Technical writing

How to make current job more fulfilling

1 Point ContributorLeave a comment

Are you losing interest in your current job? Do you feel perhaps changing it will bring more happiness? Before you go further with stronger steps, what if you re-evaluate your situation? Is it really the job you less like or something else within the job that needs adjustments? Let’s review some steps on how we can make our current job more meaningful:

  • Know why behind the job you’re in: Understand why you’re in this job. Is it your financial responsibility that’s bringing you to the work every day? Why did you choose this line of work? Does it relate to your interests?
  • Is there anything that is lacking in the current job? Evaluate reasons what can increase the engagement at work. What growth perspective you’d like to focus on? Is it more salary? Is it the increase in the impact to the people or business you support? Is it the technology/skills you use at work? Here’s a quick trick to evaluate. Would you prefer to continue doing the work you do, if:
    • You’re offered 2 times the current salary or
    • You’re offered a promotion or
    • You’re offered to increase your business impact or
    • You’re offered to work with more people or
    • You’re offered to switch to use different technology/skills
  • Evaluating it would help to increase the engagement. If we find out that we should move for a better option, we should plan for it. Sometimes, we find out that we’re already in a great environment. For example, a friend of mine was not engaged at work. After the analysis, he realized that given two times the salary, he’d continue working at the same place with the same set of people, supporting the same business. So, there was nothing wrong in his current job except a low pay situation. He decided to explore if he’s really paid low. The reality was that he’s reasonably paid. His comparison of low-pay was also somewhat unreasonable. Sometimes, it’s our comparison, judgement, and other such inner-enemies that cause us distress. Sometimes some people have a feeling that we’re not doing enough or we’re simply not happy where we are. After the realization, he’s now focusing on his current job with a desire to increase his impact. Another person in a similar situation found out that he can get a better salary outside. He planned to move to a better paying job.
  • What’s the minimum required from you at work: Make a list of minimum required expectations from you at work. How knowing the minimum required expectations help? Many times, we’re overwhelmed and lost with many demands at work. Sometimes, it becomes confusing to navigate the day at work successfully. Making a list of minimum required tasks provides us the clarity of expectations. For example, if someone is a software developer at work, they can write minimum required tasks as:
    • Prime tasks like driving projects A, B, and, C. Write down what’s the minimum required expectation for each project.
    • Supporting tasks: update on progress every week, month, and quarter. What does it means? How much times does it require every week?
    • Trainings: are there any mandatory trainings that you have to attend?
  • After making a list of minimum required tasks, plan your time at work to first address these tasks. In my experience, I noticed that many times, ambiguous situations and confusions take up a lot of time. Without the clarity of minimum required tasks, we maybe wasting time. After a certain time, we start being overwhelmed with many things undone.
  • Look at the bigger picture of your current job. Let’s look at three areas to evaluate it:
    • People: how are your work relationships with people? Do you feel connected with people at work? Can you make friends at work to share your life and listen to their life situations? Connecting with people at work can make the work more interesting. Also, knowing that you’re not alone in life’s common challenges can help to find mentor support. For example, a friend found that his colleague also has interest in writing. What if he and his colleague can find moments in the day to share their writing content?
    • Business: what business does your job supports? What’s the bigger picture of it? For example, are you in an IT development job? What social cause your company support? If it’s a public company, what is the sector of your company’s stock? How your work impacts the society? When we relate our work with a larger cause, it could help in bringing the engagement.
    • Technology/Skills: What skills do you need, to perform your current job? What are your areas of expertise? How valuable are those skills in the outside market? Is there any skill that other would like to learn from you? If you’re interested in mentoring others, would you like to mentor someone at work? Is there any particular skill you’d like to learn more at work?
  • Plan for periodic leaves. Sometimes, we feel that we have no choice other than going to the same work every week. Taking leaves helps us to reflect upon life. If it’s feasible, plan for a longer leave, like a week. If you’re free without any work tasks, what do you feel to do? For example, my one friend likes learning about technologies even when he’s out of the work. He’s naturally inclined towards technologies. By taking leaves, he realized that even if he doesn’t need a job for his financial needs, he’d like to write programs in his free time. He realized that it’s not the work type that he wants to change. Sometimes, he’s overwhelmed at work. So, the solution is to find out how can he prioritize his work so he can set the right expectations.
  • Plan for continuous evaluations: Plan for a periodic evaluation to assess where you’re. If you’re interested in something else, keep pivoting from the current situations. For example, after a decade of experience in a technical skills set, if you’re interested in moving to a new skills set, start planning for a steady move.

