Don’t worry, this isn’t going to be a post explaining how some new buzzword is either complete rubbish marketing or the next great thing. I was having a discussion with a co-worker around the latest term floating around our industry- software-defined datacenters (SDDC). We were debating whether SDDC was the “future” of virtualization, how it fits with cloud, et cetera. I think some good came from those discussions so I figured why not share them here.

Let’s Start with Cloud

Not to beat a dead horse, but let’s talk about cloud first. I still find Simon Wardley’s talk from OSCON ’09 to still be incredibly relevant here. I agree with his premise that Cloud is mainly talking about an technology operation model, not a technology. Yes, recent technology advances have made this model more feasible, but it’s still just a way of doing things.

So what about IT as a Service (ITaaS)?

To me, (whatever) as a service in the IT context is the overlaid business model. ITaaS has existed for quite some time and some companies have adopted it years ago; IT acts as a service provider to the business and charges the business cost centers for resources consumed. Now its only been recently that technology like virtualization has allowed IT to share more resources and charge business units back for much more granular consumption. So ITaaS is not even an operational model, it’s the business/financial model that sits on top. It’s the IT department taking the IT service provider model and using it internal to the company.

ITaaS leverages Cloud

As i mentioned, companies have been doing (or at least trying) the ITaaS model long before the cloud concept was feasible. What cloud brings the the party is a model which allows providers to deliver service faster, cheaper, and with a better customer experience. This applies to both service providers as well as IT departments working in the ITaaS model.

What about software defined datacenters (SDDC)? Does that replace cloud?

This is where we actually start talking about actual technology. There are two similar but slightly different takes on SDDC:

In the first, the promise of SDDC is that current manageability limitations of physical hardware like network switches, firewalls, and storage arrays can be overcome by separating out the control plane from the data plane. The idea being that moving the control/management to a virtual layer can allow for simpler, more agile operations.

Nicira (the recent VMware acquisition) is a great example of this concept- software defined networking (SDN). It has/had a great fit with very large organizations and service providers that have very complex network topologies that usually include hardware from multiple vendors.  SDN allows all the ‘smart’ networking to happen completely in software switches and nodes and allows the underlying hardware to be the lowest common denominator and focus on pushing packets with little ongoing reconfiguration. This makes it even easier for orchestration/automation tools to configure networking on the fly via APIs.

The second SSDC methodology is one that has been around a bit longer- the vAppliance. The concept is that both the control and data planes go through a virtual version of what was traditionally physical hardware. This is where VMware vShield Edge firewalls, Avamar Virtual Edition, NetApp’s ONTAP Edge virtual storage appliances, and others fit. I can provide multitenant support by just spinning up an additional copy of that vApp for the new tenant.

Obviously under both of these models, the underlying hardware requires less intelligence. The big difference I see is that the first option fits much larger enterprise and service provider environments where the second is tuned for small-medium business (SMB). The net outcome for companies that are trying to operate in a cloud model, SDDC can make that easier by limited touch points for their orchestration/automatic framework of choice.

In Summary

The way I see it:

ITaaS->Cloud->SDDC

is

Business->Operations->Technical.
Thoughts?

That said, I don’t expect you to vote for it just based on my winning personality, so here is a quick rundown on the session:

Summary

There are many common misconceptions around storage networking for both Fibrechannel and IP. These misconceptions lead to bad design, and these bad designs lead to poor performance and lower availability. We will review real-world examples of these designs and discuss what the architect intended and what the actual outcome was. We will educate you on all the advantages and limitations of different configurations and arm you with the knowledge so you can make a well-informed design that works best for your environment, including protocols, multipathing, link aggregation, and more.

Three Key Takeaways

• More wires don’t always mean more performance- some of the accepted ‘truths’ about storage networking are pure fiction.
• There is no ‘right’ storage protocol for vSphere- they all have their pluses and minuses. Get educated on all of them and select the one that makes the most sense for your environment
• Storage connectivity for vSphere is not a black art- with a good understanding of how it all works, it can be easy for any VMware or storage administrator.

