We get an interesting perspective on the market place. Our clients depend on us to steer them in the right direction for their IT and Cloud needs. The engagements we participate in that allow us to really make decisions along side our clients and watch the implementation and growth occur. Surveys, interviews and other assessments aren’t things we currently do but can give an analyst thousands  of data points to examine. In our work, we gather fewer data points but each point is much deeper giving us a different perspective.

So here’s what we see in the Cloud space right now:

Cloud providers have so effectively automated and accounted for costs down to the computing unit and proven they have vast capacity it’s becoming harder sell the following:

1. Start costs unless massive infrastructure is being deployed

2. Long term contracts unless substantial savings are involved

3. Non – automated tasks / no control over any aspect of the environment. (i.e. you need to open a ticket to get something done

What do purchasers of cloud look like? We are seeing 2 types of clients:

Small to Midsize organizations

These buyers need full control over the environment. Other than some replication and light API use they don’t need the ability to burst nor to expand their footprint dynamically from their own datacenter. Their own datacenter holdings maybe light or non existent at all. They are quite happy having all their computing infrastructure with the cloud provider. These can be traditional companies or high tech companies that don’t want to build out an operations team.

Large Organizations

These are the firms that need to dynamically scale their environment off their platform and onto the Cloud providers platform. This is sometimes referred as  Cloud bursting. Security, scalability and integration are key here. This is probably someplace where contracts will be signed since the footprint is ongoing and connected through a unified management console that the client owns and controls.

The OpEx vs CapEX argument revisited

This topic has also been well addressed online so we only have a few points to add to the discussion.

Smaller organizations that preferred to write one check and keep month operating expenditures lower are realizing that you never really do escape some type of on going fee without sacrificing uptime. Maintenance on hardware, software, consultants for upgrades and in house staff to keep things moving are unavoidable. Even as reliable as systems are becoming, the complexity and the work required to keep things running remains.

Larger organizations that can exceed the economies of scale of a cloud OR those who are in finance or in the defense industries will get to the point soon where their datacenters will be so automated, measured, cost predicable and scalable that they can even contemplate selling off excess capacity and billing it to outside firms. This is how Amazon started.

So the Cloud is truly becoming a utility but it’s not free and still costs money and requires expertise to manage even though it’s external. We’ll revisit this subject as 2012 continues to see how it’s changing from our quick and dirty analysis above.

Steve will be very missed and I pray that family and friends find comfort during this difficult time.

1. The cafeteria service (they should have been paid more than us!)

2. The shuttle service

3. Direct Deposit

Working at a smaller firm has allowed us to increase our efficiency and reduce everything to the basics which enable more to be done with less. In the past, I remember wading through so much garbage to get the essence of what was being presented. 75 page powerpoint slides could have been whittled down to 8. Emails that babbled on for pages could have been summarized in 6 bullet points.

Since we assist folks in locating datacenter space all around the world, our colocation database is critical. We have things in there like region, address, certifications, telco connections and network providers. There’s a lot of data that we’ve removed such as quantity of available square footage and number of UPS generators a facility has. Keeping track of dynamic data like that is difficult and there’s a good chance that even the provider themselves won’t track it accurately. Here are some examples of very helpful data:

restaurant availability

lodging availability

access to airport and railways, distance to the airport/train station

We’d love to get to the point where someone could ask us if there was a Chinese takeout near the datacenter. For now, we’ll use google maps. (I am sure someone will mention that the database could dynamically grab this data based off the address)

I remember back to my “large company days” – our websites did a great job of offering info on compliance, financial reports and the management team. They all failed terribly at telling folks what the heck we did. I found a few things were prevalent:

1. The more info that was crammed into the site, the better management seemed to think it was.

2. The wording was so vague and complex that it really defeated the purpose of the website all together.

3. The offering of services was so vast that it seemed unlikely that we were really good at any services listed.

To combat this, we decided to put up a video that would graphically illustrate our “datacenter locator service” so anyone could grasp our model which is the matching of datacenter providers with potential clients at no charge to the client.  This was back in February of last year and we did our best with the resources we had at hand. Here’s the “v1″ video:

(direct link here)

The content was ok but the colors were a bit pale making the graphics unreadable. We also recorded the video at different times so you can tell that the tempo and pitch changed between recording sessions.

Deciding to redo the video here’s what we came up with:

(direct link here)

We’re much happier with this version of the video for sure. Within 1 minute 49 seconds most folks in the industry will understand what we do.  The dots concept we used came from Jeff who specializes at modeling complexity with dots.

We find that our colocation datacenter partners fall into 3 categories (this is our own take as there are many official classifications in use out there):

A. Super Mission Critical Datacenters

These facilities are the Tier 3+ facilities (using the Uptime Institutes rating) that are designed for applications that cannot tolerate downtime of any sort. They offer N+2 redundancy (2 being the needed amount of utility then there are 2 spares to backup the primary utility). These facilities have no compromise in design and are generally engineered in every way not to go down. We have 20% of our clients that require these types of facilities.

