Liftr Insights FAQ
Liftr Insights FAQ

How does Liftr Insights gather data about public clouds?

We implement DevOps—a combination of development and operations. We write and maintain cloud-based apps to gather cloud data via each CSP’s application programming interface (API) and other methods of software automated data collection. If you are into software development, we’re an agile software development shop.

 

Does Liftr Insights field surveys to gather data?

No. We’re not against surveys—they’re very useful for finding information that is not publicly available. But we collect publicly-available data, so surveys are unnecessary. Because our public data collection is based on DevOps, we automatically collect data on everything we’re interested in. It’s fast, it’s comprehensive, there is no sample bias and it is much less expensive to implement than surveys.

 

How does Liftr Insights analyze the data it gathers?

We have data scientists on staff. They do data sciencey things like ensure we’re gathering the right data, normalize data so we can compare cloud service provider offerings and perform data analytics to compare the cloud service providers.

 

Does Liftr Insights use machine learning?

Yes, we occasionally use machine learning. We spend a lot of time on making sure we gather the right data and that the data we gather is clean (data scientists call it “data hygiene”). Then, through our Liftr Cloud Distiller database and other internal systems of record, we normalize the data we gather from each cloud service provider so that we can make fair comparisons between the providers. Machine learning is part of our analysis mix. We also use a lot of other analytic techniques.

 

Any data through API?

Yes, as part of the automated data pipeline, we harvest the bulk of our data through API. Having programmatic access to data ensures that our analysis is less prone to errors and always up-to-date.

 

How do I know your data is unbiased?

We reduce bias by not fielding surveys and not sampling. We reduce bias by not cherry-picking the best data or cutting corners to collect the convenient data. We count everything.

What does Liftr Cloud Components Tracker measure?

We measure…

  • All the public data and metadata we can find
  • Describing each infrastructure-as-a-service compute instance
  • In every region
  • For each cloud service provider we track

Such as…

  • Instance Metadata
    • CSP
    • Type
    • Size
    • Region
    • Price
  • Processor Specs
    • Vendor
    • Instruction Set
    • Brand
    • Model Series
    • Model Number
    • Generation
    • Core Count
    • Core Speed
    • Memory Size
  • Accelerator Specs
    • Vendor
    • Type
    • Brand
    • Model
    • Chip Count
    • Memory Size

 

How often do we measure all of that stuff above?

We measure all that stuff every month, at the end of the month. That’s twelve (12) times per year.

 

What’s a “core”?

We use the term “core” to standardize between Alibaba Cloud’s “CPU Core”, AWS’s “vCPU”, Microsoft Azure’s “VM” and Google Cloud Platform’s “Virtual CPU”. In practical terms, all of these translate to the number of physical processor cores available to an application. 

 

What is a “compute instance”?

A public cloud compute instance is a provisioned and ready to use virtual or bare metal server, accessible across the public Internet. For practical purposes, an instance can be all of or a portion of a physical server. 

 

What is an “instance type”?

An instance type describes a group of identical physical servers. Type defines basic configuration, regional availability, and service level agreement (SLA) for a class of virtual servers at a CSP. Type specifies non-configurable specifications, such as processor vendor, brand, model, speed, and maximum available cores. 

 

What is an “instance type size”?

Instance type sizes define configurable specifications, exact regional availability, and pricing for rentable virtual servers. An instance type size is a rentable virtual server with an exact configuration, by a CSP, in a region, with a price, and with a production SLA. 

Size configurable specifications include the number of virtual cores, amount of system memory, dedicated accelerator options. Operating system choice is usually associated with types but may affect pricing and SLAs for each size. 

 

What is “bare metal”?

Bare metal instance types provision one entire physical server – no more and no less. For the time a customer rents a bare metal instance, it is not a shared resource, the entire server is controlled by the customer. Typically, CSPs offer bare metal provisioning as an instance type size. Customers have the option to load their own operating systems on bare metal instance type sizes. 

 

What do we mean by “instance type share”?

We calculate instance type share as a percent of production instance type sizesThe metric enables us to compare the relative sizes of deployed configurations without knowing how many physical servers are deployed behind each instance type. We exclude pre-production instance types and sizes. 

 

There are a lot of price options for each instance type size, how do we compare pricing?

