Loading…
LISA18 has ended
Back To Schedule
Monday, October 29 • 7:00pm - 8:00pm
AuriStor Vendor BoF: AuriStor File System and Linux Kernel AFS Milestones

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
Speakers: Jeffrey Altman and Daria Brashear

The AuriStor File System, a next-generation successor to IBM's AFS, is deployed around the world in more than 100 cells since May 2016. Its global file namespace (/afs), platform-agnostic file paths, online backups, and an atomic publication model have proven to be critical components of successful software and content distribution systems. AuriStorFS cells scale to hundreds of thousands of systems. The /afs global name space has survived multiple OS and processor architecture changes over more than three decades making it the perfect match for global distribution of open data sets.

The AuriStorFS security model incorporates combined-identity authentication, multi-factor authorization, and mandatory security policies. AuriStorFS permits a global name space to extend across internal, DMZ and cloud infrastructures; and to store a mix of open and restricted data.

This session will discuss the most recent updates to AuriStorFS and the native Linux kernel implementation. The session will include a demonstration consisting of:
  • Containers with binary executable files stored in /afs
  • Containers mounting private AFS volume for scratch space
  • AuriStorFS and (k)AFS file system implementations running side-by-side Linux namespaces for /afs

AuriStorFS milestones since LISA 2017 include:
  • Successful migration and replication of volumes exceeding 5.5TB. The largest production volume so far is 250TB storing hundreds of 1TB or larger files.
  • Deployment of a single AuriStorFS cell spanning an internal data center, AWS and GCP with more than 25,000 nodes for distribution of software and configuration data.
  • Meltdown and Spectre remediation. In response to nearly 30% performance hit from Meltdown and Spectre the AuriStor team optimized the Rx stack, Ubik database and fileserver to reduce the number of syscalls by more than 50% AES-NI, SSSE3, AVX and AVX2 Intel processor optimization of AES256-CTS-HMAC-SHA1-96 cryptographic operations for kernel cache managers reduces computation time by 64%.
  • AuriStor's RX implementation has undergone a major upgrade of its flow control model. Prior implementations were based on TCP Reno Congestion Control as documented in RFC5681; and SACK behavior that was loosely modelled on RFC2018. The new RX state machine implements SACK based loss recovery as documented in RFC6675, with elements of New Reno from RFC6582 on top of TCP-style congestion control elements as documented in RFC5681. The new RX also implements RFC2861 style congestion window validation.
RX peers implementing these changes are more likely to sustain the maximum available transmission throughput while improving fairness towards competing network data flows. The improved estimation of available pipe capacity permits an increase in the default maximum window size from 60 packets (84.6 KB) to 128 packets (180.5 KB). The larger window size increases the per call theoretical maximum throughput on a 1ms RTT link from 693 mbit/sec to 1478 mbit/sec and on a 30ms RTT link from 23.1 mbit/sec to 49.39 mbit/sec.
  • macOS Mojave support
  • Linux native kernel AFS highlights:
    • Full support of AuriStorFS protocol extensions
    • IPv6 support for AuriStorFS cells
    • dynamic root, @sys and @cell support
    • multipage read and write support
    • Acceptance into Fedora Linux Rawhide 4.19 kernel


Monday October 29, 2018 7:00pm - 8:00pm CDT
Music Row Room 4

Attendees (5)