I’d love to know yours feedback on it. Thank you.

Common Design Patterns Overview

1 Point ContributorLeave a comment

In this article, we’ll go through the overview of basic design patterns.

Circuit Breaker: This is a pattern that helps to manage calls from one service to another. There are three states of it:

  • Open: Calls from one service to another service are not allowed.
  • Closed: Calls from one service to another service are allowed.
  • Half-Open: A few calls from one service to another service are allowed but not all calls are allowed.

Two implementations for circuit breakers: Hystrix and Resilience4J.

Bulkhead: It allows to set maximum concurrent users that can connect to a service.

Backpressure: We will add details of it later.

Bloom filters: it is a data structure to search an element in a data set quickly with the level of certainty. Guava is a known Java API implementation of bloom filters.

HyperLogLog: it is a data structure that can provide the probabilistic calculation of the cardinality of a data set. Let’s say we want to understand how many unique visitors visited a mall. We can use HyperLogLog data structure to do it efficiently.

Gang of Four Design Patterns: It consists of three types of patterns: structural, creational, and behavioral, In total there are 23 of patterns.

References:

Bloom filters: https://richardstartin.github.io/posts/building-a-bloom-filter-from-scratch

HyperLogLog: https://www.baeldung.com/java-hyperloglog

Strength app part 5: Enable https on AWS

1 Point Contributor1 Comment

This is the part 5 of application development series. Refer to part 4 for the previous information. On our strength application, we wanted to enable https certificate. As it is for learning purpose, we wanted to keep it low cost.

Here were our options:

  • Enable AWS provided https option.
  • Get a free https certificate via letsencrypt and enable it on AWS. For our Sprint Boot application, we needed to generate a keystore.p12 file. We decided to opt for option2: get a free https certificate via letsencrypt website.

Our next challenge is to access the generated certificate into Spring Boot application in a way that is scalable in the future and does not go away if we terminate our EC2 instance on the ECS cluster. Here are options for us:

  • Manually copy https certificate to EC2 instance. We did not opt for this option. Reason is, if we terminate our ECS instance (attached to the ECS cluster), the https certificate will be deleted with the termination of the EC2 instance.
  • Keep the certificate at Amazon S3. Then, copy it to EC2 instance manually. We did not opt this option because every time we have a need to recreate an EC2 instance, we will have to manually copy the certificate.
  • When creating an EC2 instance within ECS cluster, add commands in user data option, to copy the certificate from AWS S3. We think this is an optimum option. But we couldn’t enable it. Free version of ECS enabled EC2 instance did not allow adding user data properly. To allow running user data into EC2 instance, we had to run an EC2 agent configuration. Running these configurations were either not easily available or too complicated within the free tier EC2 instance. So, we did not opt this option.
  • Add the https certificate within the Spring Boot application via S3 copy using SSL configuration. This could have been a considerable option. Within Spring Boot code, we can add SSL configuration bean to copy the certificate from AWS S3 and recreate a certificate file within the Spring Boot application. Below is a sample code to do it:

import java.io.File;

import org.apache.catalina.Context;
import org.apache.catalina.connector.Connector;
import org.apache.tomcat.util.descriptor.web.SecurityCollection;
import org.apache.tomcat.util.descriptor.web.SecurityConstraint;
import org.springframework.boot.web.embedded.tomcat.TomcatServletWebServerFactory;
import org.springframework.boot.web.servlet.server.ServletWebServerFactory;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;