 Conclusion

All the pieces are there to build a scalable and reliable storage network, no matter what your protocol is. The key is understanding how they work and what are sound design principles to follow.

I frequently hear this about server platforms, but it is also sometimes said about x86 storage hardware; meaning that “it’s all the same.” It seems to make sense, and most people nod their head knowingly when they hear it.

The problem is that it is completely false.

What does it mean to be commoditized?

To clarify here, the words commodity/commoditized are misused in this instance. Commodity just means “something of use or value.” The term fungible is really what is meant; replaceable or interchangeable- like currency or crude oil. The best hallmark of a fungible good is price is usually the sole determiner of demand. That means absolutely zero differentiation between goods. Let’s test this definition on servers and storage.

Commodity Servers

Let’s start with servers- Are there server vendor’s who’s entire value proposition is based around acquisition cost? Sure- Super Micro is the first that comes to mind and Dell built an empire on “good enough and cheaper.”

The problem is that there are still vendors innovating in this space, and capturing sales with their innovations. HP is still the leading x86 server vendor overall as well as x86 blades. If Proliant servers are really the same as PowerEdge servers, why isn’t the lower cost option from Dell dominating the market? Cisco really sinks the server commodity theory. If x86 servers were truly commodity, would a new vendor entering the market be able to capture such large amounts of marketshare in a short period of time unless they were significantly cheaper (which by all accounts UCS is not)? Absolutely not.

Commodity Storage Hardware

Clearly servers haven’t been completely commoditized yet, what about storage hardware? Leaving tiny niche players out and focusing on mainstream storage vendors, there are only really two that use what are basically off the shelf servers as their hardware- Dell Compellent and HP Lefthand. And if you take the Dell/HP names out from in front of them, they are really ignorable niche products too.

What about EMC? Don’t they use industry standard x86 components in their storage platforms?

Now we’re starting to hit the mark- the key phrase in that question is industry standard. This is really what x86 servers and the components that make them up are- industry standard. Well understood and lower cost due to economies of scale. They allow server and storage vendors to stop wasting development and engineering cycles on custom ASIC and CPU designs, and focus on packaging these industry standard components in new ways to add value to their products.

Chad Sakac has a great video on Virtual Geek where he talks with Bill DePatie from EMC’s hardware engineering team. As Bill demonstrates the hardware for Chad, count how many times you think “hey that looks just like that supermicro server in my home lab.”

Conclusion

Are there parts of the x86 market that are a sea of clones that complete solely on price? Of course. But to say it’s all commodity does a great disservice to the hardware engineering teams at HP, Cisco, EMC, and others.

The vExpert program is a recognition to the contributions to the VMware, virtualization and cloud computing communities.

Like VMware , but for storage!

I read a note today from one of EMC’s competitors that made me chuckle. It spoke of a “paradigm shift in storage the same way VMware change the x86 server

market.” Now this isn’t the first time I’ve heard a storage company use the tired “we’re like VMware, but for storage!” slogan, but this time it got me thinking.

What this company is really announcing is ”we’re like ESX2.5, but for storage!” (Not as catchy.) It is the ability to move storage workloads between discrete storage controllers non-disruptively. Manually, no DRS or HA. And any one workload is still bound to the resources in the controller that owns it. A nice feature but not exactly a revolutionary paradigm shift.

Back to the Future

Let me paint you another picture. Some time in the future, VMware announces a new version of vSphere. This new version allows users to truly treat server resources like a pool- you have a total amount of CPU cores, memory, and NICs added up for the whole cluster and you can assign as much or as little as you need. Not limited to the resources of a single vSphere host for a single workload. If I have 2.5TB of memory and 120 cores across my vSphere cluster, I can assign as much as I want to a VM and the vSphere nodes will use a high speed interconnect (like Infiniband or RapidIO) to share all the resources between vSphere hosts.

Sounds amazing, right? But what does VMAX have to do with this?