B. Mission Critical Facilities

These tend to be Tier 2-3 facilities that make up the other 80% of our client’s needs. Definitely redundant facilities and in the worst case will be N + 1 redundant, meaning that they provide the needed facility + 1 spare. Also rigorous in their design we find that for the price point per square foot makes  sense for most of our clients.

C. Non-Mission Critical / Secure Facilities

Sometimes we have clients that need secure facilities for hosting an alpha or beta test application. Sometimes firms looking for test labs find that these non-critical facilities are great since they are economical. If this facility takes a power hit or loses network connectivity it’s an inconvenience but no revenues or major productivity is usually lost.

We have a partner that fits into the A section which we’ve done business with successfully in the past. Even though our deal volume with them is quite light, they’ve recently gone out of their way to stay in touch with us, offer us support and training and are generally pleasant to work with. The periodic (but not overwhelming) updates from them keeps them at the top of our list whenever we have a client in need of facilities in their regions. In essence, they are a great partner. They keep us educated, they don’t overwhelm us and we enjoy working with them

Contrasting this, we had a partner that fit into category C, which is the non-critical datacenter choice. We signed up with this company because we do find this need on occasion. Before we were given the opportunity to partner with these folks on a single deal we received a certified letter in the mail telling us that due to our “low volume” we were officially being removed from the partner program. Never did we get a call asking us if we needed help positioning their offering or any assistance at all. I’d like to say no hard feelings but I am glad that certified mail costs around $5.00 per item these days.

In the past, I had issues with our late 1990′s datacenter designs that would allocate an entire HP DL360 just to be used as an FTP server before virtualization was popular like today. Nowadays, virtualization has helped out greatly but in designs with very focused applications Dell may have a better solution. Datacenter Knowledge has a great article today about Dells foray into this very interesting area of computing. Let me spoil the surprise by giving you a clue- I remember specifically thinking in 2001 that I’d love servers with the size and power of my laptop minus the screen and keyboard to complete these designs when very small applications just needed to be run.

We recently had our own thoughts on a similar method of computing done by SeaMicro which uses a similar methodology to provide dense computing power in a small footprint. It would be interesting to see an ROI of a Blade Farm with one of the popular virtualization softwares vs. running these smaller footprint computing platforms and seeing what yields more.

I hadn’t really thought much more about it until I realized that running dual monitors and tons of apps for 14 hours a day on my 4 year old MacBook Pro was starting to take a toll on it when I decided I would use the MBP for traveling, working on the terrace or working just away from my desk. I considered the iMac  but I already owned 3 monitors and couldn’t justify the iMac and the Mac Pro was just overkill. I settled on the Mac Mini and now I know why there are firms out their that colocate these amazing machines.

First of all, I got the 2.4 GHz, 2 GB RAM model which I thought would be power starved. Actually it’sresponsive, boots up lightning fast and really takes a beating daily without showing any signs of slow down. Not only do I run lots of applications at once, but instead of 2 monitors, I pulled out a 3rd that I had stored and now run 3 monitors on the Mac Mini without any performance hits. (If you are wondering how, the Mini has HDMI which I’ve converted to DVI and it has a mini-display port which I’ve also converted to DVI. The last monitor was connected with a display link adapter that converts one of my USB ports to a DVI port.)

I could definitely see with 4GB or perhaps even 8GB’s of RAM that this small machine could be an excellent server in a datacenter while barely consuming power and space. This whole thing got me to thinking about a company called Sea Micro who uses small PCI board based computers to deliver 512 Intel Atom cores in a single unit. So rather than virtualizing really large servers with hundreds of virtual instances, you’d literally be able to have lots of “mini computers” in a single unit that has a really high back plane through-put for network and redundancy on all elements. So rather than splitting up large scale computers into smaller chunks, you would just buy smaller chunks of physical computing. Interesting things to consider these days

I have a relative (really smart business guy) who was heavy into IT back in the 80′s who asked me to assist in setting up a client/server application a few years back for his company. They were always in the same office but as they spread their firm out a bit they needed remote access to the application. Going to a web application wasn’t an option due to a large investment in the software and to be honest it was running well except for the remote access part of it.

I decided to just install the application on a server and enable terminal services so that all the employees (even in the same office) would just remote desktop (RDC) into the new server. When my relative asked me to explain what RDC did in simple terms I put it this way to him:

me: remember the days of connecting to a mainframe with a dumb terminal and everything was green and text only?

relative: yes

me: well, imagine that the dumb terminal has a much fatter connection to it and it’s in color, all the screens are transmitted to your end user. That’s all we’ve done here.

relative: got it

So when I think about it, we’ve all gone full cycle on many things. But I remember making fun of mainframes as they are a bit before my time and considered “old school”. Now if you look at these large virtualized blade server farms with all the users connecting via web browser we’ve gone back full circle to centralized computing with a single place where all the security and maintenance takes place. The clients can be somewhat un-secure and all is normally fine since the interaction is purely by browser. So as much as we may be hyping cloud it’s not a ton different that we had in the past with time sharing systems and similar.