domeet webmaster

• The use process is not transparent, and the cost cannot be quantified.

Snowflake, the third company under the control of Slootman, not only achieves the ultimate in cost reduction and efficiency enhancement, but also attempts to change the aforementioned “black box” problem.

On the occasion of listing, the company invited Forrester, a well-known research institution, to publish a report “The Total Economic Impact of Snowflake’s Cloud Data Platform” that measures the economic benefits of its core product Cloud Data Platform.

After investigating four large customers from software, finance, FMCG and other industries, with a three-year cycle, the following fully quantified indicators were obtained, including:

• By helping customers improve the efficiency of product launches and reduce the investment in infrastructure such as databases,A total of about 21.5 million US dollars of economic value has been created;

• At the same time, the total cost of internal investment and paid to Snowflake by the customer in this process is 1.1 million US dollars;

• Therefore, the customer’s investment within three years is ROI of 612% >.

(From: Forrester)

At the 2019 User Conference, Slootman also mentioned that a customer clearly told him that he could do 15 times more work than before with the same database investment , At this time he realized that “This is a revolutionary moment-When the economic benefits undergo fundamental changes, customers will completely change their attitudes towards new technologies.”

Reflected in the business model, Snowflake customers no longer pay a fixed annual fee in advance, and the platform will settle the settlement completely based on the calculation and storage usage actually consumed by the customer during use. The cost is measurable.

Secondly, based on the real statistics of unit resource consumption, the customer can compare the investment in each link with the traditional model, and finally get the same accurate ROI as in the report—realizing a measurable return.

Behind this more real “pay-as-you-go” model is a new set of product design and technical architecture.

Build a “cloud” on the cloud

In May 2017, the Economist published an article “The world’s most valuable resource is no longer oil, but data” which caused a lot of discussion. (The most valuable resource in the world is no longer oil, but data). Regardless of whether the view is correct or not, Buffett is probably the one who will truly implement this proposal after three years.

(From: The Economist)

Even if data is not oil, it has already become the power source that drives various industries from design, production to business decision-making. In the information age, all applications and software are built on a structure used to store data-the database (Database).

On the occasion of Snowflake’s listing, the Economist’s new article “Steam engine in the cloud” (“steam engine” on the cloud) Vividly likened to “a kind of parking lot”.

I extend this further and try my best to uncover the mystery of this “new technology” that has completely changed customers.

First of all, if you compare the “car” to files or information that need to be stored, the traditional disk storage method is like a normal parking, where there is a vacant space, no matter the model and size of the vehicle. But disks are expensive and have limited capacity.

(From: Pinterest)

So, in order to save space, people think of disassembling the car into various parts-the seat is put together, the tires are put together, the same is true for Audi and Mercedes-Benz. By analogy, all types of components are stored in specific areas.

In this way, each area has its own characteristics, and the components are systematically related. ThisThe storage method is called “relational” storage, and such a parking lot is “relational database”. This field is now dominated by a few companies such as Oracle.

The surge in the amount of relational data began in finance. A large number of data reads and writes were used to record related transactions, and they were stored in a timely and accurate manner. It is also necessary to establish a certain relationship between different databases to facilitate cross-calculation. But at this time the data format is relatively simple, mainly structured information such as numbers and text.

This form of data processing is called OLTP(On-Line Transactional Processing, online transaction processing), which consumes a lot of< strong>Storage of (Storage) resources.

(From: WordPress, a typical relational database map)

However, with the diversification of data usage, data formats have become more complex, including unstructured types such as pictures, sounds, or videos.

It’s as if we stopped the car and used a pen to write down the parking space number on paper. Now we directly take a photo or record a video of the surrounding environment and put it on our mobile phone. When we return to the parking lot and import the video from the album into an application, we will immediately get the navigation route.

At this time, we need a new storage architecture that can store structured and unstructured data at the same time, which is what we know as “Data Warehouse (Data warehouse)“.