We record on-demand pricing for every instance type at the end of every month. On-demand pricing is used for analysis and comparisons of all instance types, because all other pricing, including reserved and spot pricing, are discounted from on-demand pricing: 

  • Reserved pricing is based on a customer commitment to a minimum number of hours. 
  • Spot pricing is typically subject to availability and usage restrictions. 

Price analysis and comparisons are restricted (where possible) to Linux instance types with no installed software or dedicated accelerators and on-demand contract terms. Otherwise, Windows instance type prices are used under the same conditions. 

 

What’s an “accelerator”?

We use the term accelerator to refer to chips and add-in boards that can offload specific types of workloads from a server processor and calculate those workloads much faster than a processor. There are many types of accelerators, from general purpose merchant accelerators, based on graphics processing unit (GPUs) and field-programmable gate array (FPGA) chips, to specialized and proprietary accelerators, such as Google’s tensor processing unit (TPU) and AWS’s upcoming Inferentia, both designed to accelerate deep learning workloads. 

 

What does “disaggregated mean?

We use the term “disaggregated” to describe separating a server’s processor from any compute accelerators, such as FPGAs and GPUs, so that the two functions have to communicate with each other over a cloud data center’s network. A disaggregated instance is located in a different physical server from an accelerator-only instance, they are connected over a network. 

 

What is a “processor-only” instance type?

Processor-only instance types use only a server processor for performing calculations – they have no accelerators included in the instance type. 

 

Disaggregated instance graphic

What is a “dedicated accelerator”?

Dedicated accelerators are compute offload chips or add-in boards that are directly attached to a server processor with a hardwired connection. Directly attached means that the accelerator and processor are socketed on the same motherboard, the accelerator is attached to an add-in card slot on the same motherboard as the processor, or the accelerator is attached to a dedicated cabled connection with one server motherboard and is not a shared resource. 

 

What is a “disaggregated accelerator”?

A disaggregated accelerator is an accelerator located remotely from a server processor with a dynamic network connection between the two. Dynamic means the network connection can be reconfigured, so that the disaggregated accelerator can be paired with one of many physical servers, as long as both the disaggregated accelerator and the server are on the same network. 

Disaggregated types can be paired with a variety of processor-only instance types, but CSPs often limit the host instance types available for pairing. 

 

How many physical servers are deployed behind each instance type we record?

We don’t know for sure, and there is no legitimate way to find out with any accuracy. This info is a deep secret for each cloud. There are ways to infer how much hardware each cloud is investing in though, especially when we track new instance deployments month-to-month. Wait a few months, we’ll have some really cool insights about how many servers and chips are being deployed.

 

What workloads or applications are running on a specific instance type?

We have no idea what kinds of workloads are running on any instance type and no way to find out. Neither do the cloud service providers themselves, if they are adhering to their customer privacy agreements. Finding that kind of non-public info would take partnering with other companies that have invested in implementing large-scale surveys. First, we’d have to find cloud customers using the specific instance type in question, then ask them to divulge which apps they are running in that instance type. Even if we find a few customers using that type, many will not tell us what they are running in that type. We’d be open to partnering with other companies who are practiced at surveys, but our core value is DevOps-based automated collection of publicly-available data.

 

How do you compare the value of a specific instance type configuration between regions in a cloud?

We do that by clustering instances with a narrow range of configurations to determine a center of gravity for pricing. Then we assume that comparing similar configurations in different regions within a cloud or between different clouds is similar. This is a reasonable approximation, but system architectures may vary between clouds, affecting the performance of instance types with similar high-level specs.

 

Can you compare similar instance type configurations using different processor architectures and/or system architectures?

We can’t directly compare the value of instances based on different processor architectures or system architectures… yet. We’ll need to run workload-specific performance analysis, also known as benchmarks. This applies to:

  • Comparing instance types with different processor architectures.
  • Comparing instance types that appear to have similar instance type specifications but are implemented with different system architectures in different clouds.

The above apply whether comparing instance types within one cloud or across different clouds.

 

Which cloud service providers do you track in the Liftr Cloud Components Tracker?

  • Alibaba Cloud (also known as “Aliyun”)
  • Amazon Web Services (AWS)
  • Google Cloud Platform (GCP)
  • Microsoft Azure (Azure)

Who is Liftr Insights LLC?

  • How many employees do you have?
  • Approximately how many customers do you currently have?
  • What is your customer mix?
  • What portion of your customers are in the investment/financial services industry?
  • Are your customers institutions, individuals, or both?
  • Etc…

We are a privately-held company of under 20 employees.