//@Configuration
public class SslConfiguration {

@Bean
public ServletWebServerFactory servletContainer() {
    TomcatServletWebServerFactory tomcat = new TomcatServletWebServerFactory() {
        @Override
        protected void postProcessContext(Context context) {
            SecurityConstraint securityConstraint = new SecurityConstraint();
            securityConstraint.setUserConstraint("CONFIDENTIAL");
            SecurityCollection collection = new SecurityCollection();
            collection.addPattern("/*");
            securityConstraint.addCollection(collection);
            context.addConstraint(securityConstraint);
        }
    };
    tomcat.addAdditionalTomcatConnectors(redirectConnector());
    return tomcat;
}

private Connector redirectConnector() {

    Connector connector = new Connector("org.apache.coyote.http11.Http11NioProtocol");

    connector.setPort(8443);
    connector.setSecure(true);
    connector.setScheme("https");
    connector.setAttribute("keyAlias", "tomcat");
	connector.setAttribute("keystorePass", "<hidden>");
	connector.setAttribute("keyStoreType", "PKCS12");

    Object keystoreFile;
    File file = new File("");// ADD PATH
    String absoluteKeystoreFile = file.getAbsolutePath();

    connector.setAttribute("keystoreFile", absoluteKeystoreFile);
    connector.setAttribute("clientAuth", "false");
    connector.setAttribute("sslProtocol", "TLS");
    connector.setAttribute("SSLEnabled", true);
    return connector;

}

}
  • Add the https certificate within the Spring Boot application via S3 using properties file. To use this option, we need to read the https certificate file from application.properties. Below is a sample code to do it:

private static void copySSLCertificateFromS3() {

try {

Properties props = readPropertiesFile("src/main/resources/application.properties");

String clientRegion = props.getProperty("clientRegion");

String bucketName = props.getProperty("bucketName");

String sslFileNameWithPath = props.getProperty("sslFileNameWithPath");

String keyStoreFileName = props.getProperty("server.ssl.key-store");

AmazonS3 s3Client = AmazonS3ClientBuilder.standard().withRegion(clientRegion)

.withCredentials(new ProfileCredentialsProvider()).build();

S3Object object = s3Client.getObject(new GetObjectRequest(bucketName, sslFileNameWithPath));

InputStream objectData = object.getObjectContent();

// Process the objectData stream.

File file = new File(keyStoreFileName);

try (OutputStream outputStream = new FileOutputStream(file)) {

IOUtils.copy(objectData, outputStream);

} catch (FileNotFoundException e) {

e.printStackTrace();

// handle exception here

} catch (IOException e) {

e.printStackTrace();

// handle exception here

}

objectData.close();

} catch (Exception e) {

e.printStackTrace();

}

}

public static Properties readPropertiesFile(String fileName) throws IOException {

FileInputStream fis = null;

Properties prop = null;

try {

fis = new FileInputStream(fileName);

prop = new Properties();

prop.load(fis);

} catch (FileNotFoundException fnfe) {

fnfe.printStackTrace();

} catch (IOException ioe) {

ioe.printStackTrace();

} finally {

fis.close();

}

return prop;

}

private static void copySSLCertificateFromS3() {

try {

Properties props = readPropertiesFile("src/main/resources/application.properties");

String clientRegion = props.getProperty("clientRegion");

String bucketName = props.getProperty("bucketName");

String sslFileNameWithPath = props.getProperty("sslFileNameWithPath");

String keyStoreFileName = props.getProperty("server.ssl.key-store");

AmazonS3 s3Client = AmazonS3ClientBuilder.standard().withRegion(clientRegion)

.withCredentials(new ProfileCredentialsProvider()).build();

S3Object object = s3Client.getObject(new GetObjectRequest(bucketName, sslFileNameWithPath));

InputStream objectData = object.getObjectContent();

// Process the objectData stream.

