VMware

Introducing VMware EVO:RAIL a new hyper-converged offering!

Duncan Epping · Aug 25, 2014 ·

About 18 months ago I was asked to be part of a very small team to build a prototype. Back then it was one developer (Dave Shanley) who did all the development including the user experience aspect. I worked on architectural aspects we had an executive sponsor (Mornay van der Walt). After a couple of months we had something to show internally. In March of 2013, after showing the first prototype, we received the green light and the team expanded quickly. A small team within the SDDC Division’s Emerging Solutions Group was tasked with building something completely new, to enter a market where VMware had never gone before, to do something that would surprise many. The team was given the freedom to operate somewhat like a startup within VMware; run fast and hard, prototype, iterate, pivot when needed, with the goal of delivering a game changing product by VMworld 2014. Today I have the pleasure to introduce this project to the world: VMware EVO:RAIL™.

EVO:RAIL – What’s in a name?
EVO represents a new family of ‘Evolutionary’ Hyper-Converged Infrastructure offerings from VMware. RAIL represents the first product within the EVO family that will ship during the second half of 2014. More on the meaning of RAIL towards the end of the post.

The Speculation is finally over!

Over the past 6-plus months there was a lot of speculation over Project Mystic and Project MARVIN. I’ve been wanting to write about this for so long now, but unfortunately couldn’t talk about it. The speculation is finally over with the announcement of EVO:RAIL, and you can expect multiple articles on this topic here in the upcoming weeks! So just to be clear: MARVIN = Mystic = EVO:RAIL

What is EVO:RAIL?

Simply put, EVO:RAIL is a Hyper-Converged Infrastructure Appliance (HCIA) offering by VMware and of qualified EVO:RAIL partners that includes Dell, EMC, Fujitsu, Inspur, Net One Systems and SuperMicro. This impressive list of partners will ensure EVO:RAIL has a global market reach from day one, as well as the assurance of world class customer support and services these partners are capable of providing. For those who are not familiar with hyper-converged infrastructure offerings: it combines Compute, Network and Storage resources into a single unit of deployment. In the case of EVO:RAIL this is a 2U unit which contains 4 independent physical nodes.

But why a different type of hardware? platform What will EVO:RAIL bring to you as a customer? In my opinion EVO:RAIL has a several major advantages over traditional infrastructure:

Software-Defined
Simplicity
Highly Resilient
Customer Choice

Software-Defined

EVO:RAIL is a scalable Software-Defined Data Center (SDDC) building block that delivers compute, networking, storage, and management to empower private/hybrid-cloud, end-user computing, test/dev, remote and branch office environment, and small virtual private cloud. EVO:RAIL builds on proven technology of VMware vSphere®, vCenter Server™, and VMware Virtual SAN™, EVO:RAIL delivers the first hyper-converged infrastructure appliance 100% powered by VMware software.

Simplicity Transformed

EVO: RAIL enables time to value to first VM in minutes once the appliance is racked, cabled and powered on.. VM creation, radically simplified via the EVO:RAIL management user interface, easy VM deployment, one-click non-disruptive patch and upgrades, simplified management and scale-out.

Highly Resilient by Design

Resilient appliance design starting with four independent hosts and a distributed Virtual SAN datastore ensures zero VM downtime during planned maintenance or during disk, network, or host failures.

Customer Choice

EVO:RAIL is delivered as a complete appliance solution with hardware, software, and support through qualified EVO:RAIL partners; customers choose an EVO:RAIL appliance from their preferred EVO:RAIL partner. This means a single point of contact to buy new equipment (single SKU includes all components), and a single point of contact for support.

So what will each appliance provide you with in terms of hardware resources? Each EVO:RAIL appliance has four independent nodes with dedicated computer, network, and storage resources and dual, redundant power supplies.

