Large-Scale Systems follows one four-tier system, browser to web app to business app to database, from a single server to a global fleet. Forty-six episodes, one idea each: replication, sharding, load balancing, the laws that cap concurrency, caching, failure and the nines, security end to end, deployment from hand-built boxes to Kubernetes, the products that implement all of it, and the foundations that tie it together. Every pattern is named with the trade it actually costs. Two hosts, no slides, no exam. Large-Scale Systems follows one four-tier system, browser to web app to business app to database, from a single server to a global fleet. Forty-six episodes, one idea each: replication, sharding, load balancing, the laws that cap concurrency, caching, failure and the nines, security end to end, deployment from hand-built boxes to Kubernetes, the products that implement all of it, and the foundations that tie it together. Every pattern is named with the trade it actually costs. Two hosts, no slides, no exam. From one server to a fleet, and what that actually costs large-scale systems, software architecture, distributed systems, scalability, reliability, system design, cloud https://www.cloudadorn.com/ en-us Cloudadorn Academy false Cloudadorn Academy Cloudadorn Academy info@cloudadorn.com serial yes 7ae922c7-4581-5b17-b74a-6aededa03b8d no Sun, 30 Aug 2026 12:00:00 +0000 Large-Scale Systems follows one four-tier system, browser to web app to business app to database, from a single server to a global fleet. Forty-six episodes, one idea each: replication, sharding, load balancing, the laws that cap concurrency, caching, failure and the nines, security end to end, deployment from hand-built boxes to Kubernetes, the products that implement all of it, and the foundations that tie it together. Every pattern is named with the trade it actually costs. Two hosts, no slides, no exam.

]]> cloudadorn-academy.large-scale-systems.s1e01 Sun, 30 Aug 2026 12:00:00 +0000 1004 1 1 full false performance, scalability, latency, throughput, horizontal scaling One box, one user, and the difference between being fast and being ready for a crowd. Shelley and Rob start the four-tier system the whole season follows and name the two enemies every later pattern exists to remove. One box, one user, and the difference between being fast and being ready for a crowd. Shelley and Rob start the four-tier system the whole season follows and name the two enemies every later pattern exists to remove. One box, one user, and the difference between being fast and being ready for a crowd. Shelley and Rob start the four-tier system the whole season follows and name the two enemies every later pattern exists to remove.

]]> cloudadorn-academy.large-scale-systems.s1e02 Sun, 30 Aug 2026 12:02:00 +0000 932 2 1 full false reverse proxy, front door, load balancing, connection, edge The first thing you put in front of the box: one address the world knows, a pool it never sees. Why the front door breaks the connection in two, and why it can never be the only one. The first thing you put in front of the box: one address the world knows, a pool it never sees. Why the front door breaks the connection in two, and why it can never be the only one. The first thing you put in front of the box: one address the world knows, a pool it never sees. Why the front door breaks the connection in two, and why it can never be the only one.

]]> cloudadorn-academy.large-scale-systems.s1e03 Sun, 30 Aug 2026 12:04:00 +0000 956 3 1 full false replication, stateless, sticky sessions, session store, scale out Clone the web tier and the service tier, and discover that copies only help if none of them keeps your file in its own head. Sticky sessions, externalized state, and where the bottleneck moves next. Clone the web tier and the service tier, and discover that copies only help if none of them keeps your file in its own head. Sticky sessions, externalized state, and where the bottleneck moves next. Clone the web tier and the service tier, and discover that copies only help if none of them keeps your file in its own head. Sticky sessions, externalized state, and where the bottleneck moves next.

]]> cloudadorn-academy.large-scale-systems.s1e04 Sun, 30 Aug 2026 12:06:00 +0000 1002 4 1 full false database replication, read replica, multi-region, active-active, eventual consistency The one tier that remembers, copied. Read replicas, single-writer and multi-writer, synchronous and asynchronous, and the honest sentence behind every active-active region: some data loss is possible. The one tier that remembers, copied. Read replicas, single-writer and multi-writer, synchronous and asynchronous, and the honest sentence behind every active-active region: some data loss is possible. The one tier that remembers, copied. Read replicas, single-writer and multi-writer, synchronous and asynchronous, and the honest sentence behind every active-active region: some data loss is possible.