File file = new File(keyStoreFileName);

try (OutputStream outputStream = new FileOutputStream(file)) {

IOUtils.copy(objectData, outputStream);

} catch (FileNotFoundException e) {

e.printStackTrace();

// handle exception here

} catch (IOException e) {

e.printStackTrace();

// handle exception here

}

objectData.close();

} catch (Exception e) {

e.printStackTrace();

}

}

public static Properties readPropertiesFile(String fileName) throws IOException {

FileInputStream fis = null;

Properties prop = null;

try {

fis = new FileInputStream(fileName);

prop = new Properties();

prop.load(fis);

} catch (FileNotFoundException fnfe) {

fnfe.printStackTrace();

} catch (IOException ioe) {

ioe.printStackTrace();

} finally {

fis.close();

}

return prop;

}

  • Use a docker container to copy https certificate form S3 to EC2 instance. Every time we have a new EC2 instance, we can copy the https certificate file from S3 to EC2 instance using a very light weight docker container task. So far, this seems to be the best possible approach within the free tier EC2 instance of ECS type. We’re exploring this option.

If anyone has suggestions to us for a better approach, feel free to share your comments.

Interview questions on leadership skills

1 Point ContributorLeave a comment

Audience: anyone conducting/preparing for a technical leadership interview position.

What’s the best way to work with executives?

To work with executive, I prefer to follow the formal process per the company standards. Executive’s time is generally limited and they prefer a brief and to the point communication. Before reaching out to executives, it is important to understand if the topic worth the need of executive’s time. If it needs the time from executive, setting the expectation is important. I will determine if the topic to share with executives is to provide an information, or ask for any feedback or a decision. If it is about asking a decision, it’s helpful to prepare options. If a help is needed for an executive, it should be clear what kind of a help is needed. Executives prefer a clear communication and expectations. I will plan for it accordingly.

Is consensus always a good thing?

Consensus is not a good thing always. Depending on the situation, this strategy should be used. In my experience, there was a product that was working on a stable way but the company needed something completely innovative. To get such an environment, a leader challenged the existing environment. He did not have the consensus to proceed. But he did the right thing. In other situation, a consensus helped when I had to make a go/no-go decision on a success of failure of a User Acceptance Testing. Consensus helped as there were only few people in oppose and all such open issues could either be delayed or worked around.

What is the best way to work with customers and users?

The best way to work with customers or users is to understand their perspective and requirements. I believe the main goal of a product is to meet a customer’s or a user’s needs.

What kinds of people do you like to work with?

I like working with people who can collaborate well towards a common goal. This needs keeping individual approach secondary and thinking and planning for a common objective. This needs individuals to think beyond their individual achievements and focus on a teamwork. It is important to understand everyone’s perspective in the situations. The final decision must be that is needed for the success of the common goal.

What kind of people do you have a hard time working with?

I get hard time working with people who do not think for the success of the common goal. Instead, they may be driven from some other individual or team goals. To deal with such situation, I prefer to remind everyone about the common goal.

What would you do to get a team to stick to a schedule?

This depends on the type of the task. I will first provide a high level context of why we have a schedule and what do we want to achieve as a team within a schedule. Then, I will prepare a plan with everyone’s collaboration. Once everyone agrees with the plan and understand why we want to stick to the schedule, I suggest someone to schedule recurring meetings to check the progress status, blockers, and next steps. Ib the middle of the schedule, I prefer to remind everyone why this schedule is important , why we all agreed to it, and what are positive consequences of sticking to the schedule.

What’s the difference between leadership and management?

Leadership is about influencing, inspiring, and enabling others to make a positive impact. Management is about controlling group or entities to complete a goal on time.

Reference:

As I learn more, I will update this page. Thank you !

Behavioral interviews questions

1 Point ContributorLeave a comment

Who should read this article: Anyone appearing in or conducting in an interview for a developer, engineering manager, project/program, or a product manager or above role.

