What's new in Seastar - issue 4
A summary of recent developments in the C++ Seastar project .
Noah Watkins
Recent content on Noah Watkins
Latest posts
What's new in Seastar - issue 3
A summary of recent developments in the C++ Seastar project .
What's new in Seastar - issue 2
A summary of recent developments in the C++ Seastar project .
A virtual machine approach to fuzzing interfaces
Clang’s libfuzzer can design adverserial programs.
What's new in Seastar - issue 1
A summary of recent developments in the C++ Seastar project .
Seastar's semaphore utility
Everyone’s favorite concurency control device.
Seastar's timer utility
Efficiently manage a bunch of scheduled, asynchronous upcalls.
Seastar's gate utility
Synchronize access to resources by tracking active fibers.
Seastar's weak pointer
The non-owning counterpart to Seastar’s shared pointer.
Seastar's temporary buffer
A staple utility for memory management in Seastar.
About
Welcome to makedist.
Seastar's deleter utility
A souped up tool for building RAII objects.
Seastar's shared pointer
A reference counted smart pointer with unique optimizations.
Video of my Cephalocon 2019 Talk
Creating active objects with Lua in the Ceph distributed storage system.
Video of my Cephalocon 2018 Talk
How we built a distributed shared-log on top of Ceph.
Getting started with the smf rpc framework
Learn how to use the fastest RPC in the west
Video of my talk at Lua Workshop 2017
Creating active objects with Lua in the Ceph distributed storage system.
Ceph OSD state machine visualization
Using clang to extract a state machine diagram from Ceph’s C++ implementation.
Management of the CruzDB database catalog
Taking the CruzDB data model beyond a simple, flat key space.
How CruzDB manages transaction afterimages
Increasing transaction throughput by writing database snapshots in parallel.
An overview of transaction management in CruzDB
A shared-log creates interesting challenges for transaction management.
CruzDB is one large copy-on-write tree
Each transaction in CruzDB is stored in a log as a copy-on-write tree delta.
An introduction to the CruzDB database system
CruzDB is an immutable database built on top of a distributed shared-log.
Systems research I'm watching for in 2018
Distributed shared memory systems, language runtimes, and server frameworks.
Sharding the LRU node cache in CruzDB
Improving read performance by sharding the LRU node cache.
Debuting the performance of CruzDB
Hot of the press performance results show plenty of room to improve.
Slow placement group read operations in Ceph
An unsolved mystery involving slow reads and Ceph placement groups.
Adding a new placement group operation in Ceph
Teaching Ceph new tricks on entire groups of objects.
Measuring userfaultfd page-fault latency
How fast can my application service its own page fault?
Building an immutable key-value store on zlog
Persistent data structures are the dopest shit.
ZLog project update (mid-2016 edition)
Project funding, design updates, and improved performance.
RADOS object class development activity
Spoiler alert: object classes are HOT .
Toward dynamic RADOS object class management
Persisting Lua-based RADOS object classes in Ceph.
Load Lua RADOS classes from local file system
Supporting the installation of Lua-based RADOS object classes.
ZLog asynchronous I/O support
Bringing zlog into the future with a new asynchronous interface.
Hadoop on Ceph: usability survey
Hadoop interoperability with Lustre, Quantcast, Gluster, and more.
Hadoop on Ceph: diving in
Using Hadoop to process data stored in the Ceph distributed file system.
Distributed search for Beal's Conjecture counterexamples
Speeding up a brute force search in an enormous state space.
Setting up iSER-enabled TGT RAM disk
Bigger and faster networked RAM disks with RDMA and RAID.
Remote RAM disk with RDMA
Piecing together iSER, iSCSI, and LIO to build a networked RAM disk.
Setting up RoCE on aarch64 Ubuntu
Resolving kernel module dependency issues with RoCE.
Performance of the zlog sequencer service
How fast can an integer be incremented and read over the network?
ZLog: a distributed shared-log on Ceph
Building the CORFU distributed shared-log protocol on top of Ceph.
Ceph OSD request processing latency
How fast can Ceph’s RADOS object store process a request?
Tracing Ceph with LTTng
Collecting execution traces in Ceph with LTTng userspace tooling.
The latency of Ceph placement group splitting
How fast can Ceph split up and redistribute a collection of objects?
OpRequest flow in RADOS OSD server
Tracing the flow of object request handling in Ceph.
Performance of compound object operations in Ceph (part 2)
Reducing contention by striping operations across placement groups.
Performance of compound object operations in Ceph (part 1)
We can combine operations on Ceph objects, but should we?
Fixing a Ceph performance WTF
Tracking down and fixing a poorly performing Ceph application workload.