Salesforce Architecture Tutorial

Salesforce Architecture Tutorial

Last updated on 29th Sep 2020, Blog, Tutorials

About author

Vignesh (Associate Director - Cloud Architecture )

He is a Proficient Technical Expert for Respective Industry & Serving 11+ Years. Also, Dedicated to Imparts the Informative Knowledge to Freshers. He Share's this Blogs for us.

(5.0) | 14563 Ratings 1339

What is Salesforce Architecture?

Salesforce delivers a highly customized experience to the customers, employees, and partners of an organization. Such a platform is used to customize standard functionality and create custom pages, components, apps, etc. Also it is done faster, mainly because of the superb architecture on which it is built

Terminologies of Salesforce Architecture

Here are some key terminologies that you should familiarize yourself with before we head on to discussing the core technical parts of the architecture.

Subscribe For Free Demo

Error: Contact form not found.

 Terminologies -of -Salesforce -Architecture
  • App : The main thing for which we store data is App, then the main heads under which it is sorted, different information heads are called ‘Objects’ and further details of those objects are called ‘Fields’. So an app is a combination of objects and info on those categories is a field. A single row is called ‘Records’. It is the information stored in fields.
  • Instance : an entire arrangement of frameworks, system and capacity infrastructure, both shared and non-shared, that gives the service to a subset of its clients. For instance, is an example.
  • Superpod : an arrangement of frameworks, system and capacity foundations, including outbound intermediary servers, stack balancers, mail servers, SAN texture and other frameworks supporting different instances. Superpods give the benefit of disconnection inside a datacenter so issues with shared or complex segments can’t affect each instance within that datacenter.
  • Org (a.k.a., organization) : a single client of the Salesforce application. Each trial that began on or produces another org. An org is very adaptable and can have unmistakable security settings, record permeability, and sharing settings, UI look and feel, work processes, triggers, custom articles, custom fields on standard CRM questions, and even custom REST APIs. An org can bolster somewhere in the range of one to a large number of authorized individual client accounts, entry client records and Sites client accounts.
  • Sandbox : an instance of the benefit that hosts full copies of creation organizations for client application advancement purposes. Clients utilizing the Salesforce platform can have full application advancement lifecycles. These are test conditions for clients to do client acknowledgment testing against their applications previously conveying changes into their production org. 

ypes of Relationships in Salesforce

There are  Two Types of Relationships in Salesforce

  • Lookup relationships : These are relationships that allow you to “lookup” related objects from within an object itself. This is the most basic relational link within Salesforce.
  • Master-Detail relationships : This is identical to lookup relationships with one big difference. Master-Detail relationships share a hard relationship. So, one object is a master object the other is a detail object. The master object controls the detail object. So, let’s say that you had a Master Object of Accounts and a Detail object of contacts. If you deleted Accounts, it would delete all of your contacts. These can get hyper-complex and create intricate webs of objects There are a few other types of relationships that fall under the umbrella of these two relationships.

These are:

  • Self-relationships
  • External lookup relationships
  • Indirect lookup relationships
  • Many-to-many relationships
  • Hierarchical relationships
  • relationships.

Core Architecture of Salesforce

The architecture of Salesforce can be put into layers for better understanding. The purpose and function of each layer is described in detail in the below stated paragraphs- 

 Core -Architecture -of -Salesforce

1) Multi-Tenant

Multitenancy is an incredible word for influencing you to sound like a geek at supper parties, all things considered, all it implies is that you’re sharing your assets. Salesforce gives a central arrangement of administrations to every one of its clients in the multi tenant cloud. Regardless of the span of your business, you gain admittance to a similar registering power, information stockpiling, and central features. Whereas a customary single-inhabitant application requires a committed arrangement of assets to satisfy the necessities of only one association, a multi tenant application can fulfil the requirements of numerous occupants (organizations or divisions inside an organization, and so forth.) utilizing the equipment assets and staff expected to oversee only a solitary programming case. 


In Metadata-Driven Architectures multi tenancy is viable just when it can support applications that are dependable, adaptable, upgradeable, secure, and quick. It’s hard to make a statically ordered application executable that can meet these and another one of kind, difficulties of multi-tenancy. Characteristically, a multi tenant application must be dynamic in nature, or polymorphic, to satisfy the individual desires of different inhabitants and their clients. Thus, multi tenant application plans have developed to utilize a runtime motor that produces application parts from metadata—information about the application itself. In an all-around characterized metadata-driven engineering, there is an unmistakable division of the ordered runtime engine (kernel), application information, the metadata that portrays the base usefulness of an application, and the metadata that relates to each occupant’s information and customizations.

Course Curriculum

Get Salesforce Training with Master Advanced Concepts By Experts

  • Instructor-led Sessions
  • Real-life Case Studies
  • Assignments
Explore Curriculum

3) API

On a very basic level, APIs enable diverse bits of programming to interface with each other and trade data. On the off chance that it sounds sort of theoretical, just take a quick glance at the PC you’re working on right at this moment. You can presumably discover a progression of connections of different shapes and sizes that help various types of associations. These resemble the equipment form of APIs. You don’t need to know how the USB port functions. You should simply understand that when you connect your telephone to a USB port, it passes data to your PC.

APIs are comparable. Without knowing the subtle elements, you can interface your applications with different applications or programming frameworks. The fundamental innovation deals with the specifics of how data goes all through the framework. 

The architecture of Salesforce can be put into layers for better understanding. The purpose and function of each layer is described below −

rusted Multi Tenant Cloud

Here multiple instances of one or multiple applications operate independently in a shared environment. The instances are referred to as tenants and they logically separate from each other while physically remaining in the same hardware. It is called trusted because of both its robust nature and high security.