Each of the four EVO:RAIL nodes have (at a minimum):

Two Intel E5-2620 v2 six-core CPUs
192GB of memory
One SLC SATADOM or SAS HDD as the ESXi™ boot device
Three SAS 10K RPM 1.2TB HDD for the VMware Virtual SAN™ datastore
One 400GB MLC enterprise-grade SSD for read/write cache
One Virtual SAN-certified pass-through disk controller
Two 10GbE NIC ports (configured for either 10GBase-T or SFP+ connections)
One 1GbE IPMI port for remote (out-of-band) management

All of this leads to a total combined of at least 100GHz CPU resources, 768GB of memory resources, 14.4TB of storage capacity and 1.6TB of flash capacity used by Virtual SAN for storage acceleration services. Never seen one of these boxes before? Well this is what they tend to look like, in this example you see a SuperMicro Twin configuration. As you can see from the rear view, 4 individual nodes with 2 power supplies and in the front you see all the disks which are connected per group of 6 to each of the nodes!

For those of you who read this far and are still wondering why RAIL, and is it an acronym, the short answer is “No, not an acronym”. The RAIL in EVO:RAIL simply represent the ‘rail mount’ attached to 2U/4-node server platform that allows it to slide easily into a datacenter rack. One RAIL for one EVO:RAIL HCIA, which represents the smallest unit measure with respect to compute, network, storage and management within the EVO product family.

By now you are probably all anxious to know what EVO:RAIL looks like. Before I show you, one more thing to know about EVO:RAIL… the user interface uses HTML-5! So it works on any device, nice right!

If you prefer a video over screenshots, make sure to visit the EVO:RAIL product page on vmware.com!

What do we do to get it up and running? First of all rack the appliance, cable it up and power it on! Next, hit up the management interface on https://<ip-address>:7443

Next you start entering the details of your environment, look the following screenshot to get an idea around how easy it is! You can even define your own naming scheme and it will automatically apply that to joining hosts (both the current set, and any additional appliance added in the future)

Besides a naming scheme, EVO:RAIL allows you to configure the following:

IP addresses for Management, vMotion, Virtual SAN (by specifying a pool per traffic type, see screenshot below)
vCenter Server and ESXi passwords
Globals like: Time Zone, NTP Servers, DNS Servers, Centralized Logging (or configure Log Insight), Proxy

Believe me when I say that it does not get easier then this. Specify your IP ranges and globals once and never think about it any more.

When you are done EVO:RAIL will validate the configuration for you and then when you are ready apply it. Along the way it will indicate the stage and provide an indication of how far it is in terms of configuration.

When it is done it will point you to the management interface and from there you can start deploying workloads. Just to be clear, the EVO:RAIL interface is a simplified interface. If for any reason at all you feel the interface does not bring you the functionality required you can switch to the vSphere Web Client as that is fully supported!

The interface will allow you to manage virtual machines in an easy way. It has pre-defined virtual machine sizes (small / medium / large) and even security profiles that can be applied to the virtual machine configuration!

Of course, EVO:RAIL provides you monitoring capabilities, in the same easy fashion as everything else. Simple overview of what is there, what the usage is and what the state is.

With that I think it is time to conclude this already lengthy blog post. I will however follow up on this shortly with a series that looks a bit more in-depth to some of the details around EVO:RAIL with a couple core team members. I think it is fair to say that EVO:RAIL is an exciting development in the space of datacenter infrastructure and more specifically in the world of hyper-convergence! If you are at VMworld and want to know more visit the EVO:RAIL booth, the EVO:RAIL pavilion or one of the following sessions:

