs, …). Lots of work being done by the partner community and VMware.
9:54 Some long distance VMotion applications: DRS accross DCs, datacenter maintenance/move, DC disaster avoidance (eg. when a hurricane is on its way) and “follow the sun” environments.
9:55 All the cloud providers are nowadays creating their own self-service portals. VMware will offer their own base portal, based on Lab Manager.
9:56 Next demo: vCenter vCloud plug-in to manage clouds.
9:58 Bruce adds credentials of external cloud provider (Telefonica in this case).
The capacity offered by the cloud provider becomes available in vCenter.
Applications can be dragged from internal clouds to the external provider, just like between internal esx clusters.
9:58 Personal note: wow!
10:00 VMware wants to open up the vCloud API, to enable a rich ecosystem of clouds.
10:00 I wonder what Amazon thinks about this…
10:00 Last chapter: vClient initiative.
10:01 Another recap of the “follow the user” model/vision.
10:01 Walking through some View features: View Composer (linked clones, central patching), pushing full VMs to “thick clients” to run on top of client-side hypervisor.
10:02 Patching is made easier by using shared base disks, and application virtualization (ThinApp, formerly known as Thinstall).
10:02 Centralized policy management, based on ACE.
10:03 Major focus is the best user experience for all environments. Wan/Lan: PCoIP (PC over IP), local: rich portable desktop.
10:04 Jerry Chen from the View team is going to demo this.
10:04 The WAN technology is based on a co-development with Teradici, supporting CAD/CAM and 3D over WAN through hardware-assisted optimization.
10:06 LAN use case: traditional VDI: Thin clients, high-bandwidth, true pc experience (HD video, multimonitor, …)
10:09 Demo: thin client connected to rack workstation (a standard workstation blade, in other words, no virtualization). Demo with Google Earth working fluidly over a LAN connection.
10:10 vClient summary: Best user experience, central management, partnership with Intel for client-side virtualization (offline VDI).
10:11 “One more thing”: evolution of the mobile phone
10:12 VMware acquired Trango Virtual Processors, maker of mobile phone hypervisor.
10:12 Mobile phones bring familiar challenges: security/manageability, home/work life convergence, persona management, third-party applications.
10:13 VMware’s Mobile Virtualizaton Platform is a hypervisor for ARM-based devices.
10:15 Live demo with a Nokia N800. (If I’m not mistaken this internet device doesn’t include a phone part, but let’s not bicker about that).
10:15 Demo: downloading Windows CE, running productivity apps (Solitaire).
10:16 Another VM is downloaded and running side by side: Android.
10:18 And with that, the keynote is finished.
10:18 Thanks for following!
Leave a Reply