This concept was proposed by Teradata in the late 1970s, and its core purpose is to support the ever-increasing business intelligence analysis.(BI analysis), and nowadays complex big data and machine learning operations.

This form of data processing is what we often hear about OLAP(On-Line Analytical Processing, online analysis and processing), the main consumption is Compute (Compute) resources are CPU or even GPU.

Traditional data warehouses are still deployed on local hard drives, and the storage space is much higher than the computing space, and the ratio is fixed. This is like in a fixed parking lot, the storage is the number of parking spaces inside, and the calculation is the barrier system for vehicles entering and exiting.

In the digital age, requests for data analysis began to far exceed the demand for storage, leading to a surge in demand for data warehouses. However, the demand for storage and computing is not increasing in the same proportion, that is, the same database may be called several times a minute to hundreds of times, and the calculation speed may become analytical bottleneck.

For example, when we often park in rush hours, we see long lines on both sides of the barrier gates-the driver’s payment and the lifting and lowering of the barriers take too much time.

(Barrier gate system often becomes the bottleneck of parking lot access speed)

At this moment, Snowflake made two key changes:

First, put the data warehouse on the cloud.

“Going to the cloud” is easier to understand: a parking lot can’t fit, and multiple nearby parking lots work together. For example, one parking lot is dedicated to storing tires and the other is dedicated to storing seats.It can store components of similar attributes or brands, reducing handling time.

In addition, when the three major public cloud vendors of AWS, Microsoft’s Azure and Google’s GCP are segregated, large enterprises prefer to diversify risks, not only using hybrid cloud deployment, but also want to put data and applications on multiple platforms . This presents an opportunity for a neutral third-party supplier.

When third-party vendors can manage offline and cloud, relational and non-relational databases in the cloud, the problem of data islands caused by being stored in different places in the past disappears. BI analysts or data scientists can retrieve data in any format anytime, anywhere, even raw data that has not been processed.

Such an all-encompassing “Yunsucang” has a more vivid name——Data lake(Data lake). And Snowflake created a unified entrance, allowing users to access all the databases behind them in the language of accessing relational databases in the past, almost without changing the original usage habits.

(Shuhu: Collect→Import→Mix→Transform→Release→ Distribution)

Secondly, separate storage and calculation.

Relying on the three major public cloud vendors, the “Cloud Data Warehouse” architecture allows customers to expand and shrink at any time with low cost and high efficiency.

Returning to the example of the parking lot, when the car needs to leave the parking lot, Snowflake can accurately calculate the time for each part to be moved, assembled, and waiting for the barrier.

Storage costs are already relatively transparent, but transportation, assembly and other gates need to undergo different forms of calculations. The costs are not the same. Users can choose according to their own budget.You can leave the parking lot as soon as possible.

So, by decoupling storage and computing, user costs can be quantified and controlled.

In the billing model, the price of storage is the same as other public clouds, but the calculated price is divided into eight levels so that customers can individually expand storage or computing according to their needs. And when demand drops, the platform also automatically scales down for customers.

(From: Snowflake)

When almost all SaaS companies are thinking about how to make customers spend more money, Snowflake does the opposite, thinking about how to help customers save more money.

In the past, when public cloud was used, how much storage and computing resources were consumed, no manufacturer would disclose to customers. But in Snowflake, storage is storage, calculation is calculation, and service is service, which constitutes the three-tier architecture of its products.

(From: Snowflake)

Only when the consumption is post-positioned and the use process is completely decoupled, SaaS truly realizes “pay on demand”.

In other words, Snowflake has built a highly-scalable and centrally managed data warehouse on the public cloud vendor, IaaS-built a “cloud” on the cloud.