Well that’s pretty much how the Symmetrix VMAX and VMAXe work. Each engine, which contains two controllers (directors) are connected to a Virtual Matrix (hence the name VMAX) that connects all the directors together and allows sharing of resources. I can add more CPU (engines), Memory (cache), or NICs (FC Ports) to my workload (LUN) without an outage and without being constrained by the resources of a single director.

What about DRS & HA? VMAX Availability & Scalability

Because the resources are all shared and the resource control is so fine-grained, it also allows for unmatched fault tolerance and minimized performance degradation during a fault. In any dual controller architecture, a loss of one of the controllers means a 50% loss of potential performance during that fault. Your data may be available, but it might not be useable. In the VMAX world, workloads can be (and should be) assigned resources across many different directors in different engines. This way, any component failure, means just the loss of that component with minimal impact on performance without administrative intervention. This is even more critical as virtualization means continued consolidation. Which basket would you put all your proverbial eggs in?

This architecture also allows me to add additional resources to the VMAX on the fly- I can start with a single engine and add engines, disk, ports, and cache as I grow.

That’s all great, but what’s it going to cost me?

In my opinion, this is one of the biggest misconceptions around the Symmetrix for years. Yes, some configurations can cost a lot, but as you see by its marketshare plenty of customers see the value. There are also plenty of VMAX configurations that are cost-competitive with dual controller arrays, and depending on the use-case, the VMAX can have a better TCO.

Fully Automated Storage Tiering (FAST VP) has made a huge change here- customers are now buying much more SATA, much less FC, a few EFDs, and letting VMAX sort it out automatically. This has lead to even better TCO.

Enter the VMAXe

The VMAXe has changed this even further- it trims out some features not needed by as many customers (Mainframe support, MASSIVE scale), adds in features customers do want (RecoverPoint replication), and simplified operations (all virtually provisioned, install to first IO in 4 hours).

And of course it has all the VMware integrations of its big brother VMAX 

Now you can understand why the VMAX and VMAXe are seeing great growth in the storage industry today.

So if and when sometime in the future VMWare announces a new version of vSphere like I described, they can say “like VMAX, but for  server virtualization!”

I’m not going to summarize every session or every announcement, as there will be plenty who have or will. I’d like to just give you a few thoughts that stuck with me from Vegas:

EMC

Now, I might be a bit biased, but I am amazed how well EMC is firing on all cylinders with regards to VMware. I remember attending VMworld 2007 while working for an integrator and being shocked at EMC’s presence, or lack thereof. Fast forward a few years and the amount of integrations and just overall focus of EMC on VMware integration is incredible. Obviously the credit goes to the team at the top, but I really think their decision to invest in the vSpecialist organization has had a huge impact.

vSpecialists

I’ve now been on the team just a little bit over a year. I can very easily say that this is the best team I’ve ever been apart of. It’s incredible to be around so many amazingly talented people who also don’t take themselves too seriously and have lots of fun. I’ve been looking forward to VMworld, not just for the content but also to see people on the team that I don’t get to see on a regular basis. It is a testament to the management team that they hire such great people as well as foster this kind of culture. It’s a culture that get’s such great results because it encourages everyone to pitch in and help and at the same time to get outside their comfort zone. All I can say is it feels like home.

VMware Community

On the subject of culture- hat’s off to VMware for cultivating such a strong sense of community. Customers, Partners, and Manufacturers all coming together to learn and connect. It was great to see various competitors visiting each others booths and sessions to catch up and expand their knowledge. In the end, the entire ecosystem wins. If you don’t already, find and participate in your local VMUG (VMware Users Group).

Solutions Exchange

As usual, lots of great booths- all the ‘big guys’ down to the small start-ups. It was the second straight year of the EMC Hands on Lab booth which was extremely popular again. Hats off the many, many people involved that make it work. One thing in particular I wanted to point out was the appearance of the VMAX VSA, complements of Dave Robertson (@daverdfw). Check out his blog for more details.