]]> cloudadorn-academy.large-scale-systems.s1e05 Sun, 30 Aug 2026 12:08:00 +0000 984 5 1 full false microservices, message queue, asynchronous, api gateway, load leveling Split the business tier by job, then put a queue between accepting work and doing it. Why you can stop paying for peak, and what you start paying in instead. Split the business tier by job, then put a queue between accepting work and doing it. Why you can stop paying for peak, and what you start paying in instead. Split the business tier by job, then put a queue between accepting work and doing it. Why you can stop paying for peak, and what you start paying in instead.

]]> cloudadorn-academy.large-scale-systems.s1e06 Sun, 30 Aug 2026 12:10:00 +0000 998 6 1 full false sharding, partitioning, shard key, consistent hashing, database per service Split the data itself: by function, then by row. The shard key decides which questions stay cheap, and database-per-service is the moment cross-service transactions stop being free. Split the data itself: by function, then by row. The shard key decides which questions stay cheap, and database-per-service is the moment cross-service transactions stop being free. Split the data itself: by function, then by row. The shard key decides which questions stay cheap, and database-per-service is the moment cross-service transactions stop being free.

]]> cloudadorn-academy.large-scale-systems.s1e07 Sun, 30 Aug 2026 12:12:00 +0000 984 7 1 full false load balancer, layer 7, service discovery, dns, gslb One address per tier, a registry of who is healthy, and the rules for spreading the work. Layer four versus layer seven, the four algorithms, and why DNS is the cheapest balancer and the slowest to forget. One address per tier, a registry of who is healthy, and the rules for spreading the work. Layer four versus layer seven, the four algorithms, and why DNS is the cheapest balancer and the slowest to forget. One address per tier, a registry of who is healthy, and the rules for spreading the work. Layer four versus layer seven, the four algorithms, and why DNS is the cheapest balancer and the slowest to forget.

]]> cloudadorn-academy.large-scale-systems.s1e08 Sun, 30 Aug 2026 12:14:00 +0000 944 8 1 full false auto-scaling, elasticity, monitoring, warm-up, capacity Let the fleet grow and shrink itself. The feedback loop that does it, the warm-up minute it cannot buy back, and which kind of peak wants pre-warming, a schedule, or a bigger box. Let the fleet grow and shrink itself. The feedback loop that does it, the warm-up minute it cannot buy back, and which kind of peak wants pre-warming, a schedule, or a bigger box. Let the fleet grow and shrink itself. The feedback loop that does it, the warm-up minute it cannot buy back, and which kind of peak wants pre-warming, a schedule, or a bigger box.

]]> cloudadorn-academy.large-scale-systems.s1e09 Sun, 30 Aug 2026 12:16:00 +0000 1004 9 1 full false microservices, saga pattern, distributed transactions, two-phase commit, compensation Each service owns its data, so a single business change becomes a chain of local steps, each with its own undo. Two-phase commit versus the saga, and why compensation is not a rollback. Each service owns its data, so a single business change becomes a chain of local steps, each with its own undo. Two-phase commit versus the saga, and why compensation is not a rollback. Each service owns its data, so a single business change becomes a chain of local steps, each with its own undo. Two-phase commit versus the saga, and why compensation is not a rollback.

]]> cloudadorn-academy.large-scale-systems.s1e10 Sun, 30 Aug 2026 12:18:00 +0000 980 10 1 full false latency, throughput, queueing, percentiles, saturation Every slow system is a queue forming somewhere. Three reasons a queue forms, three levers to fix it, four signals to watch, and why the average hides the warning the tail is giving you. Every slow system is a queue forming somewhere. Three reasons a queue forms, three levers to fix it, four signals to watch, and why the average hides the warning the tail is giving you. Every slow system is a queue forming somewhere. Three reasons a queue forms, three levers to fix it, four signals to watch, and why the average hides the warning the tail is giving you.