HhhypergrIn an article on Snowflake’s in-depth analysis, owth pointed out the company’s products and commercial characteristics:

“They are a cloud-native data platform that acts as a cloud-within-a-cloud…and so are capable of near limitless scale(They are a cloud-native data platform, which is the “cloud” on cloud facilities, so it can be expanded almost unlimitedly). “

Has SaaS 3.0 come?

From Salesforce to ServiceNow, as pioneers in the era of “software swallowing everything”, they used disruptive business models and product architectures to fire the first shots like traditional software vendors, aiming at the core information of the enterprise Chemical facilities.

In the 21st century, 2.0 manufacturers led by Atlassian, Zoom and Slack, etc., carry the natural genes of cloud native, and use the network effect brought by the mobile Internet to penetrate into the enterprise from the bottom up.All work scenarios, greatly improving the efficiency of staff work and management.

And Snowflake seems to announce the arrival of a “3.0 era” that is more destructive to traditional manufacturers and more customer-friendly from multiple angles.

First, the software will define all infrastructure (Infrastructure) services.

From the last era of software-defined networking (SDN), software-defined radio (SDR) to software-defined storage and computing, and even software-defined cars, a new generation of virtualization technologyThe power of art will be further amplified in this era of “clouds”.

After reading Snowflake’s prospectus, the first thing I thought of was Twilio, which went public four years ago.

As I said in “API originator Twilio’s counterattack and wild vision , Twilio has defined itself as “AWS for telecom(AWS in the telecom industry)” since Day 1-dedicated to solving many black box problems in the communications industry, Including contract prepaid system and cumbersome configuration requirements, etc., we also hope to minimize the threshold for customers.

So, you will find many commonalities between the two companies, including:

• By the Integration is highly flexible and neutral: Twilio is integrating the three major U.S. operators, and Snowflake is integrating the three major public cloud vendors;

• Realize “pay on demand” by decoupling software and hardware through technology : Twilio is based on the number of messages or call volume, and Snowflake is based on the storage or computing resources consumed;

• Quickly win large customers by lowering the barrier to use: In the early years, Uber and WhatsApp is critical to Twilio, and Cisco and Capital One are critical to Snowflake’s revenue.

The database is only a category of infrastructure. In this market, there are also Databricks(Complete F round of financing at the beginning of the year, valued at US$6.2 billion), and hosted document database company MongoDB(current market value of 15.6 billion US dollars) and so on.

The dominance of Oracle was being disintegrated at a speed that was visible to the naked eye, and we will also see software and hardware gradually decoupled in all infrastructure. The hardware update iteration speed will inevitably slow down further or become less important, and the cost will become more transparent and standard.

(From: The Economist, Gartner)

So, in the future, when “going to the cloud” becomes 100% the first choice for enterprise digitization, it may also be when the core competitiveness of the suppliers behind it really fades.

But now, the development and competition of IaaS is far from the end.

Secondly, openness and network effects will be the labels of the new generation of SaaS.

Snowflake’s Marketplace, a trading platform that provides secondary development and sharing based on database content, may have the same value to the company as the original App Exchange to Salesforce. It is both exciting and slightly vague.

(From: Snowflake)

The excitement is that this reflects the most critical core competitiveness, because neutrality and flexibility are the characteristics that all third-party platforms will have, but the spread of content sharing and transactions gives the platform the opportunity to build a bottom-up Network effect.

I thought that only tool-based SaaS such as Zoom and Slack that are frequently used by individuals have this feature, but Snowflake has captured a group that has grown rapidly in this century-BI analysts and data analysts strong>, and then quickly expanded to data scientists (Data scientist), AI and machine learning engineers, etc.

Nowadays, the influence of core IT personnel on enterprise information procurement decisions is increasing, and the epidemic has accelerated this trend.

Secondly, Marketplace gave birth to a new profession: Data broker(Data broker). There have been intermediaries that perform rapid analysis on the database on the platform, and after paying for data content and computing resources, they can obtain, for example, an analysis report on the growth trend of the epidemic. These reports are then sold to customers in need for profit.