Sessions

As usual, the sessions ranged from high-level to nitty-gritty details. I can’t wait for the media to be posted for the sessions that I missed that I really wanted to see. One of my favorite sessions I did see was on Storage DRS (SDRS). I was aware of the basics of how it would operate before the session, but came away completely impressed with the logic working in the background to decide what to move where. Especially the ability to predict the effect of moving VMDKs on the response time and IOPs of a particular datastore. Very impressive.

Party!

The Killers, well, killed it. A great choice by VMware, and their opening act Recycled Percussion put on a great show as well. I will say one thing- if you are 20 feet from the stage and want to break out your phone and snap a pic or two to tweet or whatever, fine. You don’t need to video the entire concert on your iPhone or bring a digital SLR on a monopod to get 156,534 pictures. Try and actually, i dunno, enjoy the concert?

What’s Next?

On the flight home, I placed an order with NewEgg to update my home lab, and the parts should arrive tomorrow. While there’s been a billion posts on home lab hardware & experiences, I figure “What the hell. One more.” I plan on using the lab to update my VCP to 5, training for the VCAP, and expanding my vFabric knowledge.  Stay tuned.

This phrase has been showing up a lot more since the most recent AWS outage. It’s a pretty simple concept- your design should expect component failure and be able to deal with it. It’s something enterprise organizations have done for years in their infrastructure. The key word there (and why this is getting a lot of press now) is infrastructure. RAID, clustering, redundant networks, load balancers, etc. have been used by service providers and internal IT for years to avoid single points of failure within their infrastructure design, and provide availability for their applications. The applications (for the most part) have not been designed for failure.

Some applications have evolved to add some levels of increased availability (Oracle ADG, Microsoft Exchange DAGs, etc) and some new applications are being written from the ground up to handle all sorts of component failures. A great example (and has been touted frequently) is the Netflix design that leverages the famed chaos monkey.

The problem is, this is the exception not the rule. Most organizations have tons of apps from tons of vendors, and I’m sure some still contain legacy code from punch-card days. It’s easier to say “rewrite your apps for the cloud” than it is to actually do it. Even if most of your apps have some level of availability baked in, each one has its own methodology which can be a management nightmare.

This has been one of the driving forces behind virtualization of legacy applications beyond simple consolidation; leveraging the increased availability provided by solutions like VMware HA, DRS, FT, SRM, etc.  This way your creaky old app written with a foxpro back-end can get higher levels of availability. VMware (and it’s customers) have been highly successful with this methodology- designing for failure in the infrastructure is cheaper & quicker than rewriting old apps or migrating to new ones.

Commoditization

This dovetails nicely with the previous discussion around hardware commoditization and value added software layers. Hardware is commodity, but your ‘availability/designed for failure’ secret sauce isn’t. If you’re Netflix or Google and can put that secret sauce in your software, sure you can use whatever white-box hardware you can buy by the pallet and be fine. But if you’re the average enterprise and your apps aren’t built that way, well then you’re going to need some layer between those dumb disks/cpus that provides that availability. Today those layers are called things like VMware vSphere, EMC Enginuity, VNX OE, and Cisco UCS Manager.

Oh look, another self-important (insert favorite profanity here), here to tell us what the AWS outage means. It’s either:
1. This proves public cloud doesn’t work
2. The effect of the outage shows how big cloud is
3. He’s here to tell us AWS customers how dumb we are and how if we were on his technology this would have never happened
4. OpenStack would have prevented the outage, cured cancer, and beat Kasperov at Chess

Actually, no. I think the outage customers of AWS are experiencing really exposes two key issues:

1. “Cloud” doesn’t magically abdicate your responsibilities as an IT department to BC/DR planning and testing. It can be a component of your strategy not your replacement for one.

2. Vendor lock-in is still Vendor lock-in, even in a shiny cloud wrapper.

The key to avoiding vendor lock-in is workload portability- something VMware brought to our own internal data centers. Without portability, IaaS/PaaS is just ASP2.0. This is where our friendly neighborhood open source evangelists jump in and yell “OpenStack!” I personally don’t see how taking a Harrison Bergeron approach to IT infrastructure is good for companies; it’s lowest-common-denominator IT. What is key and what vendors and customers should be pushing for is standard APIs and data formats to give us portability without giving up on the decades of intellectual capital that have been poured into things like storage, compute, and networking.

