CloudEnterprise.info

Posts Tagged ‘IaaS

With the recent changes in the leadership of one of Microsoft’s key business units – Server and Tools – from Bob Muglia to Satya Nadella one can’t help speculating what this means for the business unit and how it will affect Microsoft’s cloud strategy, specifically Windows Azure – Microsoft’s platform as a service.

Here’s my uneducated guess based on the assumption that given a new task humans tend to use the same approaches which worked well for them last time, and that Satya definitely got this post as a recognition for successfully rolling out Bing and transforming Microsoft’s search business from nothing to a competitor really frustrating Google.

Here’s what I think Satya will bring to Microsoft’s Server and Tools Business:

  • More focus on online (Azure) than on Windows Server: Bob Muglia made Windows Server business a success, this was his kid, while Windows Azure (one could argue) was kind of a step-child, imposed on him and added to his business during a re-org. Satya will likely feel much different: for last few years he has been “living in the cloud” leading Bing, and Steve Ballmer very explicitly made lack of cloud focus the reason for changing the business unit leadership.
  • Compete against the market leader: Bing clearly was developed to compete against Google. I guess this means that now Azure development will become aggressively anti-Amazon.
  • Acquisitions and partnerships: so far Azure has really been a ground-up effort by Microsoft engineers, Bing team tried to buy Yahoo, and when this did not work hired a lot of top talent from Yahoo and finally essentially acquired its search and ad business. Satya was directly involved in these efforts. So who is a runner up in IaaS business who Microsoft could acquire to get more visible in that space? Rackspace? Savvis? Although, one could argue that search share was more relevant in search advertising business in which the big get bigger (why even bother advertising with small players?) and this advantage of scale is not as relevant in hosting, so acquisitions might not be as effective. We will see…
  • Not sure if Azure appliance emphasis will persist: Azure appliance made a lot of sense under old leadership. Server and Tools Business knows how to sell to enterprises, so let’s turn Azure into an appliance which we can sell to our existing biggest partners and customers. Will Satya feel the same? I don’t think Bing folks were paying much attention to Microsoft’s search appliance strategy leaving this all to SharePoint/FAST and concentrating on pure cloud play…

There were speculations after Ray Ozzie left that Azure might get de-emphasized – after all Azure was one of Ray’s pet projects. With Satya’s appointment, I would say that we should expect Azure to only gain priority at Microsoft. We’ll see how applicable will Bing experience be for making Windows Azure a top player in the cloud platform space.

Advertisements

Is there hard ROI to use a cloud IaaS instead of a server in your garage/basement/on-premise datacenter? I think there increasingly is and justifying self-hosting is getting increasingly tough.

I would actually go as far as posit that you can now get a server in a public datacenter at price comparable to your electricity bill alone!

If you don’t believe me – let’s do a quick math.

Mark Kolich noticed in his blog that the server he had running at his home was consuming 220 W, which at the consumer electricity costs of 12-cents per kWh means:

0.220 kWh * 12 cents = 2.64 cents per hour

Almost 3 cents/hour for electricity alone not taking into account: labor, server hardware amortization, data-storage costs (replacing a failed disk), cooling costs, ISP costs, security costs (routers, firewalls, etc.), power backup costs (a UPS) and so on. Mark notes that he could have probably bought a newer more energy efficient server – but the required investment would not justify the savings.

The shocking part is that the recent price competition of cloud infrastructure (IaaS) and platform (PaaS) vendors took the current cloud servers costs to roughly the same order of costs. Here’s a quick survey of a few major cloud players:

  • Microsoft is rolling out their 5 cent/hour option (with additional further discounts if you pre-pay for reserved use – e.g. say you have a bunch of instances which you have running all the time and you are willing to pre-pay for the next few months).
  • Same thing with Amazon: minimal price (although for a slightly more limited version) is already in 2 cent for Linux / 3 cent for Windows instance area, with reserved/pre-paid option getting as low as 0.7 cents/Linux & 1.3 cents/Windows.
  • Rackspace pricing starts at 1.5 cents/hour for Linux, and 8 cents/hour for Windows.

My take on these numbers is that you need to have a really good reason to go into hosting when there is so much price competition in that space and the margins are going down so fast.

The only good reason I can think of is hosting being your competitive advantage in some way. For example, being a local hosting company in a country which legislation is making it hard to use foreign datacenters. Or offering some level of compliance which public hosters cannot provide. And as a matter of fact both of these differentiators are gradually going away with the vendors quickly getting all the possible certifications and compliance stamps you can think of, as well as opening datacenters around the globe.

Cloud is cheaper than your own hosting regardless on how you calculate the costs. Get used to it.

Dmitry

What do you do once you become the top bookseller and web-startup hoster? You shoot for the enterprise market!

That seemed to be the sentiment of Amazon’s Cloud for the Enterprise event which the company held in Sofitel Los Angeles last week. The pitch boiled down to:

  • Amazon’s datacenters are the most reliable, secure, and cost-effective option you can find,
  • Amazon Virtual Private Cloud (VPC) lets you securely connect an isolated subnet of Amazon Web Services (AWS) to your intranet,
  • If you are already virtualizing your applications, why not then run them in AWS and not spend money and efforts on whatever datacenter expansion you might need on your end?