Scalable Metadata Platform

The metadata driven platform makes it easy for customization and scaling up as the amount of data or concurrent user instances increase.

Enterprise Ecosystem

The Enterprise Ecosystem of Sales is very large as a large number of partners contribute by creating and maintaining applications in this platform.

CRM and Related Functionality

Salesforce includes all aspects of CRM in its list of features and also extends it by providing features for creation of apps and integrating analytics, etc.

Hardware/Software Architecture

Logging In To The Salesforce.Com Service

  • We maintain a pool of servers to handle login traffic for all instances. A handful of servers from many (but not all) instances accept login requests and redirect the session to the user’s home instance. This is what happens when you log in via
  • Customer traffic starts with our external DNS. Once a lookup has successfully returned the IP address for an instance, standard Internet routing directs it to the appropriate datacenter.
  • Once the traffic enters our network in that datacenter, it is directed to the load balancer pair on which that IP lives. All of our Internet-facing IPs are VIPs configured on an active/standby pair of load balancers.

Inside The Instance

The load balancer directs the traffic to the application tier of the given instance. At this tier, we service both standard web page traffic as well as our API traffic. API traffic makes up over 60% of the traffic serviced by our application tier overall. Depending on the needs of the customer’s request, it will be directed to additional server tiers for various types of backend processing.

Core App

  • The core app tier contains anywhere from ten to 40 app servers, depending on the instance. Each server runs a single Hotspot JVM configured with as much as a 14 GiB heap, depending on the server hardware configuration.
  • The batch server is responsible for running scheduled, automated processes on the database tier. For example, the Weekly Export process which is used to export customer data in a single archive file format as a form of backup.
  • offers a number of services including basic and advanced content management. We have a content search server and a content batch server for managing asynchronous processes on the content application tier. The content batch servers schedule processing of content types, including functions such as rendering previews of certain file types and file type conversion.


  • The primary data flow occurs between the core app server tier and the database tier. From a software perspective, everything goes through the database so database performance is critical. Each primary instance (e.g. NA, AP or EU instances) uses an 8 node clustered database tier. Customer sandbox (e.g. CS instances) have a 4 node clustered database tier.
  • Since is such a heavily database-driven system, reducing load on the database is critically important. To reduce load on the database tier, we developed ACS — API Cursor Server. This was a solution to 2 problems which enabled us to improve our core database performance significantly. First, we used to store cursors in the database but the deletes were impacting performance. Second, after moving to using database tables to hold cursors, the DDL overhead became a negative impact. Thus was born the ACS. ACS is a cursor cache running on a pair of servers, providing a method to offload cursor processing from the database tier.


  • Our search tier runs on commodity Linux hosts, each of which is augmented with a 640 GiB PCI-E flash drive which serves as a caching layer for search requests. These hosts get their data from a shared SAN array via an NFS file system. Search indexes are stored on the flash drive to enable greater performance for search throughput.
  • Search indexing currently occurs on translation servers which mount LUNs from storage arrays via Fibre Channel SANs. Those LUNs make up a QFS file system which allows single writer but multi-reader access. Like most other critical systems, we run these in active/passive with the passive node doing some low priority search indexing work. It then ships its results to the active partner to write into the QFS file system.
  • The translation occurs when these same LUNs are mounted read-only from a group of four NFS servers running Solaris 10 on SPARC. These SAN mounted file systems are then shared via NFS to the search tier previously described.


  • We maintain a tier of servers that provide object storage, similar in concept to Amazon’s S3 or OpenStacks’ Swift project. This system, Fileforce, was developed internally to reduce the load on our DB tier. Prior to the introduction of Fileforce, all Binary Large Objects (BLOBs) were stored directly in the database. Once Fileforce came online, all BLOBs larger than 32 KiB were migrated into it. BLOBs smaller than 32 KiB in size continue living in the database. All BLOBs in Fileforce have a reference in the database so in order to restore Fileforce data from backups, we have to start a database instance based on a database backup from the same restore point.
  • Fileforce includes a bundler function, developed to reduce the disk seek load on the Fileforce servers. If 100+ objects smaller than 32 KB are stored in the database, a process runs on the app servers to bundle those objects into a single file. A reference to the bundled file remains in the database along with a seek offset into the bundle. This is similar to Facebook’s Haystack image storage system but built into an object storage system.


Each instance contains various other servers for support roles such as debugging application servers and “Hammer testing” app servers in the app tier, hub servers which monitor each instance for health and monitor servers running Nagios. Outside of the instance itself reside supporting servers like storage management, database management, log aggregation, production access authentication and other functions.

Future Directions

  • On the database tier, we’re carefully examining several options for data storage systems. We’re also evaluating higher speed, lower latency interconnects such as Infiniband for our cluster interconnects with our existing database solution.
  • In the future, we expect to make the search hosts do all reads and writes. We are in the process of rolling out Apache Solr for search indexing. Solr runs locally on our existing search hosts. The SAN, NFS servers and SPARC based indexer hosts will all go away which will lead to a dramatic reduction in complexity of the entire search tier.
  • Our application container was previously Resin but over the last year, we’ve been migrating to Jetty. We’re also in the midst of a hardware refresh of our application tier hardware which will increase RAM sizes anywhere from 30% – 266% and introduce Sandy Bridge processors into our stack.
  • I hope this overview of the technology architecture and stack has been interesting and informative. Please leave a comment if you’re interested in additional guest posts deep diving into other parts of our expansive technology infrastructure. 
Salesforce Sample Resumes! Download & Edit, Get Noticed by Top Employers! Download


Salesforce architecture is not a surprising result of any random series of hit and trial experiments. Every feature of its architecture has been carefully planned and placed right where it requires to be. If you get a hand at its architecture you can understand most of its functionality.

Are you looking training with Right Jobs?

Contact Us

Popular Courses