I agree that there is a march towards comoditization- you can see it in the storage industry. Most of the major players are using off the shelf hardware components, but they are innovating in the software layer. Platforms like Engenuity and VNX OE can add a ton of value through features like FAST VP. As long as they are addressable natively by your cloud management platform of choice, what’s the problem? Let customers/providers worry about building their stack with the components they deem most valuable, and lets spend our energy on making all those stacks able to talk to each other.

I think Microsoft is dead, they just don’t know it yet.

Well, not completely dead, just mostly dead. Now obviously I believe if they make some changes they can recover, but I’m not sure they will.

What killed Microsoft?

I think there are a few macro market trends and their lack of response to the pressure have hurt them significantly:
1. The consumerization of IT/Tech by Apple, Google, etc.
2. Virtualization has (and continues to) marginalize the role of the OS in the stack.
3. The rise of Firefox (and AJAX, SOAP, HTML5, etc.) as a force in the browser market, practically killing ActiveX and IE’s stranglehold(thank goodness!)

The biggest cultural problem I think Microsoft has that stops them from responding to these pressures and succeeding is one main thing:

They focus on the What instead of the Why.

Huh?
Look at every major product failure from Microsoft (the Zune, the Kin, Every windows tablet ever, etc.) and see what they all have in common, and probably still leaves Microsoft execs scratching their heads as to why they all failed miserably. They looked at the what, aka the features, and made a product that had those features.  I’m sure this discussion happened somewhere on Microsoft’s campus:

“Since the iPad released, tablets are hot again. We need a 10 inch tablet that’s thin with a long battery life. It needs two cameras too. And run Windows. It has to run Windows 7. Enterprises will love it! iPad is a toy for playing Angry Birds. Our new tablet will be the iPad for business!”
(BTW- I’m sure a similar conversation happened at RIM’s headquarters as well. That’s how we ended up with the nonsense that is the Playbook. RIM is a goner too.)

And that product will bomb, just like..every…windows…tablet…ever…sold. Why Apple keeps hitting home run after home run is Steve Job’s unrelenting focus on the end user experience, and how that is carried out by the design teams in Cupertino.

“How would a user use this? What would they expect it to do? What would they want it to do? What would they love it to do that they don’t know it’s even possible?”

Aka, the Why. If Microsoft really understood why people liked the iPod or why they like the iPad and asked some of these types of questions, they might actually come up with a product that people actually want. Jobs said he realized quickly two things that had killed tablets before- the stylus and the convertible tablet w/keyboard (aka touchscreen laptop). The user experience was terrible.

And that key question of “what do they want that they don’t know they want” is crucial. That’s why focus groups are a waste of time and usually make mediocre products. I guarantee Microsoft uses a lot of focus groups. No focus group would have described the iPad before it existed. Why? People in general aren’t very imaginative and think in terms of things they know.

I’m really interest to see what becomes of our friends in Redmond. Do they ever evolve? Or Do they end up as the Microsoft Office company?

So who am I and what is this blog for? My name is Tyler Britten and I am an IT infrastructure geek. I’m currently employed by EMC as a vSpecialist on the VMware Affinity Team. Previous to that I worked as technical consultant for a small infrastructure reseller, a network engineer for a Fortune 1000 company, and I also freelanced as an IT consultant for small businesses. I’ve also cleaned pools, directed traffic, been a short order cook, telemarketer, and made ice cream. I have certifications from EMC, Cisco, VMware, NetApp, Riverbed, Microsoft, CommVault, and I’m sure a few others.

I’m obviously fascinated by technology and I decided to start this blog (besides all the usual narcissistic reasons) to share not only my opinions, but also tips and tricks I discover along my travels with enterprise tech as well as consumer electronics. Of course, all opinions expressed are my own and do not represent my employer.

Hopefully you’ll find something interesting here. Of course feedback is always welcome.

Tyler