Disclaimer: Answers to these questions below are based on my experience and I may be wrong for some answers or you may have another opinion or an answer to it. It’s advisable to come up with your own answers.

Tell me about a time when you faced a challenge and overcame it.

Nugget: Sure. I will describe a challenge of integrating with a system that provides flag of who’s eligible to contact and who’s not.

Situation: We had a migration project on marketing campaigns, to reach out to customers, to get their feedback on products. We were using a central user identification system, to know who is eligible to be contacted. I was not getting the answer to it.

Result: After more than 5 attempts to convince this group with repeatedly asking for same information in different ways, finally they were convinced. We did it in a way that provided us information.

Tell me about how you interact with customers or clients?

In my various roles, this has been done in a. different way. In my current role, I interact with my internal customers as my business partners. We discuss about customer experiences, new changes to the business, and how it can impact my owned infrastructure or business area.

Talk about how you overcame product failures/challenges or poor feedback.

There was a time when our product was not meeting the customer’s needs. We met with the group of customers periodically and explained our limitations. Unfortunately these failures increased day by day to a point that these were not manageable. Then, collectively, we took an innovative approach to come up with a next suite of products that overcame the failures for a longer time.

Tell me about a time when you had to influence a team.

Nugget: Sure, let me tell you a time when I had to convince a team to increase the scope of testing in a project.

Situation: For a customer facing campaign emails go-live, engineering team was in a hurry, to go-live with the campaign. They had a pressure of reaching an end of life of an application. I still had to convince them to increase testing coverage as it was a customer facing application.

Action: I listened to their situation/proposal. In this situation, I had to disagree with proposed testing coverage. I convinced them with past data when a problem occured, due to lack of testing. I shared my past mistake when customers experienced an issue. I also told them that I would have agreed to their proposal if it was an internal release BUT we represent the company and can’t risk the customers.

Result: Unfortunately, there were tough deadlines, due to a legacy end of life. Team agreed to increase testing coverage upto 80% of what I sugegsted. I compromised to reduce my ask a little bit. But it was a good enough testing plan that gave both the teams a win/win feeling.

Tell me about a time when you have made a mistake.

I will describe a situation of a project in that I provided I underestimated the level of efforts for the development. This situation added burden on me to finish the development task within the given timeline. When I came up with it, I didn’t know it’s a large effort task. I estimated it as a medium task. To mitigate the delay of the go-live, I had to work extra hours, to develop the component within the expected timeline. Ultimately, after working long hours, I managed to get through it and complete the task in the given timeline.

How would you handle if two executives are asking to prioritize two different features and you can plan only for

one?

I will describe a prioritization exercise for the situation. I will go through the features requested by both the executives, understand level of efforts in implementing it, and business priority of it. After going through it, to both executives together, I will explain the current bandwidth of how much we can accomplish/deliver. To do that, I will utilize the appropriate forum that is suitable for such a prioritization. Then, after the discussions, I will decide the outcome of selecting one feature to deliver it on time with the expected quality.

Tell me about a time you used data to make a decision.

I will describe a situation when we determined the priority of a fix depending on the volume of the issues. In a customer campaign, we wanted to fix an issue. But we had limited bandwidth to fix the issue. We analyzed the data to understand the criticality of the issue. It turned out that the chances of issues were less than 0.5%. We had to delay this issue fix over other higher priority issues. So, this is how data helped us making the decision.

As I learn more, I will update this page. Thank you !

Databases basics

1 Point ContributorLeave a comment

Who should read it: It is for you if you are looking for an overview of this topic for a project, to conduct/appear in an interview, or in general. As we learn more, we will update this article.

Why we need distributed databases: 

  • Difficult to store entire data set into a single database
  • Single point of failure
  • Slow in performance
  • If we make a big computer, it will be more expensive

For the distributed database architectures, we have a master database and can have multiple secondary databases. Single node databases are classic databases like PostgreSQL, MySQL, etc. Distributed databases are made of multiple nodes. These are fault tolerant. Database clusters means multiple database instances. In general, we have leader nodes and follower nodes. Leader node is in charge of returning the final data results. Followers receive the data. If the leader node fails, a follower node can become the leader node.