Software-Defined Data Center through Hyper-Converged Infrastructure (SDDC4245) Monday 25th August, 2:00 PM – 3:00 PM – Moscone South, Gateway 103 with Mornay van der Walt and Chris Wolf
VMware and Hyper-Converged Infrastructure (SDDC2095) Monday 25th of August, 4:00 PM – 5:00 PM – Moscone West, Room 3001 with Bryan Evans
VMware EVO:RAIL Technical Deepdive (SDDC1337) Tuesday 26th of August – 11:00 AM – Marriott, Yerba Buena Level, Salon 1 with Dave Shanley and Duncan Epping
VMware Customers Share Experiences and Requirements for Hyper-Converged (SDDC1818) – Tuesday 26th of August, 12:30 PM – 1:30 PM – Moscone West, Room 3014 with Bryan Evans and Michael McDonough
VMware EVO:RAIL Technical Deepdive (SDDC1337) Wednesday 27th of August – 11:30 AM – Marriott, Yerba Buena Level, Salon 7 with Dave Shanley and Duncan Epping

VMware / ecosystem / industry news flash… part 1

Duncan Epping · Aug 21, 2014 ·

I did these startup news flash newsletters for a long time and it kind of died down. Reason for it being that it felt like I missed out on updates of companies who were way beyond the stage of being a startup, which is a shame. I do feel there is a need, or there is a market for these types of blog articles. So lets give it another try but this time with a different name and a slightly different angle. In this VMware / ecosystem / industry news flash I will feature interesting (to me) news items of VMware and its ecosystem. This could range from Nexenta releasing a new version of their software to an acquisition or new round of funding. I hope you will find it useful. Besides a pointer to the originating article, I will also provide some of my thoughts, and I would like to encourage you to do the same via the comment section!

VMware announced the intention to acquire CloudVolumes
I looked in to CloudVolumes about 5 months ago. I was very intrigued with what they developed and how they go about layering applications on top of virtual machines both for the server and desktop space. Especially when it comes to updating applications I was impressed how seamless and flexible CloudVolumes was. What stood out most was that they simply connect a VMDK to a VM instead of streaming data across the network… it was so simple, but yet so innovative and smart at the same time that I am not surprised VMware acquired them.
Ubuntu Now Available on VMware vCloud Air
I guess there are two things here… vCloud Air formerly known as vCloud Hybrid Services, a new name. And Ubuntu being available in the catalog by default which is great. If I have to pick any Linux distribution to work with it typically is Ubuntu, and as it is one of the most used distributions out there in cloud environments this makes sense to me.
Nexenta Enables File Services on VMware VSAN
I did not see this one coming to be honest and I am not sure what it will look like from an implementation perspective. It is interesting though to see the ecosystem layering services on top of VSAN and I am wondering what else is in the pipeline!
DataGravity out of stealth… (Steve Foskett opinion, Howard Marks thoughts, Robin Harris)
I have read multiple articles on DataGravity and it seems they are playing in the same space as Nimble and Tintri with their hybrid storage solution. But what DataGravity adds on top of that in-sight in to the data being stored. It sounds interesting but I do wonder what the target market is as I would suspect that many who need this type of data analytics already have layers on top of their storage systems which will provide this detail for them. I will try to set some time aside at VMworld to chat with them.
Pernix Data announced a new round of funding (Business Insider, Venture Beat)
35 million dollar C round of funding is significant I would say. On top the growth numbers that Pernix published are impressive, but not surprisingly if you look at what they have to offer and still have in the pipeline.
Atlantis announced USX 2.0 (vClouds.nl, Andrea Mauro)
I have always been intrigued by USX. I played around with 1.0 on top of Virtual SAN and was impressed by the capabilities added and the efficiency of their data services, however at the same time I also realized that layering USX on top of Virtual SAN kind of diminished the Virtual SAN policy based management capabilities. However with USX 2.0 it appears that USX will support VVOL and as such SPBM to a full extent, and hopefully also will require a lower number of USX related virtual appliance to be deployed. Will try to explore USX 2.0 in the near future.

Great read: Virtualizing SQL Server with VMware: doing IT right

Duncan Epping · Aug 4, 2014 ·

A couple of months back I received an email from Jeff Szastak. He asked me if I wanted to write a foreword for the book he was authoring with Michael Corey an Michael Webster. When he mentioned the book was on the topic of virtualizing SQL Server I knew it would be an instant classic, and I did not hesitate one second and said I would be honoured to write the foreword. All three have a wealth of knowledge and experience when it comes to virtualizing SQL Server (and tier 1 apps in general) and my expectations were set extremely high.

