Systems Seminar - CSE

Flat Datacenter Storage

Ed NightingaleArchitectOnline Services Division (OSD) at Microsoft

Flat Datacenter Storage (FDS) is a high-performance, fault-tolerant,
large-scale, locality-oblivious blob store. Using a novel combination
of full bisection bandwidth networks, data and metadata striping, and
flow control, FDS multiplexes an application's large-scale I/O across
the available throughput and latency budget of every disk in a
cluster. FDS therefore makes many optimizations around data locality
unnecessary. Disks also communicate with each other at their full
bandwidth, making recovery from disk failures extremely fast. FDS is
designed for datacenter scale, fully distributing metadata operations
that might otherwise become a bottleneck.

FDS applications achieve single-process read and write performance of
more than 2 GB/s. We measure recovery of 92 GB data lost to disk
failure in 6.2 s and recovery from a total machine failure with 655 GB
of data in 33.7 s. Application performance is also high: we describe
our FDS-based sort application which set the 2012 world record for
disk-to-disk sorting.
Ed Nightingale is an architect working in the Online Services Division
(OSD) at Microsoft. Ed runs the Cosmos storage system, which is a
batch processing store for Bing that runs on tens of thousands of
machines and processes petabytes of data every day. Before joining
Bing, Ed was a senior researcher at Microsoft Research. Ed has always
enjoyed building operating systems and distributed systems. More
details on Ed's work can be found at

Sponsored by