 

Is Liftr Insights Involved in Trading?

No

 

Do you have an affiliated asset manager?

No.

 

Do you have a trading desk?

No.

 

When was the last time you conducted training on insider trading issues?

Not applicable.

 

Has your Firm undergone any routine regulatory examinations over the past two years?

No.

 

Do you recommend trading actions or stocks or anything else that would prohibit you from distributing your research during trading hours?

No.

 

Do you facilitate one‐on‐one communications between clients and third‐party industry experts (“consultants”) who you compensate, whether in cash or in‐kind?

No.

 

Do you conduct surveys at the specific request of a client, where such survey results are disseminated to one or a select subset of clients only?

No.

 

In the context of researching a company, do you conduct interviews with people or entities who do business with that company, whether in the supply chain, distribution channel, or otherwise?

No.

 

Do you provide information relating to government (including legislative or agency) developments, consult with members or employees of government, and/or facilitate meetings with members or employees of government?

No.

 

Do you provide information relating to non‐public drug trials, consult with medical doctors or other healthcare industry professionals regarding non‐public drug trials, and/or facilitate meetings with medical doctors or other healthcare industry professionals regarding non‐public drug trials?

No.

 

Do you consult with investment bankers?

No.

 

Do you have a Chief Compliance Officer or a Chief Information Officer?

No.

 

Who is in charge of compliance?

The Chief Operating Officer: https://liftrinsights.com/dt_team/uttam-reddy/

 

Approximately how many compliance personnel do you employ?

Two.

 

What role does your compliance personnel play in the pre‐publication review of your research products?

Peer review of content.

 

Do you have a compliance manual?

We have documented processes and we gather public data from public companies.

Access to customer queries is restricted on a need to know basis, with direct oversight by CEO and COO.

 

Do you have a legal opinion or process in place to ensure that the information you share with or provide is not material non-public information obtained in breach of a duty?

Yes. Our data gathering process uses only public sources made available by the companies we track.

 

Do you have policies and procedures in place to handle receipt of material non-public or confidential information?

Not applicable.

 

Do you perform diligence on your data sources to confirm that they have the requisite consent from their users to collect and share the data with you?

No, our data sources provide data describing their own unique services.

 

Do you review the privacy policy?

Yes.

 

Do you review the terms of service?

Yes.

 

Please describe what other steps you take.

We ensure we are following any requirements regarding the use of third-party data.

 

What is your pre‐publication compliance review process for information included in your written research?

Our Principal Analyst and team review each report internally prior to release to subscribers.

 

Please provide the name of a representative data source or provide a representative example of the type of consent language you expect those data sources to have.

Microsoft Corporation:

Do you enter into contracts with your data sources?

No.

 

Do you ever compensate anyone for information (including any in‐kind benefits that are not de minimis)?

No.

 

Is any information used to prepare your research subject to a non-disclosure/confidentiality agreement?

No.

 

Do you consult with outside counsel?

Yes.

 

Have you received an opinion of counsel on your business model or consent process for data sources?

Yes.

 

Does your firm collect and/or provide data, and/or provide analysis on data?

Yes.

 

Can you represent that you are fully authorized to sell such data to institutional investors?

Yes, all raw data used in our analysis are publicly available to anyone.

 

Are you providing data on companies traded in the US, outside the US, or both?

Both.

 

What are our Liftr Cloud Components Tracker deliverables?

Monthly data analytics and insights about the public cloud market, specifically the Infrastructure-as-a-Service (IaaS) market. Our standard service includes:

  • A monthly report document (PDF format)
  • A monthly summary spreadsheet containing the data used to generate the monthly report’s charts and graphs along with high resolution images of the charts and graphs (XLSX format)
  • Inquiry (phone, email, other text messaging channels) to answer questions about report topics, supply clarifications and definitions, etc.

 

Will you be doing the same data extraction/collection and sharing for other customers whether or not we hire/contract with you?

Yes.

 

Do you enable every customer to access all of your data and insight or do you withhold data from some customers?

  • Do we receive the same information and in the same format as all of your clients or are the results customized to us and our interests?

  • Do you conduct custom research projects at the specific request of a client, where such custom research results are disseminated to one or a select subset of clients only?

  • Etc.

