vSphere 5: The next level.

So the big news is out: vSphere 5 is here! With a lot of exciting new features, of which I will be posting several deepdives I’m sure.

For now, let’s just look at the new and exciting features vSphere 5 brings to the table.

VMware’s clear messaging

With the coming of vSphere, VMware has a very clear messaging around this new release which raises the bar once again:

  • Deploy Business Critical Applications with Confidence;
  • Respond to the Needs of the Business Faster with Cloud Agility;
  • Move to Cloud Computing with Trust.

Some great point to strive for I think. vSphere 5 will enable even bigger VMs to run. That is the ESX part of the game. Next, agility is on the agenda. Where competing hypervisors may deliver the same performance, VMware is aiming on their toolset to further improve their agility. Cloud computing is a very important step I think.

Some years ago, cloud used to be something like this:

“Looking at the cloud it is hairy, within the cloud it is misty, under the cloud it rains”

Thanks to recent developments, the cloud is rapidly becoming very real. I am more than happy that VMware is really pushing this technology. People who still think the statement above is true, please go and see your local EMC vSpecialist. The journey to the cloud is real. Very real. I encourage everyone to hop on!

New Technical coolness

There is just so much new in vSphere5! Check out some of the awesome new features you’ll find in vSphere 5:

32 vCPUs and 1TB per VM

Yes indeed, you now can configure a VM in vSphere5 with up to 32 vCPUs and up to 1TB of RAM. This is directly in line with VMware’s statement to deploy business critial appliactions.

Bye bye Service Console
ESX is dead. Long live ESXi! From the vSphere5 release, it will be ESXi only, which is now called the VMware vSphere Hypervisor. I have always loved the Service Console, but everyone has seen this one coming and as we move towards stateless installs (part of the VMware journey to the cloud!) the Service Console is no longer required.

Broadened support
There is a bigger base for hardware supported now. As a short proof: My new vSphere whiteboxes have their onboard NIC (Realtek 8111E) now working in vSphere5!

The VMFS filesystem has entered a new level: VMFS version 5. This version of the VMFS filesystem has one single blocksize, which is always 1Mbyte. You can actually upgrade existing VMFS filesystems to version 5, but in that case the blocksize of the “old” VMFS will remain. Also, an existing VMFS which is upgraded wil remain limited to 2TB-512bytes. But when you create a new VMFS5 filesystem from scratch, the filesystem will be GPT formatted and can be sized up to 64TB without extents! This is the first time you can also add physical raw device mappings (pRDMs) to VMs which are larger than 2TB. Unfortunately, VMDK files (virtual disks) still have the 2TB-512bytes limit.

vSphere storage appliance
A basic DAS-to-SAN built natively into vSphere. Wow! Can we get rid now of all those NAS and SAN boxes? Well, you might if you run a very small shop. Stated otherwise, very small implementations would normally not be able to afford a SAN-like hardware appliance. For these situations you might be able to use the vSphere storage appliance!

(Stateless) Auto Deploy
In an attempt to further automate tasks like adding and configuring nodes, VMware now introduces Auto Deploy. This feature is much like products like UDA. Auto Deploy runs from vCenter, and establishes PXE booting for vSphere nodes. Since all configuration resides in vCenter, you can now boot your servers from the network into vSphere, completely stateless!

VAAI enhancements
VAAI has been around already in vSphere 4. A short introduction may be in place. VAAI is not an API like many people seems to think. VAAI is basically a set of inline SCSI commands that VMware can try to shoot to storage arrays. These commands can offload certain storage-intensive workloads from the hypervisor down to the array. In vSphere 4 we have seen three implemented functions so far:

  1. VAAI offload for “write same” (eg. zero out blocks on disk for FT-enabled disks) for block;
  2. VAAI offload for “block copy” (eg. cloning, deploying templates or storage VMotions) for block;
  3. VAAI atomic test-and-set (metadata locking on a granular level) for block.
  4. vSphere 5 is going to add new VAAI features:

  5. Thin provision stun for block. This is a new feature. When you thin provision disks and space runs out, the VM used to be unable to write data and eventually crash. Now the VM can be stunned until space becomes available;
  6. Zero space reclaim for block. This will enable to notify the storage that it can mark blocks as empty, returning them to thin pools;
  7. VAAI is going NFS! For now Full Clone, Extended stats (finally better insight in space usage on NAS) and Space Reservation have become part of the VAAI spectrum.

Fault Domain Manager: VMware HA complete rewrite
Even though VMware HA in vSphere4 was a compelling feature, the underlying software was dated (actually an old EMC product called Legato) and some limitations applied. Now, the software that makes VMware HA tick has been completely rewritten to form FDM (Fault Domain Manager). There are no multiple master nodes anymore, but the nodes in a cluster perform an election for a master. If the master dies, the slaves will re-elect. Even better: When a cluster is split in two, the two halves will both elect a new master! This is a very important step into the safe construction of stretched clusters. To make things even more interesting: The new VMware HA is going to be able to leverage Ipv6 but also datastores to monitor each other as well. Another small improvement: HA has no longer any DNS dependency! Owwwww yes. Expect a deepdive post on this subject very soon!

Storage DRS
Most people will know VMware DRS. This feature in vSphere can perform initial placement or even realtime movement of VMs across vSphere nodes in order to balance the CPU and memory load between nodes in a cluster. Storage DRS is the logical next step: By looking at the datastores, storage DRS can perform initial placement for VMs, or even perform storage VMotions to balance the load on the different datastores!

VASA (vStorage API for Storage Awareness)
A feature already available in vSphere 4 is SIOC (Storage I/O Control). VMware measures the latency to the datastores, and is able to limit the numbers of I/O to that store if latency gets too wild. Features like this enable storage DRS (see above). But VMware realized that a lot of storage vendors are doing auto-tiering on their storage (for example EMC’s FAST-VP). It would be very useful to be able to gather information on this. This is what VASA was designed for. With VASA it becomes possible for VMware to understand how the underlying storage is configured. As time progresses I expect to see more and more details being delivered by storage arrays to VMware. The storage and the hypervisor fuses more and more!

Storage VMotion enhancements
In vSphere5, storage VMotions can be performed on VMs that have snapshots (and the snapshots will be migrated as well). Storage VMotion will also support linked clones. Storage VMotion now has another customer: Storage DRS (see above).

All in all, great stuff once again from VMware. Where other hypervisors are just maturing, VMware has taken a huge leap once again in a journey to a really cool world!

2 Responses to “vSphere 5: The next level.”

Soon to come