However, the market has many doubts about the prospects of this business.

Because after Salesforce launched the App Exchange or PaaS platform-level business in 2005, it was not until 2019 that is 14 years later that the platform business only surpassed Service Cloud with a year-on-year growth rate of 57%, becoming second only to the core CRM business The second largest source of revenue.

Without a huge user base, partners, and ISVs, the value of data cannot be mentioned. For Snowflake, whether business analysts, data engineers or data intermediaries, the base and perception of this group need to be further developed.

But there is no doubt that the new generation of cloud-native SaaS companies, including Snowflake, Twilio, Zoom, and Slack, will pay more attention to their integration and openness with other cloud-native applications, just like a catfish.May penetrate into every corner of the work scene.

Finally, the new generation of SaaS may truly provide “pay on demand.”

In fact, more and more SaaS companies have begun to provide usage-based billing methods (usage-based). In addition to Twilio and Snowflake, there is also Fastly whose market value has increased nearly six times this year.

Fastly is leveraging the cake of traditional CDN(content delivery network) vendors, and its edge cloud platform helps TikTok, Pinterest, Twitter and other large The platform delivers videos, pictures and other related streaming media data to users around the world faster, greatly reducing loading time and improving user experience.

(From: Xueqiu, Fastly became the SaaS company with the fastest increase during the epidemic)

The business model is to charge based on usage, and at the same time open a wealth of development tools and components for developers, so that customers’ applications can be quickly deployed to the edge cloud.

So in the future, once these infrastructure-providing vendors realize the standard of charging by usage, is it possible for the SaaS vendors above to also charge according to the actual usage of users?

Because the most important cost of SaaS products is the cost of infrastructure such as servers. Once the cost structure becomes clearer, the original monthly or annual fee model based on the size of the customer and the number of users will definitely be able to achieve a more refined design .

In the primary market, I also see that many new generation SaaS companies charge lower subscription service fees first, and then charge extra based on the number of orders or items processed on the platform, because this consumes a lot of calculations Resources to ensure the stability of large-scale and concurrent operations.

Return to “first principles”

Eight years are not long for an enterprise service company, but it has created the largest IPO in the history of the software industry.

Snowflake’s historical process is not only the process of gradual commercialization of the cloud native technology system, but also the tireless exploration and pursuit of several business geniuses and technical geniuses represented by Slootman for 20 years. In the end, given the market and product, the process of releasing commercial value, everything seems accidental, but it is logical.

I was impressed by what Slootman said in an interview:

“I have no playbooks, I am entirely situational and operate on so-called first principles.(I don’t have a script, I act completely by chance and follow First Principles.)“

The CEO of Okta, a publicly listed identity management SaaS company, summed up the reasons for choosing Snowflake on Twitter:

• The price is cheaper;

< /ul>

• Expansion is more flexible;

< /ul>

• Sharing is easier.

Returning to the first principles, aren’t these three points exactly the core values ​​that a new generation of enterprise service companies should provide?

Looking ahead, I believe that under the leadership of Slootman, Snowflake will likely achieve faster growth than ServiceNow more than a decade ago.

But what makes me more excited and looking forward to is–

Will this really open up the next golden decade of SaaS?

Reference material:

Snowflake, Salesforce, Twilio prospectus and previous financial reports;

https://thegeneralist.substack.com/p/the-s-1-club-snowflake-and-the-data

https://hhhypergrowth.com/a-snowflake-deep-dive/

https://www.datanami.com/2019/06/04/slootman-makes-it-snow-at-snowflake-summit/

https://www.computerworld.com/article/3443044/snowflake-ceo-frank-slootman-talks-vision-aws-tension-and-ipo-plans.html

https://blocksandfiles.com/2019/11/11/frank-slootman-snowflake-computing-interview/

This article comes fromWeChat public account: I think I’m here (ID: angelplusdevil), author: SY, GN