Basic service: Everyone receives the same information in the same format.

Custom services: Custom queries use the same database used in our basic service, but specific query fields and report formats will be selected to answer specific customer queries. Custom responses are only provided to the requesting customer, the responses are not shared with other customers, although customers with similar queries will receive similar responses based on the same underlying data.

 

Even if the results are provided in a customized format, is the underlying data that is being provided the same for different clients?

Yes, see above.

 

Do you anonymize, aggregate or summarize the data before providing it, or do you provide it in raw form?

We never provide raw data to customers.

We do not need to anonymize our data, there is no personally identifiable information (PII) in the data we collect.

We aggregate, standardize, normalize, and then summarize data before reporting the data to customers.

 

Do you receive individual user data from your data sources?

No.

 

Has any personally identifiable information (PII) contained in the data been masked or deleted?

No, there is no PII contained in the data we gather.

 

Does the data contain any personally identifiable, health or medical or other information that is protected by any law, rule or regulation?

No.

 

What information can we expect to receive regarding the source of the data?

Our data sources are public. Each company’s public data starts at its top-level URL and then permeates a legion of sub-URLs:

 

Can the data be directly linked to the sites/apps from which you received it?

No.

 

How do you store and process the data collected?

We store data in Amazon Web Services (AWS) S3 buckets in AWS’s US East (northern Virginia) zone. Our Liftr Cloud Components Tracker data collection is processed solely using AWS “serverless” functions in that same zone.

 

Does Liftr Cloud Components Tracker use web scraping?

No.

 

Do you have specific jurisdictional rules on what activities you will or will not undertake in certain states or countries?

No.

 

In what jurisdictions are your analysts located?

USA

 

In what jurisdictions are your servers located?

USA

 

Have you received any cease-and-desist letters related to scraping?

No.

 

How will you respond to a cease-and-desist letter that you receive?

We will find alternate legal means to collect or best approximate the data from other sources, calculations, and/or extrapolations.

 

Will you notify us if you receive a C&D letter?

Only if specifically requested to do so and it is information specifically related to data a customer is engaged to receive.

 

Have you been subject to any copyright infringement claims?

No.

 

Have you ever received a subpoena or been subject to any investigations by a regulator, law enforcement agency or governmental organization related to your scraping activities?

No.

 

When scraping a site, do you review the Terms of Service (TOS) prior to gathering information?

Not applicable.

 

If an account requires you to create an account to gain access to the information:

Do you create an account?

Yes.

 

Does the created account reflect your actual association?

Yes.

 

Are you accessing known APIs?

Yes.

 

Are you subject to any API or developer agreements on these sites?

Yes, where applicable.

 

How do you prevent interfering with the site functionality?

Not applicable.

 

What IP addresses will appear in the site’s logs?

Our own IP addresses.

 

Do you engage in any behavior to mask your identity / IP addresses?

No.

 

Do you have agreements with any of the sites from which you collect data?

Yes, where applicable.

 

If so, what limitations, if any, do the licenses impose on republication / reuse?

Our agreements do not limit our use of the data we collect.

 

Do you have requirements/guidelines about how your data sources gather and/or provide the information you receive?

No, our data sources provide data describing their own unique services.

 

Do you audit your data sources for changes in terms of service/privacy policies/service offerings, and if so, how often?

Yes, when we receive notifications of changes and annually.

 

Please explain how you establish the right to share data obtained from your platform or from your customers.

Data used is publicly available from the top four public cloud providers: Amazon Web Services, Microsoft Azure, Alibaba Cloud, and Google Cloud.

 

As applicable, provide a copy of your contract, terms of service, privacy policy, and/or acceptable use policy to illustrate the basis for this right.

See URLs listed above.

 

Does your right to share this data depend on the form in which the data is provided?

No.

 

Please note any other limits placed on the use of the data that we receive.

Liftr Insights data subscription is for internal use only at each customer or specified department or division and may not be shared externally with any other organization, department, or division.

 

Have you received any regulatory, governmental or quasi‐governmental, or law enforcement inquiries relating to your research and/or handling of confidential information?

No.

 

Do you have a Chief Information Security Officer (“CISO”) or other dedicated security professional who oversees the security and integrity of your computer systems?

Liftr Insights’ head of product development and operations has responsibility for the security and integrity of our IT infrastructure.

 

Do you have a Comprehensive Information Security Plan (“CISP”)?