Types of distributed databases:

  • Big compute databases: Split data across multiple nodes. These are suitable for analytical workloads.
  • High availability databases: are extremely fault tolerant. Each node has a full copy of the data.

Some key points about distributed databases:

  • Imbalance node: the problem when a node has more data load. Moving the data between nodes is slow.
  • Asking data from hard disk is slow. Asking data from RAM is fast.
  • Leader node examines the query. Leader node distributes the jobs to different nodes.
  • Sharding: It is a model in that all database instances acts as the primary databases. We segment data into multiple instances. Problem is that if we have more load on one segmented database, it will cause problems. Also, if we have to join data with two databases, we will have network connections.

CAP theorem:

  • C stands for Consistency. If we write information, we want to get same data. That is the consistency. To maintain consistency, data in the primary and secondary databases should replicate asap.
  • A stands for Availability. If we have two databases, if one machine goes down, as a whole system, users should be able to read or write.
  • P stands for Partitions. Partition tells us that even if connections to two machines not working, we should still be able to read/write the data.
  • RBDMS databases provide strong consistency. NoSQL databases generally prioritize availability, partition tolerance, and provide eventual consistency.

As per CAP theorem, generally, databases can achieve up to two features out of three. For distributed databases, assume network failures will be inevitable. So, for distributed databases, we need to choose between C and A.

PACELC theorem:

As per CAPLEC theorem, if partition happens, choose Availability and Consistency. Else, choose Latency and Consistency.

Other theorems:

BASE: BASE stands for Basically Available Soft state Eventually Consistent. NoSQL is an example of a BASE.

  • Basically Available: The system is guaranteed to be available in event of failure.
  • Soft State: The state of the data could change without application interactions due to eventual consistency.
  • Eventual Consistency: The system will be eventually consistent after the application input. The data will be replicated to different nodes and will eventually reach a consistent state. But the consistency is not guaranteed at a transaction level.

Indexes:

Database index is a data structure that helps to retrieve data faster from a table. Indexes are like library catalog that helps to know the location of a book. For more about index, refer here.

Relational databases:

Relational databases store data in rows and columns. Some famous relational databases are MySQL, Oracle, and Postgres.

Advantages of relational databases:

  1. Well defined relationships and structured: data in relational databases is structured, with foreign and primary key constraints. It helps in organizing the data. Defined relationships and structure also helps in retrieving the data effectively.
  2. ACID (Atomicity, Consistency, Isolation, Durability): As relational databases support ACID properties, it’s helpful in ensuring the data changes for a transaction.

Disadvantages of relational databases:

  1. Rigidity due to structured data: As the data is well defined and structured, it’s not easy to store a new data set for that a structure is unknown. For example, to add a new column into a table, the table has to be changed, to support it.
  2. Difficult to scale: scaling means supporting more volume of data. For relational databases, scaling is difficult. For read-only operations, it’s easier to replicate the data. For write operations, a general approach is to add more capacity (vertical scaling) to the primary database server, which is costlier than replacing read-only databases.

NoSQL Databases:

There are multiple types of NoSQL databases like:

  • Key-Value storage type: Data is stored in key-value pairs in arrays. Some examples of such databases are Redis and Dynamo databases.
  • Document databases: In these databases, data is stored in the documents. A collection is a group of documents. Each document can have a different structure. An example of such a database is MongoDB.
  • Wide-column databases: In these databases, the number of columns can vary per row in the same table. We can consider it as a two dimensional key-value storage. Some examples of wide-column databases are Cassandra and HBase. For more, refer here.
  • Graph databases: These represent data in a form of a graph. Examples of such databases are Neo4J and Infinite graph.
  • Some NoSQL databases:
    • Couchbase: It is a NoSQL database that stores the data either in key/value pair or in JSON document format. In a traditional database model, we begin with a schema. We add tables and the columns in the tables.
    • MongoDB: MongoDB is an open source document database. It works on concept of collections and documents. A collection is a document which is equivalent to an RDBMS table. A document is a set of key-value pairs. It is a schema-less database. It is easy to scale. It is a good choice for a Big Data need.

