Wednesday, October 31, 2012
Why Charge For Floorspace?
I work with quite a number of colocation providers and it still seems there is a mix of how colocation services are priced. In a retail environment, generally the major components are space and power. This has held true since I entered the colocation side of the industry in 1996. Back then, the floor space was considered the critical element as building more space was costly to the provider and most customers had very low power density requirements. Contracts would have one line item for space and another line item for power. So why, 16 years later, do we still see so many providers using the same model, despite colocation having undergone major changes?
The answer is most likely due to conditioning of the product groups, Sales teams, and even the customer. I've heard some providers say floor space is irrelavant now. That statement may hold true in certain circumstances, but we need to define those circumstances.
Each retail colocation facility will have a floor density associated to the colocation space as rated in "watts per square foot". Most people assume this is a power rating, but it actually has more to do with cooling of the power consumed within that space rather than the delivery of power to that space. If a space has a "watts per square foot" of 150, that means that for each given square foot of space used by a customer, the facility can accommodate 150 watts of cooling and power load. In a cage environment the math is simple: a 10 x 10 cage = 100 square feet and at 150 watts per square feet the cage can accommodate 15,000 watts. Simple, right? It gets more complicated in cabinets as each provider may use a different square footage for each cabinet, but for simplicity, let's assume a cabinet equates to 25 square feet. With the same 150 watts per square foot, that equates to 3,750 watts available to that cabinet.
Why does density matter when we are talking about charging for floor space? By knowing the watts a space can accommodate we can quickly determine if a customer wants to utilize a larger amount of power than the space will accommodate If so, then the floor space itself does not matter as the customer will always use less space than their power requirements require. If we tell that 10 x 10 cage customer we will sell them 15,000 watts (15 kW) of power and we know their power density requirement is higher than our 150 watts per square foot, then we also know they will not use all the floor space before they use all the power. Now we can just make our contract a single line item for power, inclusive of space.
Where we run into problems with the single line item pricing is if we have a customer that has power density requirements lower than our watts per square foot. I've seen this only once in the last 6 years. The customer was a provider that sold individual cabinets to their own clients. Most of their clients were dev environments that only needed a single 120v-20A circuit per cabinet. That power requirement equates to 1.92kW or 1,920 watts. With a 25 square foot cabinet, that means that each cabinet was really only needing 76.8 watts per square foot. In essence that customer would have used double the floor space and if we were only charging for power, the provider would lose a lot of usable space. This very antiquated dev model is being replaced by cloud computing now, but there are still some clients out there that use very low density.
Moving to a single line item power pricing model is the way of the future in retail colocation. Just make sure to run your customer density requirements first to verify they need power beyond your own floor density.
Tuesday, October 30, 2012
All Colo Is Not Created Equal
Time and Time again I hear, "colocation is a commodity". This would imply all service offerings along with all the components that make up each service offering were identical. And this simply isn't the case.
Many are familiar with The Uptime Institute and their data center tier rating where data centers are rated from a Tier I (least redundant) to Tier IV (most redundant, least downtime). The rating system does not allow 1/4, 1/3, or 1/2 points, so many times when we hear of a Tier III facility, it is really Tier II by definition, but the provider feels it warrants a higher rating for various reasons. This question of tier rating is one of the arguments I hear most often between customer and provider or provider to provider comparisons. But does it matter? It most definitely matters in terms of defining our data center and colocation world, but what I mean is 'does it matter to the customer'?
The only way any service offering and comparison of offerings matters to the customer is if the customer has a firm grasp of what they require. This means internal audits of their own infrastructure, project growth, technology changes (such as physical to virtual environments), M&A activity, DR strategy, proximity and more. This is the crux of the issue and will determine why data center tiers and the different colocation offerings will become apparent and will matter. If the customer does not know this information, the Sales team should be able to probe and probably arrange their offering to suit.
Let's assume there is no such thing as a "good" colo or a "bad" colo. If the customer understands their own requirements, there will just be offerings that are less than what is needed, offerings that are appropriate for what is needed, and offerings that are more than what is needed. Now we can quickly look at those same data center tier ratings and understand how they can help compile a short list of possible providers. We can also, without prejudice, understand that a Tier I facility does not mean it is a "bad" facility, but may mean it is more suited toward lab, dev, or SMB markets and a Tier IV facility does not mean it is a "good" facility, but might mean it is more suited toward big dollar mission critical infrastructure that can bear the expense of the added redundancies.