We have internal security around our IT systems but do not have a CISP.

 

Do you have access controls in place to ensure employees and third-party service providers only can access information necessary to perform their responsibilities?

Yes. Two factor authentication is used on our key data collection infrastructure.

 

Have you had a data security incident or potential incident in the last 5 years?

No.

 

Do you have mechanisms in place to ensure that our company data is maintained in a confidential manner?

Liftr Insights does not have access to customer data.  We supply data analytics subscriptions base on independent research to our customers.

We protect customer queries under mutual non-disclosure agreement – we do not disclose queries made by any individual customer.

What kinds of information can I see on the map?

You can see a substantial amount information we aggregate in our Liftr Cloud Distiller (see below) on the Liftr Cloud Regions Map. We can map any information that a tracked CSP makes available by region.

 

Do I have to pay or login to use the Liftr Cloud Regions Map?

No, it is free to use here.

 

Can I share the map?

Yes, it is open to anyone with a web browser that is compatible with Google Maps.

 

How should I use the map?

It is based on Google Maps, so it responds to the same mouse clicks or finger touches that Google Maps responds to. Zoom, scroll, select, all work as in Google Maps. Select a pull-down dimension and then select an active location to see what a cloud service provider in that region offers to customers.

 

Where do you get the data for the map?

We aggregate display data for the Liftr Cloud Regions Map from our monthly Liftr Cloud Components tracker scans and a lot of other publicly available metrics. We gather this data from cloud service provider public configuration and pricing APIs. Then our data scientists and developers sanity check the data.

 

How factually accurate is the map?

See the Liftr Cloud Distiller and Liftr Cloud Components Tracker sections.

 

How geographically accurate is the map?

We position each region in the city the cloud service provider identifies with the region or in the most prominent capital city in the region.

 

How often do you update the map?

We update the Liftr Cloud Regions Map at least once every month, after our end-of-month Liftr Cloud Components Tracker scan. The update is an automated pull from our Liftr Cloud Distiller database.

 

Which cloud service providers do you track in the Liftr Cloud Regions Map?

Currently, we track four CSPs who collectively represent over 60% of IaaS and PaaS revenue:

  • Alibaba Cloud (also known as “Aliyun”)
  • Amazon Web Services (AWS)
  • Google Cloud Platform (GCP)
  • Microsoft Azure (Azure)

What is it?

The Liftr Cloud Distiller is our proprietary database at the core of our multi-cloud comparisons. It compares similar IaaS instance types and also similar PaaS services across tracked cloud service providers.

 

What does it do?

The Liftr Cloud Distiller enumerates a list of high-value IaaS instance types and PaaS services from tracked CSPs, which enables in-depth comparisons and analysis of cloud technologies, cloud services and cloud industry trends. Liftr Cloud Distiller currently tracks and normalizes about 200 unique types of IaaS instance types and PaaS services. We examine each CSPs’ services to equate its services to our normalized services in an unbiased manner. Some CSP services are complex and map into multiple normalized services, other services map one-to-one. Cloud Distiller is a many-to-many mapping of services that enables us to make a fair “apples-to-apples” assessment of CSPs’ services offering.

 

Why did you develop it?

In 2017, before the Liftr Insights product line or Liftr Insights company start-up, we were part of a cloud services broker. The broker needed a fair and unbiased way to compare IaaS instance types and PaaS services to guide its customers to the best cloud for different classes of workloads and specific applications.

 

Are you selling it?

No. At the moment it is powering our complimentary Liftr Cloud Regions Map. In the future we may offer for-profit queries into the Liftr Cloud Distiller targeted at specific market needs.

 

What is its relationship to Liftr Insights products and services?

Liftr Cloud Distiller is the engine behind our products and services, including Liftr Cloud Regions Map. For example, we add new information to our Liftr Cloud Regions Map by first adding a CSP or a new service to our Liftr Cloud Distiller.

 

Which cloud service providers do you track in the Liftr Cloud Distiller?

Currently, we track four CSPs who collectively represent over 60% of IaaS and PaaS revenue:

  • Alibaba Cloud (also known as “Aliyun”)
  • Amazon Web Services (AWS)
  • Google Cloud Platform (GCP)
  • Microsoft Azure (Azure)

“We identify trends and deliver insights about deployed Internet-attached infrastructure.”

Learn more about Liftr Insights’ services