]]> cloudadorn-academy.large-scale-systems.s1e11 Sun, 30 Aug 2026 12:20:00 +0000 1008 11 1 full false latency, memory, disk, network round trip, garbage collection One request's time is network plus memory plus disk plus CPU, and the rungs are orders of magnitude apart. Three laws, the cost of a handshake, and the tray that gets cleaned less often but for longer. One request's time is network plus memory plus disk plus CPU, and the rungs are orders of magnitude apart. Three laws, the cost of a handshake, and the tray that gets cleaned less often but for longer. One request's time is network plus memory plus disk plus CPU, and the rungs are orders of magnitude apart. Three laws, the cost of a handshake, and the tray that gets cleaned less often but for longer.

]]> cloudadorn-academy.large-scale-systems.s1e12 Sun, 30 Aug 2026 12:22:00 +0000 962 12 1 full false amdahl's law, universal scalability law, contention, coherency, concurrency The math that caps what adding workers can do, and the math that makes more workers slower. The serial fraction, the coherency tax, and the knee you design to stay under. The math that caps what adding workers can do, and the math that makes more workers slower. The serial fraction, the coherency tax, and the knee you design to stay under. The math that caps what adding workers can do, and the math that makes more workers slower. The serial fraction, the coherency tax, and the knee you design to stay under.

]]> cloudadorn-academy.large-scale-systems.s1e13 Sun, 30 Aug 2026 12:24:00 +0000 969 13 1 full false locking, optimistic locking, compare and swap, thread pool, deadlock Where workers actually collide: the accept queue, the pools, the hardware, the locks. Hold locks shorter, collide less often, pick optimistic or pessimistic, and sometimes stop sharing the thing at all. Where workers actually collide: the accept queue, the pools, the hardware, the locks. Hold locks shorter, collide less often, pick optimistic or pessimistic, and sometimes stop sharing the thing at all. Where workers actually collide: the accept queue, the pools, the hardware, the locks. Hold locks shorter, collide less often, pick optimistic or pessimistic, and sometimes stop sharing the thing at all.

]]> cloudadorn-academy.large-scale-systems.s1e14 Sun, 30 Aug 2026 12:26:00 +0000 952 14 1 full false cache coherence, false sharing, memory barrier, volatile, cache line Even one machine's cores pay to agree. Cache lines, false sharing, what volatile and synchronized actually promise, and why shared mutable state is the physical reason scaling bends. Even one machine's cores pay to agree. Cache lines, false sharing, what volatile and synchronized actually promise, and why shared mutable state is the physical reason scaling bends. Even one machine's cores pay to agree. Cache lines, false sharing, what volatile and synchronized actually promise, and why shared mutable state is the physical reason scaling bends.

]]> cloudadorn-academy.large-scale-systems.s1e15 Sun, 30 Aug 2026 12:28:00 +0000 926 15 1 full false caching, cache-control, etag, hit ratio, ttl Trade a bounded staleness for latency, throughput, and money. The cache layers, the HTTP headers that run them, hit ratio, the time-to-live dial, and why a cache outage hits harder than no cache. Trade a bounded staleness for latency, throughput, and money. The cache layers, the HTTP headers that run them, hit ratio, the time-to-live dial, and why a cache outage hits harder than no cache. Trade a bounded staleness for latency, throughput, and money. The cache layers, the HTTP headers that run them, hit ratio, the time-to-live dial, and why a cache outage hits harder than no cache.

]]> cloudadorn-academy.large-scale-systems.s1e16 Sun, 30 Aug 2026 12:30:00 +0000 931 16 1 full false distributed systems, partial failure, cascading failure, gray failure, incident Failure is the steady state: partial, independent, and cascading. Gray failure, the spectrum of fault models, and the outages that started with a change and stayed long because the tools shared the fate. Failure is the steady state: partial, independent, and cascading. Gray failure, the spectrum of fault models, and the outages that started with a change and stayed long because the tools shared the fate. Failure is the steady state: partial, independent, and cascading. Gray failure, the spectrum of fault models, and the outages that started with a change and stayed long because the tools shared the fate.

]]> cloudadorn-academy.large-scale-systems.s1e17 Sun, 30 Aug 2026 12:32:00 +0000 925 17 1 full false availability, reliability, nines, slo, error budget Right answer over time versus open right now. The two formulas, why time-to-repair matters as much as time-between-failures, and the downtime budget behind each extra nine. Right answer over time versus open right now. The two formulas, why time-to-repair matters as much as time-between-failures, and the downtime budget behind each extra nine. Right answer over time versus open right now. The two formulas, why time-to-repair matters as much as time-between-failures, and the downtime budget behind each extra nine.

]]> cloudadorn-academy.large-scale-systems.s1e18 Sun, 30 Aug 2026 12:34:00 +0000 929 18 1 full false redundancy, single point of failure, high availability, fault isolation, n plus one Spare copies of everything, including the thing that hands out the work. Active, passive, cold; the ladder from component to zone to region; and why identical spares share identical bugs. Spare copies of everything, including the thing that hands out the work. Active, passive, cold; the ladder from component to zone to region; and why identical spares share identical bugs. Spare copies of everything, including the thing that hands out the work. Active, passive, cold; the ladder from component to zone to region; and why identical spares share identical bugs.

]]> cloudadorn-academy.large-scale-systems.s1e19 Sun, 30 Aug 2026 12:36:00 +0000 954 19 1 full false health checks, heartbeat, fault detection, split brain, monitoring Nothing recovers from a failure it never noticed. Outside-in checks versus inside-out heartbeats, the trade between detecting fast and crying wolf, split brain, and who watches the watcher. Nothing recovers from a failure it never noticed. Outside-in checks versus inside-out heartbeats, the trade between detecting fast and crying wolf, split brain, and who watches the watcher. Nothing recovers from a failure it never noticed. Outside-in checks versus inside-out heartbeats, the trade between detecting fast and crying wolf, split brain, and who watches the watcher.

]]> cloudadorn-academy.large-scale-systems.s1e20 Sun, 30 Aug 2026 12:38:00 +0000 966 20 1 full false failover, hot standby, disaster recovery, rpo, rto Replace the stateless, fail over the stateful. Hot, warm, and cold standby, the floating address that moves, why replication carries corruption with it, and why untested failover is a rumor. Replace the stateless, fail over the stateful. Hot, warm, and cold standby, the floating address that moves, why replication carries corruption with it, and why untested failover is a rumor. Replace the stateless, fail over the stateful. Hot, warm, and cold standby, the floating address that moves, why replication carries corruption with it, and why untested failover is a rumor.

]]> cloudadorn-academy.large-scale-systems.s1e21 Sun, 30 Aug 2026 12:40:00 +0000 986 21 1 full false timeouts, retries, circuit breaker, back-pressure, load shedding Stop one slow dependency from taking the fleet with it. Always time out, retry only what is transient and safe to repeat, and the three-state breaker that fails fast and probes its way back. Stop one slow dependency from taking the fleet with it. Always time out, retry only what is transient and safe to repeat, and the three-state breaker that fails fast and probes its way back. Stop one slow dependency from taking the fleet with it. Always time out, retry only what is transient and safe to repeat, and the three-state breaker that fails fast and probes its way back.

]]> cloudadorn-academy.large-scale-systems.s1e22 Sun, 30 Aug 2026 12:42:00 +0000 1014 22 1 full false threat modeling, security, defense in depth, trust boundary, stride Four pillars and one map of trust boundaries on the same four-tier system. Which control answers which concern, and why no single failure is allowed to be the whole story. Four pillars and one map of trust boundaries on the same four-tier system. Which control answers which concern, and why no single failure is allowed to be the whole story. Four pillars and one map of trust boundaries on the same four-tier system. Which control answers which concern, and why no single failure is allowed to be the whole story.

]]> cloudadorn-academy.large-scale-systems.s1e23 Sun, 30 Aug 2026 12:44:00 +0000 960 23 1 full false cryptography, symmetric encryption, public key, hashing, key management Three primitives, four properties. The fast one that cannot deliver its own key, the slow one that can, the one-way one that is not encryption, and why the algorithm is rarely what breaks. Three primitives, four properties. The fast one that cannot deliver its own key, the slow one that can, the one-way one that is not encryption, and why the algorithm is rarely what breaks. Three primitives, four properties. The fast one that cannot deliver its own key, the slow one that can, the one-way one that is not encryption, and why the algorithm is rarely what breaks.

]]> cloudadorn-academy.large-scale-systems.s1e24 Sun, 30 Aug 2026 12:46:00 +0000 842 24 1 full false tls, pki, certificates, handshake, mtls How two strangers agree on a secret across a hostile network, and who vouches for whom. Signatures, certificates, the chain of trust, the handshake, and the certificate that expired on a whole country. How two strangers agree on a secret across a hostile network, and who vouches for whom. Signatures, certificates, the chain of trust, the handshake, and the certificate that expired on a whole country. How two strangers agree on a secret across a hostile network, and who vouches for whom. Signatures, certificates, the chain of trust, the handshake, and the certificate that expired on a whole country.

]]> cloudadorn-academy.large-scale-systems.s1e25 Sun, 30 Aug 2026 12:48:00 +0000 933 25 1 full false firewall, dmz, network segmentation, zero trust, least privilege Which zone may talk to which. The four-part rule a firewall checks, the one zone the internet is allowed to touch, why nothing in the lobby reaches the vault, and where zero trust takes it next. Which zone may talk to which. The four-part rule a firewall checks, the one zone the internet is allowed to touch, why nothing in the lobby reaches the vault, and where zero trust takes it next. Which zone may talk to which. The four-part rule a firewall checks, the one zone the internet is allowed to touch, why nothing in the lobby reaches the vault, and where zero trust takes it next.

]]> cloudadorn-academy.large-scale-systems.s1e26 Sun, 30 Aug 2026 12:50:00 +0000 936 26 1 full false authentication, sessions, tokens, single sign-on, ldap Who are you. Credentials in transit, in storage, and at verification; sessions you can cancel versus tokens you can check anywhere; and single sign-on as one login that concentrates the risk. Who are you. Credentials in transit, in storage, and at verification; sessions you can cancel versus tokens you can check anywhere; and single sign-on as one login that concentrates the risk. Who are you. Credentials in transit, in storage, and at verification; sessions you can cancel versus tokens you can check anywhere; and single sign-on as one login that concentrates the risk.

]]> cloudadorn-academy.large-scale-systems.s1e27 Sun, 30 Aug 2026 12:52:00 +0000 933 27 1 full false authorization, rbac, oauth2, jwt, access control What may you do. Roles and the point that enforces them, delegation that is not identity, the two flows that matter, and a signed ticket every service can read and nobody can take back early. What may you do. Roles and the point that enforces them, delegation that is not identity, the two flows that matter, and a signed ticket every service can read and nobody can take back early. What may you do. Roles and the point that enforces them, delegation that is not identity, the two flows that matter, and a signed ticket every service can read and nobody can take back early.

]]> cloudadorn-academy.large-scale-systems.s1e28 Sun, 30 Aug 2026 12:54:00 +0000 939 28 1 full false data security, encryption at rest, sql injection, xss, csrf Protect the bytes at rest and the request in flight. Encrypting the disk versus encrypting the column, the key that wraps the key, the three classic web attacks and their fixes, and the breaches that walked in through an allowed door. Protect the bytes at rest and the request in flight. Encrypting the disk versus encrypting the column, the key that wraps the key, the three classic web attacks and their fixes, and the breaches that walked in through an allowed door. Protect the bytes at rest and the request in flight. Encrypting the disk versus encrypting the column, the key that wraps the key, the three classic web attacks and their fixes, and the breaches that walked in through an allowed door.

]]> cloudadorn-academy.large-scale-systems.s1e29 Sun, 30 Aug 2026 12:56:00 +0000 930 29 1 full false deployment, devops, operations, rollback, environments Application, infrastructure, operations, planned together. The loop that ships the fleet, the five signals you watch it with, and why the rollback is fast and the fix is slow. Application, infrastructure, operations, planned together. The loop that ships the fleet, the five signals you watch it with, and why the rollback is fast and the fix is slow. Application, infrastructure, operations, planned together. The loop that ships the fleet, the five signals you watch it with, and why the rollback is fast and the fix is slow.

]]> cloudadorn-academy.large-scale-systems.s1e30 Sun, 30 Aug 2026 12:58:00 +0000 942 30 1 full false infrastructure as code, provisioning, configuration management, idempotent, automation From a hand-built box to a recipe you can run again. The stack from operating system up, provisioning versus configuration, and the three adjectives a good script has to earn. From a hand-built box to a recipe you can run again. The stack from operating system up, provisioning versus configuration, and the three adjectives a good script has to earn. From a hand-built box to a recipe you can run again. The stack from operating system up, provisioning versus configuration, and the three adjectives a good script has to earn.

]]> cloudadorn-academy.large-scale-systems.s1e31 Sun, 30 Aug 2026 13:00:00 +0000 974 31 1 full false virtual machines, containers, docker, hypervisor, isolation Virtualize the hardware or virtualize the operating system. Why one boots and the other starts, what each isolates, the layered image that makes containers cheap, and why production runs one inside the other. Virtualize the hardware or virtualize the operating system. Why one boots and the other starts, what each isolates, the layered image that makes containers cheap, and why production runs one inside the other. Virtualize the hardware or virtualize the operating system. Why one boots and the other starts, what each isolates, the layered image that makes containers cheap, and why production runs one inside the other.

]]> cloudadorn-academy.large-scale-systems.s1e32 Sun, 30 Aug 2026 13:02:00 +0000 1006 32 1 full false cloud, managed services, serverless, vpc, cost governance The same stack rented by the hour. Mapping every box to a managed service on three clouds, what serverless removes and what it charges, and where to draw the build-versus-buy line. The same stack rented by the hour. Mapping every box to a managed service on three clouds, what serverless removes and what it charges, and where to draw the build-versus-buy line. The same stack rented by the hour. Mapping every box to a managed service on three clouds, what serverless removes and what it charges, and where to draw the build-versus-buy line.

]]> cloudadorn-academy.large-scale-systems.s1e33 Sun, 30 Aug 2026 13:04:00 +0000 938 33 1 full false kubernetes, orchestration, pods, control plane, reconciliation Declare the fleet you want and a loop makes reality match. Pods, the control plane, and how every feature maps onto a pain the season already met. Declare the fleet you want and a loop makes reality match. Pods, the control plane, and how every feature maps onto a pain the season already met. Declare the fleet you want and a loop makes reality match. Pods, the control plane, and how every feature maps onto a pain the season already met.

]]> cloudadorn-academy.large-scale-systems.s1e34 Sun, 30 Aug 2026 13:06:00 +0000 902 34 1 full false zero downtime, blue-green, canary, rolling update, feature flags Replace the software while the service stays up. Recreate, rolling, canary, blue-green, and the experiment that is not a safety check; flags that separate deploying from releasing; and the database that is always the hard part. Replace the software while the service stays up. Recreate, rolling, canary, blue-green, and the experiment that is not a safety check; flags that separate deploying from releasing; and the database that is always the hard part. Replace the software while the service stays up. Recreate, rolling, canary, blue-green, and the experiment that is not a safety check; flags that separate deploying from releasing; and the database that is always the hard part.

]]> cloudadorn-academy.large-scale-systems.s1e35 Sun, 30 Aug 2026 13:08:00 +0000 964 35 1 full false nginx, apache, node.js, event loop, web server The front of the system as products. Threads per request versus the event loop, why Nginx sits in front of Apache, what Node is good and bad at, and where the managed door makes sense. The front of the system as products. Threads per request versus the event loop, why Nginx sits in front of Apache, what Node is good and bad at, and where the managed door makes sense. The front of the system as products. Threads per request versus the event loop, why Nginx sits in front of Apache, what Node is good and bad at, and where the managed door makes sense.

]]> cloudadorn-academy.large-scale-systems.s1e36 Sun, 30 Aug 2026 13:10:00 +0000 988 36 1 full false cdn, object storage, edge cache, cache hit ratio, thundering herd The durable origin and the edge cache in front of it. What object storage promises, what the edge serves and what it must fetch, the herd that forms on launch day, and why smaller bytes are a budget line. The durable origin and the edge cache in front of it. What object storage promises, what the edge serves and what it must fetch, the herd that forms on launch day, and why smaller bytes are a budget line. The durable origin and the edge cache in front of it. What object storage promises, what the edge serves and what it must fetch, the herd that forms on launch day, and why smaller bytes are a budget line.

]]> cloudadorn-academy.large-scale-systems.s1e37 Sun, 30 Aug 2026 13:12:00 +0000 1028 37 1 full false memcached, redis, cache-aside, consistent hashing, in-memory The two cache products, chosen by access pattern. Cache-aside, the ring that lets you add a node without losing everything, what one forgets and the other remembers, and why the hard part is never speed. The two cache products, chosen by access pattern. Cache-aside, the ring that lets you add a node without losing everything, what one forgets and the other remembers, and why the hard part is never speed. The two cache products, chosen by access pattern. Cache-aside, the ring that lets you add a node without losing everything, what one forgets and the other remembers, and why the hard part is never speed.

]]> cloudadorn-academy.large-scale-systems.s1e38 Sun, 30 Aug 2026 13:14:00 +0000 974 38 1 full false rabbitmq, kafka, pub sub, message broker, event streaming The brokers. Push versus pull, delete-on-acknowledge versus the retained log, fire-and-forget, the cloud equivalents on three doors, and the one consumer rule at-least-once delivery forces. The brokers. Push versus pull, delete-on-acknowledge versus the retained log, fire-and-forget, the cloud equivalents on three doors, and the one consumer rule at-least-once delivery forces. The brokers. Push versus pull, delete-on-acknowledge versus the retained log, fire-and-forget, the cloud equivalents on three doors, and the one consumer rule at-least-once delivery forces.

]]> cloudadorn-academy.large-scale-systems.s1e39 Sun, 30 Aug 2026 13:16:00 +0000 1003 39 1 full false relational database, acid, replication lag, sharding, vitess What the consistency-first store gives up when it grows. The single-node ceiling, the cost hidden inside each strength, reads that scale and writes that do not, and the teams that sharded the relational model instead of leaving it. What the consistency-first store gives up when it grows. The single-node ceiling, the cost hidden inside each strength, reads that scale and writes that do not, and the teams that sharded the relational model instead of leaving it. What the consistency-first store gives up when it grows. The single-node ceiling, the cost hidden inside each strength, reads that scale and writes that do not, and the teams that sharded the relational model instead of leaving it.

]]> cloudadorn-academy.large-scale-systems.s1e40 Sun, 30 Aug 2026 13:18:00 +0000 991 40 1 full false nosql, dynamodb, cassandra, mongodb, bigtable Four families that made the scale trade explicit. The ring, the tablet, the column family, and the document; what each gives up, and the two decisions that make or break any of them. Four families that made the scale trade explicit. The ring, the tablet, the column family, and the document; what each gives up, and the two decisions that make or break any of them. Four families that made the scale trade explicit. The ring, the tablet, the column family, and the document; what each gives up, and the two decisions that make or break any of them.

]]> cloudadorn-academy.large-scale-systems.s1e41 Sun, 30 Aug 2026 13:20:00 +0000 1020 41 1 full false hadoop, spark, elasticsearch, logstash, analytics The tier that collects, stores, searches, and processes what the application cannot. Shippers and sorters, the search index, the distributed file system, map-reduce, and the engine that kept the data in memory. The tier that collects, stores, searches, and processes what the application cannot. Shippers and sorters, the search index, the distributed file system, map-reduce, and the engine that kept the data in memory. The tier that collects, stores, searches, and processes what the application cannot. Shippers and sorters, the search index, the distributed file system, map-reduce, and the engine that kept the data in memory.

]]> cloudadorn-academy.large-scale-systems.s1e42 Sun, 30 Aug 2026 13:22:00 +0000 926 42 1 full false quality attributes, non-functional requirements, architectural drivers, trade-offs, constraints Requirements that shape architecture, made measurable. Features versus attributes versus constraints, why several attributes contradict, and the number you promise that you now have to measure forever. Requirements that shape architecture, made measurable. Features versus attributes versus constraints, why several attributes contradict, and the number you promise that you now have to measure forever. Requirements that shape architecture, made measurable. Features versus attributes versus constraints, why several attributes contradict, and the number you promise that you now have to measure forever.

]]> cloudadorn-academy.large-scale-systems.s1e43 Sun, 30 Aug 2026 13:24:00 +0000 923 43 1 full false api design, rest, rpc, grpc, api gateway The contract between services. Functions over a wire versus resources over verbs, what a good contract promises, and the single front door that does everything except business logic. The contract between services. Functions over a wire versus resources over verbs, what a good contract promises, and the single front door that does everything except business logic. The contract between services. Functions over a wire versus resources over verbs, what a good contract promises, and the single front door that does everything except business logic.

]]> cloudadorn-academy.large-scale-systems.s1e44 Sun, 30 Aug 2026 13:26:00 +0000 983 44 1 full false cap theorem, consistency, availability, partition tolerance, pacelc Partitions are not optional, so the real choice is who waits when the line is cut. What the theorem actually says, two kinds of consistency that share a word, and the dial the whole season was quietly turning. Partitions are not optional, so the real choice is who waits when the line is cut. What the theorem actually says, two kinds of consistency that share a word, and the dial the whole season was quietly turning. Partitions are not optional, so the real choice is who waits when the line is cut. What the theorem actually says, two kinds of consistency that share a word, and the dial the whole season was quietly turning.

]]> cloudadorn-academy.large-scale-systems.s1e45 Sun, 30 Aug 2026 13:28:00 +0000 963 45 1 full false event-driven architecture, event sourcing, cqrs, materialized view, message broker Events as facts that already happened, the log as the truth, reads split from writes. Event sourcing, CQRS, materialized views, and the two things people confuse with them. Events as facts that already happened, the log as the truth, reads split from writes. Event sourcing, CQRS, materialized views, and the two things people confuse with them. Events as facts that already happened, the log as the truth, reads split from writes. Event sourcing, CQRS, materialized views, and the two things people confuse with them.

]]> cloudadorn-academy.large-scale-systems.s1e46 Sun, 30 Aug 2026 13:30:00 +0000 948 46 1 full false stream processing, windowing, watermark, lambda architecture, event time Finite questions of infinite data. Four kinds of window, three clocks that disagree, what to do with the event that arrives late, and the two-layer architecture that answers fast and corrects later. Finite questions of infinite data. Four kinds of window, three clocks that disagree, what to do with the event that arrives late, and the two-layer architecture that answers fast and corrects later. Finite questions of infinite data. Four kinds of window, three clocks that disagree, what to do with the event that arrives late, and the two-layer architecture that answers fast and corrects later.

]]>