I can honestly say that they did not disappoint me, even with the extremely high expectations I had. The book is indeed an instant classic, and one which I wished I had when I went through the exercise of virtualizing SQL myself… Don’t get me wrong, this book is not just a must read for people looking to virtualize SQL Server, I would argue that it is a must read for everyone who does something with virtualization. There are countless of tips, tricks and design considerations in this book making it invaluable to every virtualization admin. Michael, Michael and Jeff have done a tremendous job if you ask me and the book is definitely high up on my favourite VMware related books list!

By virtualizing business-critical databases, enterprises can drive far more value from existing IT infrastructure. But squeezing maximum performance out of a virtualized database instance is an art as much as a science. This indispensable start-to-finish guide brings together all the techniques, tips, and insights you need to succeed.

Drawing on unsurpassed personal experience, three leading experts share complete best practices for deploying business-critical database servers in virtualized vSphere 5 environments. They cover the entire project lifecycle, bridging technical and communications gaps between SQL Server and VMware professionals that often make database virtualization more difficult than it needs to be.

You’ll find specific guidance for architects and administrators responsible for systems, storage, databases, applications, or VMware virtualization. The authors also present detailed, start-to-finish coverage of performance baselining and testing: all you need to make your virtualized databases as fast as they are cost effective. Although this book focuses on SQL, the authors’ proven guidance for enhancing performance can be leveraged by any IT professional virtualizing a demanding Tier 1 application.

Just pick it up!

vSphere 5.5 and disk limits (mClock scheduler caveat)

Duncan Epping · Jul 15, 2014 ·

I mentioned the new disk IO scheduler in vSphere 5.5 yesterday. When discussing this new disk IO scheduler one thing that was brought to my attention is a caveat around disk limits. Lets get started by saying that disk limits are a function of the host local disk scheduler and not, I repeat, not Storage IO Control. This is an often made mistake by many.

Now, when setting a limit on a virtual disk you define a limit in IOPS. The IOPS specified is the maximum number of IOPS the virtual machine can drive. The caveat is is as follows: IOPS takes the IO size in to account. (It does this as a 64KB IO has a different cost than a 4KB IO.) The calculation is in multiples of 32KB. Note that if you do a 4KB IO it is counted as one IO, however if you do a 64KB IO it is counted as two IOs. Any IO larger than 32KB will be 2 IOs at a minimum as it is rounded up. In other words, a 40KB IO would be 2 IOs and not 1.25 IOs. This also implies that there could be an unexpected result when you have an application doing relatively large blocksize IOs. If you set a limit of 100 IOPS but your app is doing 64KB IOs than you will see your VM being limited to 50 IOPS as each 64KB IO will count as 2 IOs instead of 1. So the formula here is: ceil(IO Size / 32).

I think that is useful to know when you are limiting your virtual machines. Especially cause this is a change in behaviour compared to vSphere 5.1.

New disk IO scheduler used in vSphere 5.5

Duncan Epping · Jul 14, 2014 ·

When 5.1 was released I noticed the mention of “mClock” in the advanced settings of a vSphere host. I tried enabling it but failed miserably. A couple of weeks back I noticed the same advanced setting again, but this time also noticed it was enabled. So what is this mClock thingie? Well mClock is the new disk IO scheduler used in vSphere 5.5. There isn’t much detail on mClock by itself other than an academic paper by Ajay Gulati.

The paper describes in-depth why mClock was designed / developed, it primarily was to provide a better IO scheduling mechanism that allows for limits, shares and yes also reservations. The paper also describes some interesting details around how different IO sizes and latency is taken in to account. I recommend anyone who likes reading brain hurting material to take a look at it. I am also digging internally for some more human readable material, If I find out more I will let you guys know!