What I mean by “distributed system”:

a group of computers cooperating to provide a service

this class is mostly about infrastructure services

  • e.g. storage for big web sites, MapReduce, peer-to-peer sharing

lots of important infrastructure is distributed

Why do people build distributed systems?

  • to increase capacity via parallel processing
  • to tolerate faults via replication
  • to match distribution of physical devices e.g. sensors
  • to achieve security via isolation

But it’s not easy:

  • many concurrent parts, complex interactions
  • must cope with partial failure
  • tricky to realize performance potential

Why study this topic?

  • interesting — hard problems, powerful solutions
  • widely used — driven by the rise of big Web sites
  • active research area — important unsolved problems
  • challenging to build — you’ll do it in the labs

COURSE STRUCTURE

http://pdos.csail.mit.edu/6.824

Course components:

  • lectures
  • papers
  • two exams
  • labs
  • final project (optional)

Lectures:

  • big ideas, paper discussion, lab guidance
  • will be video-taped, available online

Papers:

  • there’s a paper assigned for almost every lecture
  • research papers, some classic, some new
  • problems, ideas, implementation details, evaluation
  • please read papers before class!
  • each paper has a short question for you to answer
  • and we ask you to send us a question you have about the paper
  • submit answer and question before start of lecture

Labs:

  • goal: deeper understanding of some important ideas
  • goal: experience with distributed programming

Lab 1: distributed big-data framework (like MapReduce)

Lab 2: fault tolerance library using replication (Raft)

Lab 3: a simple fault-tolerant database

Lab 4: scalable database performance via sharding

MAIN TOPICS

This is a course about infrastructure for applications.

  • Storage.
  • Communication.
  • Computation.

A big goal: hide the complexity of distribution from applications.

Topic: fault tolerance

1000s of servers, big network -> always something broken

We’d like to hide these failures from the application.

“High availability”: service continues despite failures

Big idea: replicated servers.

If one server crashes, can proceed using the other(s).

Labs 2 and 3

Topic: consistency

General-purpose infrastructure needs well-defined behavior.

E.g. “Get(k) yields the value from the most recent Put(k,v).”

Achieving good behavior is hard!

“Replica” servers are hard to keep identical.

Topic: performance

The goal: scalable throughput

Nx servers -> Nx total throughput via parallel CPU, disk, net.

Scaling gets harder as N grows:

  • Load imbalance.
  • Slowest-of-N latency.
  • Some things don’t speed up with N: initialization, interaction.

Labs 1, 4

Topic: tradeoffs

Fault-tolerance, consistency, and performance are enemies.

Fault tolerance and consistency require communication

e.g., send data to backup

e.g., check if my data is up-to-date

communication is often slow and non-scalable

Many designs provide only weak consistency, to gain speed.

e.g. Get() does not yield the latest Put()!

Painful for application programmers but may be a good trade-off.

We’ll see many design points in the consistency/performance spectrum.

Topic: implementation

RPC, threads, concurrency control, configuration.

The labs…

This material comes up a lot in the real world.

All big web sites and cloud providers are expert at distributed systems.

Many big open source projects are built around these ideas.

We’ll read multiple papers from industry.

And industry has adopted many ideas from academia.

CASE STUDY: MapReduce

Let’s talk about MapReduce (MR)

a good illustration of 6.824’s main topics

hugely influential

the focus of Lab 1

MapReduce overview

context: multi-hour computations on multi-terabyte data-sets

e.g. build search index, or sort, or analyze structure of web

only practical with 1000s of computers

applications not written by distributed systems experts

overall goal: easy for non-specialist programmers

programmer just defines Map and Reduce functions

often fairly simple sequential code

MR manages, and hides, all aspects of distribution!

Abstract view of a MapReduce job — word count

Input1 -> Map -> a,1 b,1

Input2 -> Map -> b,1

Input3 -> Map -> a,1 c,1
| | |-> Reduce -> c,1
| | ——-> Reduce -> b,2
| ————-> Reduce -> a,2

  1. input is (already) split into M files
  2. MR calls Map() for each input file, produces set of k2,v2
    “intermediate” data
    each Map() call is a “task”
  3. when Maps are don,
    MR gathers all intermediate v2’s for a given k2,
    and passes each key + values to a Reduce call
  4. final output is set of<k2,v3> pairs from Reduce()s

Word-count-specific code

  1. Map(k, v)
  2.     split v into words
  3.     for each word w
  4.       emit(w, "1")
  5. Reduce(k, v_set)
  6.     emit(len(v_set))

MapReduce scales well:

N “worker” computers (might) get you Nx throughput.

Maps()s can run in parallel, since they don’t interact.

Same for Reduce()s.

Thus more computers -> more throughput — very nice!

MapReduce hides many details:

sending app code to servers

tracking which tasks have finished

“shuffling” intermediate data from Maps to Reduces

balancing load over servers

recovering from failures

However, MapReduce limits what apps can do:

  • No interaction or state (other than via intermediate output).
  • No iteration
  • No real-time or streaming processing.

Input and output are stored on the GFS cluster file system

MR needs huge parallel input and output throughput.

GFS splits files over many servers, in 64 MB chunks

  1. Maps read in parallel
  3. Reduces write in parallel

GFS also replicates each file on 2 or 3 servers

GFS is a big win for MapReduce

Some details (paper’s Figure 1):

one coordinator, that hands out tasks to workers and remembers progress.

  1. coordinator gives Map tasks to workers until all Maps complete
    Maps write output (intermediate data) to local disk
    Maps split output, by hash, into one file per Reduce task
  2. after all Maps have finished, coordinator hands out Reduce tasks
    each Reduce fetches its intermediate output from (all) Map workers
    each Reduce task writes a separate output file on GFS

What will likely limit the performance?

We care since that’s the thing to optimize.

CPU? memory? disk? network?

In 2004 authors were limited by network capacity.

What does MR send over the network?

    Maps read input from GFS.

    Reduces read Map intermediate output.

        Often as large as input, e.g. for sorting.

    Reduces write output files to GFS.

[diagram: servers, tree of network switches]

In MR's all-to-all shuffle, half of traffic goes through root switch.

Paper's root switch: 100 to 200 gigabits/second, total

    1800 machines, so 55 megabits/second/machine.

    55 is small:  much less than disk or RAM speed.

Today: networks are much faster

How does MR minimize network use?

Coordinator tries to run each Map task on GFS server that stores its input.

  • All computers run both GFS and MR workers
  • So input is read from local disk (via GFS), not over network.

Intermediate data goes over network just once.

  • Map worker writes to local disk.
  • Reduce workers read from Map worker disks over the network.
  • Storing it in GFS would require at least two trips over the network.

Intermediate data partitioned into files holding many keys.

  • R is much smaller than the number of keys.
  • Big network transfers are more efficient.

How does MR get good load balance?

Wasteful and slow if N-1 servers have to wait for 1 slow server to finish.

But some tasks likely take longer than others.

Solution: many more tasks than workers.

  • Coordinator hands out new tasks to workers who finish previous tasks.
  • So no task is so big it dominates completion time (hopefully).
  • So faster servers do more tasks than slower ones, finish abt the same time.

What about fault tolerance?

I.e. what if a worker crashes during a MR job?

We want to hide failures from the application programmer!

Does MR have to re-run the whole job from the beginning?

  • Why not?

MR re-runs just the failed Map()s and Reduce()s.

Suppose MR runs a Map twice, one Reduce sees first run's output,another Reduce sees the second run's output?

Correctness requires re-execution to yield exactly the same output.

So Map and Reduce must be pure deterministic functions:
  • they are only allowed to look at their arguments/input.
  • no state, no file I/O, no interaction, no external communication.

What if you wanted to allow non-functional Map or Reduce?

Worker failure would require whole job to be re-executed, or you'd need to roll back to some kind of global checkpoint.

Details of worker crash recovery:

  • a Map worker crashes:
    coordinator notices worker no longer responds to pings
    coordinator knows which Map tasks ran on that worker
    those tasks’ intermediate output is now lost, must be re-created
    coordinator tells other workers to run those tasks
    can omit re-running if all Reduces have fetched the intermediate data
  • a Reduce worker crashes:
    finished tasks are OK — stored in GFS, with replicas.
    coordinator re-starts worker’s unfinished tasks on other workers.

Other failures/problems:

  • What if the coordinator gives two workers the same Map() task?
    perhaps the coordinator incorrectly thinks one worker died.
    it will tell Reduce workers about only one of them.
  • What if the coordinator gives two workers the same Reduce() task?
    they will both try to write the same output file on GFS!
    atomic GFS rename prevents mixing; one complete file will be visible.
  • What if a single worker is very slow — a “straggler”?
    perhaps due to flakey hardware.
    coordinator starts a second copy of last few tasks.
  • What if a worker computes incorrect output, due to broken h/w or s/w?
    too bad! MR assumes “fail-stop” CPUs and software.
  • What if the coordinator crashes?

Current status?

Hugely influential (Hadoop, Spark, &c).
Probably no longer in use at Google.
Replaced by Flume / FlumeJava (see paper by Chambers et al).
GFS replaced by Colossus (no good description), and BigTable.

Conclusion

MapReduce made big cluster computation popular.

  • Not the most efficient or flexible.

  • Scales well.

  • Easy to program — failures and data movement are hidden.
    These were good trade-offs in practice.
    We’ll see some more advanced successors later in the course.
    Have fun with Lab 1!