Why this makes sense?

Amazon went a long way to make their datacenters more reliable and secure, they have the technology for network connectivity, and they do get significant economies of scale. The latter is not just words. Amazon’s CTO – Werner Vogels – showed pie chart of the cost structure for their datacenters.

They have almost eliminated labor among the significant cost factors – which is great considering that labor is one of the top (if not the top) elements of typical on-premise IT environments.

However, they went further than that. In their current cost structure server hardware is by far the number one cost absorbing more than 50% of what they have to spend. This made them work hard on improving the utilization of these resources. What they did is sell these compute resources as a combination of:

  • Reserved instances (when customers commit to resources for a long period of time to get 50%-70% discounts),
  • On-demand instances (normal hourly pay as you go model), and
  • Spot instances (when the remaining resources get automatically auctioned among bidders in a name-your-own-price scheme)

This means that they can get server utilization close to 100% – which is incredible considering that typical numbers in the industry are probably within 10-30% range.

Considering all this, why bother buying a new server when Amazon can deliver a potentially better service (with additional availability options, global datacenters and so on) at a lower rate?

What is in it for Amazon?

This also seems to be a natural adjacent market for Amazon (the IaaS company – not the online retailer). If they already successfully host web startups and are the most well-known compute platform for tasks such video transcoding or text recognition – why not use that same expertise and infrastructure to sell it to enterprises?

Enterprise IT is a huge market with great margins, and as corporate CIOs are looking for ways to use the cloud to cut costs and/or become more agile – Amazon has the brand recognition to be their number one choice.

This seems to be a high priority effort for the company considering that they have their CTO attending and delivering his keynote at events like the one in LA. And it should be if Amazon does not want to be squeezed between enterprise vendors like Microsoft and VMware getting the higher margin enterprise cloud segment, while initiatives like OpenStack commoditizing lower end cloud compute services.

With so many vendors going after them, Amazon needs to keep moving fast to stay relevant.

Are we there yet?

With all that being said, today Amazon’s pitch remains a great story rather than reality for both technology, business and perception issues.

Technological challenges include inability for IT today to easily (or better automatically) move workloads between their on-premise datacenters and Amazon’s cloud. Even the virtual machine images Amazon is using are not compatible with the VMware and Hyper-V hypervisors enterprises have.

Obviously most of the existing IT management and monitoring tools that companies are using are not yet Amazon aware either – meaning that administrators cannot just get Amazon added to what they have already but instead would have to learn new ways and find new tools.

From business perspective, Amazon is just not an enterprise vendor yet. Corporations have contracts with Microsoft, IBM and others – Amazon is brand new to these customers.

Perception-wise, Amazon needs to find early adopters of that enterprise IT scenario to showcase at events like this. The 4 customers presenting at the event in LA were using AWS to:

  1. Offload computation tasks,
  2. Do image processing,
  3. Host web sites, and
  4. Host their SaaS electronic medical records application in the cloud.

Needless to say, these are not the scenarios Amazon was trying to pitch.

Summary

With the enterprises starting to evaluate their cloud options, the fight for the cloud for the enterprise is only going to become hotter. It is going to be interesting to see if Amazon finds a way to “descend” from the public cloud to the on-premise and hybrid scenarios with smart partnering and acquisition strategy, or traditional enterprise and virtualization players add public cloud to their solution sets and squeeze Amazon out.

Forrester’s Frank Gillet just published results of a big survey on Infrastructure-as-a-Service (IaaS – like Amazon’s compute cloud or internal cloud-like datacenter) acceptance across geographies and company sizes titled “Conventional Wisdom Is Wrong About Cloud IaaS“.

As the title implies the survey has a few very interesting results.

First and foremost, they found that bigger customers are actually more willing to use IaaS than smaller companies – Frank attributes that to enterprises being more familiar with virtualization trends.

Public clouds are actually more popular than private (internal) ones – which in my mind is because there is simply too much confusion about what internal clouds are and technology not really being at the commodity level yet. Frank adds a very good point on inadequate positioning on internal clouds as pay-per-use, rather than self-service on-demand resources which makes the idea look less appealing.

Production use is as high as dev/test – actually this makes a lot of sense. In the early virtualization days it was mostly used in dev/test environments and only then got accepted for production systems. In my opinion, cloud computing is now piggy-backing on this virtualization success and thus jumping ahead right to the production acceptance.

Level of acceptance of cloud computing is approximately the same across geographies (e.g. US and Europe) but not across verticals (e.g. public sector lacking way behind retail – which again makes sense because the former has more regulatory concerns and the latter is used to supply chains and is all about efficiency.)

Overall, this is a very good report with lots of useful data. If you are a Forrester subscriber or have $1999 to spend go get it here.

Technorati Tags:
, , , ,


RSS My company’s main blog

My Recent Tweets

Error: Twitter did not respond. Please wait a few minutes and refresh this page.

Blogroll

Legal

The posts on this blog are provided “as is” with no warranties and confer no rights. The opinions expressed on this site are mine and mine alone, and do not necessarily represent those of my employer Jelastic or anyone else for that matter. All trademarks acknowledged.

© 2008-2012 Dmitry Sotnikov

%d bloggers like this: