Aditya Akella
UW-Madison
Shuchi Chawla
Ashok Anand
Chitra Muthukrishnan
UW-Madison
Srinivasan
Seshan Vyas
Sekar
CMU
Scott
Shenker
UC-Berkeley
Ram Ramjee
MSR-India

2
Video
Data centers
Web content
Other svcs
(backup)
Strain on installed link capacities
ISP core


Network traffic volumes growing rapidly
Annual growth: overall (45%), enterprise (50%), mobile (125%)*


Growing strain on installed capacity everywhere
Core (Asian ISPs – 80-90% core utilization), enterprise access, data center, cellular, wireless…

How to sustain robust network performance?
Mobile users
Enterprises Home users
* Interview with Cisco CEO, Aug 2007, Network world

3
Video
Data centers
Web content
Other svcs
(backup)
CDN
Wan
Opt
Wan
Opt
Dedup/ archival
Dedup/ archival
ISP HTTP cache



A key idea: suppress duplicates


Popular objects, partial content matches, backups, app headers
Effective capacity improves ~ 2X
Many approaches

Application-layer caches
 Protocol-independent schemes


Below app-layer
WAN accelerators, de-duplication

Content distribution
 CDNs like Akamai, CORAL
 Bittorrent
Point solutions  apply to specific link, protocol, or app
Mobile users
Enterprises Home users

4
Point solutions inadequate
Point solutions:
Other links must
Wan
Opt re-implement specific
RE mechanisms
Elimination Service in the core
Wan
Opt
Architectural support to address universal need to scale capacities? Implications?
Dedup/ archival
RE: A primitive operation supported inherently in the network o Applies to all links, flows (long/short), apps, unicast/multicast
Bittorrent o Transparent network service; optional endpoint modifications
Point solutions:
Only benefit system/app attached o How? Implications?
Dedup/ archival
ISP HTTP cache

5
WAN link

Data center Cache Cache
5
Enterprise
Network must examine byte streams, remove duplicates, reinsert

Building blocks from WAN optimizers: RE agnostic to application, ports or flow semantics

Upstream cache = content table + fingerprint index

Fingerprint index: content-based names for chunks of bytes in payload


Fingerprints computed for content, looked up to identify redundant bytestrings
Downstream cache: content table

6
Wisconsin Packet cache at every router
Router upstream removes redundant bytes
Router downstream reconstructs full packet
Network RE service: apply protocol-indep
RE at the packet-level on network links
 IP-layer RE service
(Hop-by-hop works for slow links
Alternate approaches to scale to faster links…)
Internet2
CMU Berkeley

7

Improved performance everywhere even if partially enabled


Generalizes point deployments and app-specific approaches
 Benefits all network end-points, applications
Crucially, benefits ISPs
 Improved switching capacity, responsiveness to sudden overload

Architectural benefits


Enables new protocols and apps


Min-entropy routing, RE-aware traffic engineering (intra- and inter-domain)
Anomaly detection, in-network filtering of unwanted traffic
Simplifies/improves apps : need not worry about using network efficiently
 Application control messages & headers can be verbose  better diagnostics
 Controlling duplicate transmission in app-layer multicast is a non-issue

8
Network RE
 12 pkts
(ignoring tiny packets)
Wisconsin
6  2 packets
Generalizes point deployments
Benefits ISPs: improve effective switching capacity
Without RE
 18 pkts
33% lower
CMU 3  2 packets
Internet2
3  2 packets
Berkeley

9
Wisconsin
Simple RE
 12 pkts
RE + routing
 10 pkts
Minimum-entropy routing
New, flexible traffic engineering mechanisms
Inter-domain protocols
Redundancy-based anomaly detectors
Network-assisted spam filtering
New content distribution mechanisms
CMU
Internet2 Berkeley
9

10

Analysis of 12 enterprises: traffic 15-60% redundant [SIGMETRICS 09]

~1GB of cache sufficient to identify redundancies
 DRAM or PCM (PRAM) on routers

Network RE benefits both ISPs and end-networks [SIGCOMM 08]


Upto 15-50% better util, responsive TE, control inter-domain traffic impact
Centralized algorithm for min-entropy routing (using “redundancy profiles”)
 Reduces utilization by a further 10-25% in intra-domain case
 Inter-domain min-entropy routing: gains much more significant (50-80%)

Is network RE viable at high speeds ? Not in its current form…

Compression is slow: limits hop-by-hop speed at each hop to 2.5Gbps


Acceptable for access, wireless, cellular links, not for the core
Also, wastes memory on multiple routers  limits effectiveness

11





Toss out link-by-link view; treat RE as a network-wide problem per ISP
[Current work]
Memory usage: each packet compressed/un-compressed once
Throughput: allow reconstruction multiple hops away from compression


Stand-alone reconstruction much faster when freed from dependence on compression immediately upstream
Reconstructor can reconstruct a lot more, from multiple different compression agents
Resource-awareness: carefully account for network and device resources, and traffic


Compression/reconstruction/caching locations decided based on memory capacity and memory operations
Also consider global TE objectives
Just 4% from ideal RE (no memory or processing constraints)

12




RE service to scale link capacities everywhere
Architectural niceties and performance benefits
High speed router RE seems feasible
Future directions
 End-host participation
 Role of different memory technologies – DRAM, flash and PCM
 Theoretical issues – pricing and economics, routing policy, network design
 Network coding as an alternative to compression