AFS and NFS 20 Years Later Mike Kazar

AFS and NFS … 20 Years Later Mike Kazar kazar@netapp. com June 2005 Network Appliance 1

Overview 4 Inspired by talk by Brian Pawlowski (beepy) – – presented to a bunch of folks at IBM including lots of AFS people 4 What were AFS’s goals? NFS’s goals? 4 Innovation on client and server ends 4 Storage management – Then and now 4 Some thoughts about the future 4 Lessons learned 4 Questions Network Appliance | 9/10/2020 2

Who Am I? 4 Part – – – of AFS design team, with Bob Sidebotham (Rx, Volumes) Mike West (Server) Sherri Nichols (VLDB &c) M. Satyanarayanan (prototype RPC) Dave Nichols (prototype client) 4 Coded parts of AFS – cache manager – kernel port of Rx 4 Now Network Appliance | at Net. App 9/10/2020 3

AFS Goals in 1984 4 Distributed – connect lots of workstations together 4 Scalability – did I mention *lots*? – key approach was caching • the fastest RPC is the one you don’t make 4 Security – needed isolation from random students 4 Management was an afterthought – but turned out to be critical – volumes – volume *moves* – mirroring Network Appliance | 9/10/2020 4

Key points from Beepy’s talk 4 What is NFS? – IETF standard – bundled with all Unix/Linux systems – available on nearly everything – heterogeneous (systems, FSes) 4 If NFS is the answer, what was the question? Network Appliance | 9/10/2020 5

Stick a fork in it… FTAM Network Appliance | 9/10/2020 6

Clients, Servers, an’ ‘at 4 So, why am I smiling? 4 Clients, Protocols, and Servers – Different goals for each Network Appliance | 9/10/2020 7

Clients 4 OS integration 4 Bug-free – or you get calls all the time 4 Reference implementation – helps portability a great deal – even though porting still lots of work – “Vnode” layers in Windows, AIX, Solaris, &c Network Appliance | 9/10/2020 8

Protocols 4 Where NFS Really Shines – – – 4 Public protocol spec • everyone knows they can implement it Interoperability tests (Connectathon) • formally, helps certify who’s in the game • informally, helps communication! – set goals for future work Reference implementation • as education tool Where NFS sucks – – Network Appliance | cache coherence • how did they blow this in NFS version 4? but still relatively minor • few applications use DSM • locking makes most things work 9/10/2020 9

Servers 4 Where AFS Shines – Data management • global name space • cell name spaces • transparent move • transparent load balancing mirrors • flexibly allocated volumes • snapshots / clones – Usable ACLs • suprisingly – caching • using memory caches today Network Appliance | 9/10/2020 10

Today 4 NFS – – is the protocol of choice open licensing made it a no-brainer simplicity also a plus for growth 4 Some AFS data management – Snapshots on Net. App filers – flexible volumes – non-transparent moves available today 4 Some AFS data management – transparent move – load balancing mirrors – multiple cells available “soon” Network Appliance | 9/10/2020 11

On. TAP NG Architecture Network Appliance | 9/10/2020 12

Architecture Detail ONTAP / NG: 2 -Stage Distributed File System 1. 2. 3. Request switched to appropriate back-end IP-based cluster network Client Access Gigabit Ethernet Network Function • TCP termination • VLDB lookup • Protocol translation to Spin. FS Protocol X No client code changes Disk Function • Caching • Locking Fibre Channel Gigabit Ethernet Switch Disk Function • Caching • Locking Fibre Channel 12 Network Appliance | 9/10/2020 13

History as Knobs 4 No knobs – – 4 Too – – – 4 Just – – Network Appliance | initial filers • one volume, period Suns, &c • network parameters, exports, not much else many knobs tracking thousands of volumes figuring out restores where do I create a new volume? right self-managing based on guidance eg. this part of the name space is a database 9/10/2020 14

The Future 4 Policy-based management – declare part of name space “database” • inherit RAID level, drive speed • referenced abstractly – constraint engine • moves data around • with limited system impact • when desired – tied to delegation • sub-admins tied to name space parts • sub-admins constrained by resource limits Network Appliance | 9/10/2020 15

The Future 4 Ties – – Network Appliance | to historical data what volumes are heavily loaded, and when? which volumes grow and shrink? • what variation in size • what variation in load what volumes were on this server? • backup database issue charge-back 9/10/2020 16

The Future 4 Quality of Service – Important for managing applications – but no common framework • eg: to connnect job controller and storage – so, need to start somewhere • virtual servers, perhaps • per volume? • priorities vs. guaranteed bandwidth or ops Network Appliance | 9/10/2020 17

Lessons Learned 4 Perfection is highly overrated – POSIX semantics never really required – Huge effort in DCE/DFS 4 Universities are similar to enterprises – pointless politics and empire building – 24 x 7 operation – availability and reliability – coordination with users nearly impossible – slightly cheaper Network Appliance | 9/10/2020 18

Lessons Learned 4 Technology Transfer is Hard – 20 years to get AFS volume concept out • only happened because Blake Lewis and Ed. Zayas went to Net. App – More for transparent moves • Spinnaker acquisition 4 Beware second system syndrome – and ignoring customer requirements – DCE/DFS was nearly complete waste of time 4 Morris’s point about 3 rd systems – independence from inventor is critical • throw out bad ideas – allows simplification Network Appliance | 9/10/2020 19

Lessons Learned 4 No one pays attention to system management – but it is critical to any technology deployment Qo. S is part of this scaling is part of this (managing resource pools) centralized error reporting dynamic reconfiguration – – 4 Don’t be greedy! – know where your real value is – know how to get help – IBM lived in fear that Sun would productize AFS! • so bungled licensing Network Appliance | 9/10/2020 20