Advantages of NoSQL databases:

  • Flexibility with unstructured data: As the data in NoSQL databases is unstructured, these databases provide more flexibility to store the data.
  • Horizontal scaling: Horizontal scaling means distributing data into multiple server instances. Data in NoSQL databases are distributed, by using sharding. These databases support horizontal scaling for both, read and write operations.

SQL versus NoSQL databases:

  • Storage: Data in SQL databases is stored in rows. NoSQL databases have different data storage models like key-value or graph
  • Schema: SQL databases have a fixed schema. noSQL databases can have different schemas.
  • Querying: SQL databases use Structural Query Language (SQL), to retrieve the data. NoSQL databases uses UnQL (Unstructured Query Language). NoSQL are focused as a collection of documents.
  • Scalability: Horizontal scaling in SQL databases is difficult as compared to NoSQL databases.
  • Reliability: Most SQL databases are reliable and ACID compliant. Whereas, NoSQL databases may compromise reliability and ACID compliance.
  • Language: SQL databases use transactional SQL. They support core ANSI/ISO language elements. Whereas, NoSQL databases are not limited to one particular language. For example, MongoDb uses Javascript based query language.

 How to choose between SQL and NoSQL databases:

  • Consider SQL databases when:
    • Data is structured and structure is not changed frequently.
    • Supporting transaction-oriented use cases.
    • No need to scale the database.
  • Consider NoSQL databases when:
    • Data is not structured and the structure can change frequently.
    • A flexibility of dynamic schema is needed.
    • We anticipate the scaling of the database in the future.
    • Level of data integrity is not needed.

Other databases terms:

  • Purpose built databases: There are many options like relational database, key-value database, document database, graph database, in-memory database, time series database, and ledger database. Depending on the situation, today’s application developers need to pick a right database for the use case, by analyzing pros and cons of the situation.

Partitioning and Sharding: it is the process of splitting the data into columns or features. Vertical partitioning splits the data into the same database by columns or features within the same tables. Horizontal partitioning splits the table data to multiple shards ( e.g. multiple database locations). In case of sharding, a table may have a customer ID 1 on one server and customer ID 2 on another server. In case of partitioning, other customer Ids 1 and 2 are on the same database servers and in the same table.

Distributed transaction types:

  • Two-phase commit (2PC): In case of 2 Phase commit, there is a coordinator that prepares multiple transactions. Then, the coordinator either commits or rolls back all transactions together. While preparing each transactions, the database table rows (that to be updated in DB) are locked using local transactions. This prevents any updates during the 2-phase transaction. We also need to plan for a time limit for each transaction so that the coordinator is waiting to commit or rollback within a defined time period.
  • 3 phase commit: First phase is that the coordinator asks if it’s ok to commit. Second phase is to pre-commit the transaction. Third step is to commit a transaction. I’m yet to learn more and add more notes about it.
  • Try-Confirm/Cancel (TC/C): In the first phase, the coordinator asks all databases to reserve resources. In the second phase, the coordinator captures the replies from all databases. If the response is yes, the coordinator asks all databases to confirm the transaction. If any of the databases respond as no, the coordinator asks all databases to cancel the transaction.
  • Saga: It’s an asynchronous way to achieve the transaction. It’s an event driven process. Micro-services generally use Saga as their default choice. In Saga, all operations are executed in a sequence. When one operation finishes, then the next operation is executed. For the rollback purpose, we need to prepare double operations: one for execution and the another one for the rollback. To coordinate operations in Saga, there are two ways:
    • Choreography: All services do their jobs by subscribing to other services’ events. This is a decentralized coordination.
    • Orchestration: In this way, there is a single coordinator to instruct all services in a defined order.

References: