April 2015 archive

Join Infinio and the mad chief scientist from Deep Storage himself, Howard Marks for a webinar on May 7th at 10:00am PST to find out how using compute server resources to accelerate storage access can bring greater levels of storage performance and flexibility to the modern data center while simultaneously helping to keep costs under control.

The always entertaining Howard Marks has decades of knowledge and experience with storage so be sure to give this one a watch to learn more about server-side storage accelerator technology. He also has a rather unique method of simulating a storage failure which you can find all about here.

On a VMFS volume a VM is comprised of many files that all have different purposes, the configuration file for a VM (.vmx) is a small text file (<100KB) that contains various information about the virtual hardware configuration of the VM. With the Virtual Volumes (VVols) architecture the configuration information for a VM is stored in a different manner. Before we go there though lets take a look at how a VM is stored on VVols versus VMFS, the below picture illustrates the files that make up a VM on VMFS.

With VVols a VM is still comprised of different object types just like VMFS but many of them have been consolidated so there are fewer objects. A VM on VVols is made up of a minimum of two files, a Config VVol and a Data VVol, once a VM is powered on a Swap VVol is created to support memory over-commitment on a host which is then deleted when a VM is powered off just like with VMFS. There are additional VVols created for each snapshot that is taken of a VM as well as any memory states included with snapshots. So where did all those other files go that are part of a VM on VMFS such as log files, descriptor files, lock files, etc? They all get packing into the Config VVol.

While a config file on VMFS is quite small, with VVols it’s much larger as it’s allocated with 4GB of space by default. It’s allocated thin though so the actual size will be much smaller but it has the potential to grow as the data within it grows. Below is a screen shot from a 3PAR array that shows the Config VVol, you can see that is is provisioned thin (tpvv) and the actual physical space it consumes is around 2GB (1996MB). Also note the Swap file is fully allocated, this is necessary as the file is used for virtual RAM and it will be equal in size to the memory that is assigned to a VM (minus any memory reservations).

So what’s all in the Config VVol, the virtual hardware configuration of a VM of course but it also contains descriptor files for virtual disk and snapshots. With VMFS you always had 2 files that were created for each virtual disk, a small descriptor file and the larger data file, with VVols the Data VVol is the virtual disk and the descriptor is in the Config VVol. You also have all those individual .log files that exist for each VM on VMFS all combined into the Config VVol as well as any lock files that get created.

So there you have it, I don’t know if it’s possible to reclaim space inside a Config VVol like you can within the Data VVol using UNMAP. The max size of a Config VVol is 4GB, as most of the files are small, they probably truncate and remove log files as needed to ensure that it doesn’t exceed that size. Also all of the data inside a Config VVol is mostly text based data, I’m not sure how it’s structured inside the VVol container I’ll have to see if I can get a look inside one to find out.

In today’s virtual world where disk-based media rules and everything is interconnected via the internet there is a very high probability that you’ll need replication for BC/DR to get your valuable VMs protected off-site to another location. Fortunately Zerto makes that simple by offering a virtual-aware, software-only, tier-one, enterprise-class replication solution purpose-built for virtual environments. Unlike traditional storage array replication which requires like brand and model storage arrays at each site to be able to replicate with each other, hypervisor based replication is storage agnostic meaning it operates above the physical storage layer at the virtual layer so it supports any type of underlying storage on each side. Another advantage of using HBR is that it allows you to replicate at the more granular VM level instead of the larger scale LUN level.

It works by using only two components, the Zerto Virtual Manager (ZVM) which is a Windows-based service that acts as the management console and can be accessed via a plug-in to vCenter, the ZVM manages all vSphere replication and keeps track of applications and information in motion in real time. The other component is the Virtual Replication Appliance (VRA) which is deployed as a virtual appliance (VM) on each host and continuously replicates data from user-selected VMs, compressing and sending that data to the remote site. The following diagram shows how Zerto Virtual Replication is deployed across sites as a DR solution:

Zerto does one thing, replication and they do that well, I’ve always admired them and thought that had a great solution. They were founded in 2010 and their product debuted shortly after that and won the Gold in the BC/DR category for the Best of VMworld awards in 2011. It also won Best of Show which means is was chosen as the best product across all the many awards categories. I noticed a quote from Jo Maitland while on Zerto’s website that essentially stated the decision was a no-brainer. As a former Best of VMworld judge myself I know that feeling well as I was involved in a similar situation in a prior year with the product I had picked as the winner in the Security category, HyTrust which also easily won Best of Show.

So if you are in the market for a great BC/DR solution, I highly encourage you to go checkout their replication solution for VMware vSphere environments. One thing to get you started is this great overview and demo video. They have a lot of great and smart folks that work at Zerto that you should follow on Twitter as well, and be sure and read this great post entitled “I am not a booth babe, Ask me a question” by Fara Hain, Zerto’s Director of Marketing.

Key Zerto folks to follow on Twitter:

• Jennifer Gill
• Fara Hain
• Shannon Snowden
• Zerto Corp

I’ve started shipping the Top vBlog 2015 coins that were custom made this year courtesy of Infinio to the Top 50 bloggers. I’m sending them out in small batches due to all the work packing them, printing labels, filling out customs forms, etc. When you receive yours you might find a something little extra in the package courtesy of Nutanix who is providing a coin holder and lanyard so you can proudly wear your coin around your neck. Just imagine how cool you would look wearing it around your house, at the grocery store, in your office or even at VMworld.

So when you receive yours in the mail, give a shout out to Nutanix and Infinio and post a picture of yourself wearing it on Twitter with the hash tag #TopvBlog2015. Bonus points for whoever has the most creative photo.

Scott Davis from Infinio who was a special guest on the Top vBlog 2015 results show is presenting a webinar for technology geeks tomorrow, that sounds pretty cool so go sign up now! Also speaking of cool, Infinio was recently named by Gartner as one of the Cool Vendors in Storage Technologies for 2015, find out what Garnter had to say about Infinio here.

A new wave of disruptive innovations in the storage industry is shaking up familiar architectures and technologies. These innovations promise to drive down costs, simplify storage in the data center and challenge the long-established vendors and data center status quo. Hybrid Arrays are the new normal and All-Flash Arrays and Hyper-Converged Infrastructure solutions are serious contenders in the appropriate use cases.

This webinar will cover disruptive storage technologies and the solutions they are driving. You will learn about:

• Core storage technology innovations for performance (such as PCI-e and DIMM-based Flash/SSD, and NVMe technologies) and for capacity (such as SMR (shingled) drives)
• Mobile/cloud application influences on storage including a shift to scale-out architectures, object & cloud storage trends
• Storage product architectures such as All-flash and Hybrid arrays, Hyper-converged Infrastructure, Software-defined storage & I/O optimization technologies

As I’ve been pretty involved with 3PAR support for VVols during my day job at HP, my VMware storage marketing counterpart contacted me and wanted me to write a guest post on their storage blog on implementing VVols on 3PAR. So if you are interested in that head on over there and give it a read. Thanks to Mauricio Barra and Ken Werneburg for the opportunity.