Good morning — reporting live from the last day of VMworld 2007, it’s the virtualization blogging crew. Follow along here, and go visit:
- Alessandro Perilli
- Scott Lowe
- Mike Laverick
- Viktor van den Berg
- Bouke Groenescheij
- Eric Sloof
- Bob Plankers
8:41 Mendel Rosenblum is in the house. "[People think virtualization is] almost like this magic pixie dust that you sprinkle on your data center. That’s not a very helpful description."
8:44 Virtualization is interposing a layer (adding a level of indirection) that acts the same as the old interface.
8:45 Showing the standard VMware boxes diagram — slide the virtualization layer between the hardware box and the App/OS boxes. OK, now what? Machines were underutilized, so let’s do server consolidation.
8:46 You have to be careful not to overload a node. This is where we introduced VMotion — if you get overloaded on a particular machine, you pick it up and move it. We’ve totally decoupled these virtual machines from the physical hardware. How do I map these VMs for better power efficiency, for better management?
8:48 ESX Server 3i makes things simpler — just drop the box in you data center and start it up. People initially seemed skeptical, but as I talk to people this year, everybody seems to agree — virtualization is a fundamentally better way of doing things.
8:50 New Hardware Compatibility Problem — now that we have these pools of resources, if they are mixed types of CPUs, how can we move these VMs around? Extended Migration (AMD) and FlexMigration (Intel) — the whole industry is getting behind it.
Mendel: Why don’t you add this capability so your CPU can lie about it’s capabilities? Chip maker: I’m adding all these new features to my CPU and you want it to lie and say it can’t do that? No way! But they quickly figured out why this was useful.
8:52 Storage VMotion — now why can’t we pick up the disk and move it around as well?
First demo: VC, running Oracle in a VM + Swingbench, let’s kick off the Storage VMotion. It’s a big disk, in progress…
Solves issues: need extra storage temporarily, need to rebalance storage, resource contention, suddenly needed faster or more capable storage, can move it from FC to iSCSI to a NAS device.
We see the VM now moved on a new data store with no downtime. (cue applause)
We’re moving towards the idea of a data center that manages the hardware itself.
8:57 Virtual Appliances. eg Browser appliance, CRM system-in-a-box
Virtual Appliance Marketplace – go to a website, click, download a virtual appliance. But… they’re very large.
8:59 Virtual Appliance "Instant On" — we can start the download, give the VMM the blocks it initially needs, and we can start the VM right away as the rest of the blocks are still downloading.
Second demo: At the Virtual Appliance Marketplace, downloading a 400MB VA.
- Laptop 1: start of the regular download
- Laptop 2: start the streaming download – download and install a small application — the stream manager.
- The first thing it does it download the prefetch — we prioritize the blocks within the VM to first download those that we need immediately. We do that by recording what happens in the VM when it boots and we download those first.
- Laptop 1: still downloading
- Laptop 2: It’s already booted and the browser is running
- Laptop 1: still downloading
Does it work with VDI? Yes
Software Delivery Models
- Traditional (client-side)
- Software-as-a-service
- Hybrid: VM managed on the server side and run locally on the client side — streaming helps enable this
9:05 High Continuous Availability. Machine dies, HA automatically restarts it before a human has to notice.
Last year we demo’s Record/Replay — record every instruction and interrupt, and then replay later. Now as a tech preview in Workstation.
Now Continuous Availability. Have two machines, synchronizing continuously via record/replay — when the primary goes down, the secondary machine keeps running and doesn’t skip an instruction.
Third Demo: An experimental ESX running an Exchange VM being exercised by a test program. Turn on Continuous Availability — now a second VM is mirroring every single action and state on the primary VM. When we pull the plug, the secondary machine will notice in about a second the secondary will take over with zero interruption. (cue *big* applause)
9:11 Looking to the future.
Solving Hard IT Problems with the virtualization layer — making them "checkbox simple." Now: cpu & memory management storage management disaster recovery, hardware fault tolerance
Can you guess what’s next?
Optimizing the hardware. Virtualization layer optimizes hardware usages (cpu, memory, storage, i/o) — key for hardware efficiency and power efficiency. We’re still pretty conservative, but as our algorithms get better, we will do more, especially with power usage in the data center.
Conclusions
We’ve only just begun to benefit. Small fraction of industry is virtualized — it’s so clear that this is a better way of doing things, the rest will follow quickly.
We have only scratched the surface of the valuable functionality that can be offered in the layer. Existing customers will see a steady stream of innovation.
9:17 Thank you very much
Karthik Rau — next year, VMworld Europe Cannes Feb 26
VMworld 2008 Sept 16 at the Venetian
Session and labs both powerpoint and audio sessions will be available starting tomorrow.