Within each tier there are many design and aesthetic variables due to building age, data center "generation", as well as more practical components such as power and cooling designs and rack/cabinet/cage layout. A note to the customer - verify all infrastructure and service offerings before you worry about aesthetics. A note to the provider - the customer is initially only going to worry about aesthetics ;)
Monday, October 29, 2012
Datacenter vs Colocation
I hear a lot of clients, as well as providers, use the words datacenter (or data center, or even data centre for our UK friends) interchangeably with the word colocation (co-location/colo). What is the difference or is there one?
The simplest way to think of a datacenter is to think of the actual building structure. This comprises the foundation, walls, roof, and all the core components that give it life, such as power, cooling, and fire suppression. It also comprises common areas, work areas, the restrooms, offices, entry way, etc. Colocation is merely the room within the datacenter where you locate your equipment. These rooms will have sub-systems or extensions of the core power, cooling, and fire suppression, but the main or core components are really still part of the building. Think of the datacenter as your house and the colocation is just one of the rooms. Simple, isn't it?
This is very straight forward for multi-tenant datacenter (MTDC) environments as there is a single service provider/operator and many customers, each with colocation within the building. Where things get more tricky is when you have an Enterprise datacenter. Many times the Enterprise will own or run the building, as well as utilize space within the building for their compute equipment. At this point, it would seem the datacenter is the colo, but in effect, the same rules apply. The building itself is the datacenter and the equipment room within the building is the colo. Remember, the datacenter itself is comprised of all the ancillary systems and rooms in addition to the equipment room, even if that equipment room is the majority of the total facility.
But what about REIT infrastructures such as 200 Paul in San Francisco or 300 Blvd East in Weehawken? Both buildings provide wholesale offerings to their clients, who in turn provide services to their own clients? This creates a nested solution where the landlord owns the building and the tenants lease from the landlord and sublease to their own clients. The same logic applies as before. The building itself is the datacenter and each room/suite/pod within the building is still colocation. It just happens that provider colocation is split again into colocation for their clients in terms of suite, cage, and cabinet offerings. Think of that same house. All we have done is allowed different rooms to become colocation provider specific, while they lease space in that room to their clients.
This is review for many, but will give us a common building block for our future blogs.
The simplest way to think of a datacenter is to think of the actual building structure. This comprises the foundation, walls, roof, and all the core components that give it life, such as power, cooling, and fire suppression. It also comprises common areas, work areas, the restrooms, offices, entry way, etc. Colocation is merely the room within the datacenter where you locate your equipment. These rooms will have sub-systems or extensions of the core power, cooling, and fire suppression, but the main or core components are really still part of the building. Think of the datacenter as your house and the colocation is just one of the rooms. Simple, isn't it?
This is very straight forward for multi-tenant datacenter (MTDC) environments as there is a single service provider/operator and many customers, each with colocation within the building. Where things get more tricky is when you have an Enterprise datacenter. Many times the Enterprise will own or run the building, as well as utilize space within the building for their compute equipment. At this point, it would seem the datacenter is the colo, but in effect, the same rules apply. The building itself is the datacenter and the equipment room within the building is the colo. Remember, the datacenter itself is comprised of all the ancillary systems and rooms in addition to the equipment room, even if that equipment room is the majority of the total facility.
But what about REIT infrastructures such as 200 Paul in San Francisco or 300 Blvd East in Weehawken? Both buildings provide wholesale offerings to their clients, who in turn provide services to their own clients? This creates a nested solution where the landlord owns the building and the tenants lease from the landlord and sublease to their own clients. The same logic applies as before. The building itself is the datacenter and each room/suite/pod within the building is still colocation. It just happens that provider colocation is split again into colocation for their clients in terms of suite, cage, and cabinet offerings. Think of that same house. All we have done is allowed different rooms to become colocation provider specific, while they lease space in that room to their clients.
This is review for many, but will give us a common building block for our future blogs.
Subscribe to:
Posts (Atom)

