Data Centers Flashcards
What is a Cloud Service Provider (CSP)?
A cloud service provider, or CSP, is an IT company that provides on-demand, scalable computing resources (computing power, data storage, or applications) over the internet.
Typically, cloud-based service models are defined as IaaS (infrastructure as a service), PaaS (platform as a service), or SaaS (software as a service).
The CSP market includes cloud providers of all shapes and sizes. The big three—Google Cloud, Microsoft Azure, and Amazon Web Services (AWS)—are considered the established leaders. However, there are a host of other smaller or niche players that offer cloud services as well, including IBM, Alibaba, Oracle, Red Hat, DigitalOcean, and Rackspace.
What are the 3 types of services a CSP offers?
- IaaS (Infra a as a service): CSPs provide internet access to IT infrastructure, saving customers from buying and managing their own. This offers flexibility and cost savings, but customers are still responsible for their software and troubleshooting.
- PaaS (platform as a service): comes after IaaS. It gives you tools and services to make and launch apps. PaaS takes care of things like the operating system and software tools needed for your apps, as well as the hardware stuff. It’s like a handy package to build and run your apps without worrying about the technical details.
- Software as a Service: SaaS solutions are perhaps the most familiar and make up the fastest growing services in the cloud market. CSPs deliver ready-to-use applications and are responsible for maintaining and managing everything, from hardware and maintenance to development, scaling, and delivery. For example, you probably use many of the SaaS productivity applications in Google Workspace every day, such as Gmail, Calendar, Docs, and Drive.
What is the effective utilization rate of a signed data center lease?
60% to 80
Retail, wholesale, hyperscaler power demands
Retail 0 50 kw
Wholesale 50 kw - 4MW
Hyperscaler > 4 MW
What are the three key elements that determine how much a data center customer will pay monthly?
Fixed monthly lease rate $ per kw
PUI = Power Efficiency
Electricity Cost $ per MWhr
Colocation Data center
A Colocation data center is typically a multi tenant data center
Most of these data centers have 100k sqft
Data halls can be divided among just one customer or multiple
7-8 layers of security
Resin and y of power n+1 electrical, ups, generator level and the cooling side and mechanical side
Some financial companies ask for 2n
Mostly sited in the suburbs given the land requirements.
Customers of Colocation facilities are typically smaller companies or startups that don’t have massive data requirements
Healthcare, financial, insurance companies
Retail leasing rates are 2 to 3x but it requires so much work
Data Center Retrofits
Less common except in the Sillicom Valley area due to permitting difficulties and land availability
Hyperscaler data center
When a company asks for 30 MWs they topically don’t need the full amount but rather they start small and grow over time
Enterprise Data Center
Private companies build out their own data centers. Designed purposely built for one organization. Typically, 20 cabinets located at the corporate headquarters on their own premises
65% of the companies in amarrica still have enterprises data centers
Most of them are outdated and the IT migration is painful. Huge opportunity to move this enterprise users to the cloud
Carrier Hotels
Data centers that have massive connectivity to multiple fiber and network lines
One word to associate carrier hotels with ecosystems
Putting a cabinet in a carrier hotel is much more expansive than anywhere else given its connectivity
A lot of the carrier hotels are located in downtown areas
Some famous carrier hotels are the infomart in Dallas Texas which is owned by Equinix
350 Easttarmac in Chigaco is owned by digital realty
The main carrier hotel in LA is 1 Wilshire. This one of the most well connected buildings in the West Coast
And
Edge Data Centers
The goal of the edge data centers is to solve network connectivity issues
Typically small betweeen 100 to 300kw with n+1 redundancy. More flexible from a location standpoint. Placed at higher location densities, base of cell towers
Content companies that want to be as close as possible to the eye balls are the ones that want edge data centers
Primary use case is self driving cars and as the car drives through the city it gets new info from the data centers around it
When people refer to hyperscale data centers, they are most often referring to:
1) The customers who lease space in a data center
2) The developers that build a data center
3) The size of the data center
So, a hyperscale data center is one that has been built by a hyperscale company or one that has been designed specifically to meet the needs of a hyperscale company. Many of the latter type are designed in partnership with the company who is leasing the space as part of the pre-leasing process
What are the two trends that have driven the Rise of Hyperscale data centers?
Two trends have driven the need to large-scale IT infrastructure.
1) Cloud Computing: The cloud acts as demand aggregators for businesses who are looking to convert their on-premises IT workloads. Cloud platforms allow users to provision resources and scale quickly, often remotely and near-instantaneously. Cloud companies forecast future growth and “buy IT in bulk” to meet that demand.
2) The rapid rise of SaaS, streaming, and social media further drives growth in the industry. All three of these types of companies need massive amounts of storage, compute, and bandwidth to meet their customers needs.
Who Are the Hyperscale Users?
Amazon, Microsoft, Google, and IBM all have hyperscale data centers in every major region of the world, and other top players like Facebook, Apple, Oracle, and Chinese cloud giant Alibaba aren’t far behind. These A-listers can take down over 70 megawatts and hundreds of thousands of square feet at a time.
Next in line are companies with lower present requirements, but who still want the ability to grow a few megawatts at a time. Among their ranks are companies like Salesforce, SAP, Dropbox, Twitter, Uber, and Lyft.
Hyperscale Over Colocation?
Hyperscale Approach: The largest companies that use a lot of data centers often prefer to build their own data centers or lease customized ones. They do this based on various factors like cost, timing, and their future requirements. Sometimes, they even work closely with data center providers to design and build data centers exactly to their needs. They might only get an empty building (powered shell) and then fill it with their own computer equipment.
Colocation Approach: Smaller but still quite large companies opt to lease data center space from specialized providers. They use these data centers like a traditional rental service. However, these companies also plan for future growth and have strategies to expand their data center usage over time.
