![Page 1: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/1.jpg)
PrincetonUniversity
Performance Isolation and Fairness for Multi-Tenant Cloud Storage
David Shue*, Michael Freedman*, and Anees Shaikh✦
*Princeton ✦IBM Research
![Page 2: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/2.jpg)
Setting: Shared Storage in the Cloud
2
Z
Y
TFZ
Y
FT
![Page 3: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/3.jpg)
Setting: Shared Storage in the Cloud
2
Z
Y
TFZ
Y
FT
S3 EBS SQS
![Page 4: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/4.jpg)
Setting: Shared Storage in the Cloud
2
Z
Y
TFZ
Y
FT
S3 EBS SQSShared Key-Value Storage
![Page 5: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/5.jpg)
DD DD DD DDShared Key-Value Storage
3
Z Y T FZ Y FTY YZ F F F
Multiple co-located tenants ⇒ resource contention
Predictable Performance is Hard
![Page 6: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/6.jpg)
DD DD DD DDDD DD DD DDShared Key-Value Storage
4
Z Y T FZ Y FTY YZ F F F
Multiple co-located tenants ⇒ resource contention
Predictable Performance is Hard
![Page 7: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/7.jpg)
DD DD DD DDDD DD DD DDShared Key-Value Storage
4
Z Y T FZ Y FTY YZ F F F
Fair queuing @ big iron
Multiple co-located tenants ⇒ resource contention
Predictable Performance is Hard
![Page 8: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/8.jpg)
5
Distributed system ⇒ distributed resource allocationMultiple co-located tenants ⇒ resource contention
Z Y T FZ Y FTY YZ F F F
SS SS SS SS
Predictable Performance is Hard
![Page 9: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/9.jpg)
Z Y T FZ Y FTY YZ F F F
SS SS SS SS
6
Multiple co-located tenants ⇒ resource contentionDistributed system ⇒ distributed resource allocation
Z keyspace T keyspace F keyspaceY keyspace
Predictable Performance is Hard
![Page 10: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/10.jpg)
Z Y T FZ Y FTY YZ F F F
SS SS SS SS
6
popu
lari
ty data partition
Multiple co-located tenants ⇒ resource contentionDistributed system ⇒ distributed resource allocation
Predictable Performance is Hard
![Page 11: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/11.jpg)
Z Y T FZ Y FTY YZ F F FZ Y T FZ Y FTY YZ F F F
SS SS SS SS
7
Skewed object popularity ⇒ variable per-node demand
Multiple co-located tenants ⇒ resource contentionDistributed system ⇒ distributed resource allocation
Predictable Performance is Hard
![Page 12: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/12.jpg)
Z Y T FZ Y FTY YZ F F FZ Y T FZ Y FTY YZ F F F
SS SS SS SS
7
Skewed object popularity ⇒ variable per-node demand
Multiple co-located tenants ⇒ resource contentionDistributed system ⇒ distributed resource allocation
1kBGET 10BGET 1kBSET 10BSET(small reads)(large reads) (large writes) (small writes)
Disparate workloads ⇒ different bottleneck resources
Predictable Performance is Hard
![Page 13: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/13.jpg)
Zynga Yelp FoursquareTP
Shared Key-Value Storage
8
Tenants Want System-wide Resource Guarantees
Z Y T FZ Y FTY YZ F F F
SS SS SS SS
demandz = 120 kreq/s
demandf = 120 kreq/s
Skewed object popularity ⇒ variable per-node demand
Multiple co-located tenants ⇒ resource contentionDistributed system ⇒ distributed resource allocation
Disparate workloads ⇒ different bottleneck resources
![Page 14: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/14.jpg)
Zynga Yelp FoursquareTP
Shared Key-Value Storage
8
Tenants Want System-wide Resource Guarantees
Z Y T FZ Y FTY YZ F F F
SS SS SS SS
80 kreq/s 120 kreq/s 160 kreq/s40 kreq/sdemandz = 120 kreq/s
demandf = 120 kreq/s
Skewed object popularity ⇒ variable per-node demand
Multiple co-located tenants ⇒ resource contentionDistributed system ⇒ distributed resource allocation
Disparate workloads ⇒ different bottleneck resources
![Page 15: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/15.jpg)
Zynga Yelp FoursquareTP
Shared Key-Value Storage
9
Pisces Provides Weighted Fair-shares
wz = 20% wy = 30% wf = 40%wt = 10%
demandz = 30%Z Y T FZ Y FTY YZ F F F
SS SS SS SS
Skewed object popularity ⇒ variable per-node demand
Multiple co-located tenants ⇒ resource contentionDistributed system ⇒ distributed resource allocation
Disparate workloads ⇒ different bottleneck resources
![Page 16: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/16.jpg)
Pisces: Predictable Shared Cloud Storage
10
•Pisces- Per-tenant max-min fair shares of system-wide resources
~ min guarantees, high utilization
- Arbitrary object popularity
- Different resource bottlenecks
![Page 17: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/17.jpg)
Pisces: Predictable Shared Cloud Storage
10
•Pisces- Per-tenant max-min fair shares of system-wide resources
~ min guarantees, high utilization
- Arbitrary object popularity
- Different resource bottlenecks
•Amazon DynamoDB- Per-tenant provisioned rates
~ rate limited, non-work conserving
- Uniform object popularity
- Single resource (1kB requests)
![Page 18: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/18.jpg)
Tenant A
Predictable Multi-Tenant Key-Value Storage
11
Tenant BVM VM VM VM VM VM
GET 1101100
RR
Controller
![Page 19: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/19.jpg)
Tenant A
Predictable Multi-Tenant Key-Value Storage
11
Tenant BVM VM VM VM VM VM
GET 1101100
RR
Controller
PP
![Page 20: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/20.jpg)
Tenant A
Predictable Multi-Tenant Key-Value Storage
11
Tenant BVM VM VM VM VM VM
RS
GET 1101100
RR
Controller
PP
![Page 21: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/21.jpg)
Tenant A
Predictable Multi-Tenant Key-Value Storage
11
Tenant BVM VM VM VM VM VM
RS
FQ
GET 1101100
RR
Controller
PP
![Page 22: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/22.jpg)
Tenant A
Predictable Multi-Tenant Key-Value Storage
12
Tenant BVM VM VM VM VM VM
WeightA WeightB
RS
FQ
PP
WA WA2 WB2
GET 1101100
RR
Controller
![Page 23: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/23.jpg)
Tenant A
Predictable Multi-Tenant Key-Value Storage
12
Tenant BVM VM VM VM VM VM
WeightA WeightB
RS
FQ
PP
WA WA2 WB2
GET 1101100
RR
Controller
WA2 WB2WA1 WB1
![Page 24: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/24.jpg)
Strawman: Place Partitions Randomly
13
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
RS
FQ
WA WA2 WB2Controller
RR
![Page 25: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/25.jpg)
Strawman: Place Partitions Randomly
14
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
RS
FQ
WA WA2 WB2
RR
ControllerOverloaded
![Page 26: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/26.jpg)
Pisces: Place Partitions By Fairness Constraints
15
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
RS
FQ
WA WA2 WB2
RR
Collect per-partition tenant demand Controller
![Page 27: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/27.jpg)
Pisces: Place Partitions By Fairness Constraints
15
Bin-pack partitions
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
RS
FQ
WA WA2 WB2
RR
Collect per-partition tenant demand Controller
![Page 28: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/28.jpg)
Pisces: Place Partitions By Fairness Constraints
16
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
Results in feasible partition placement
RS
FQ
WA WA2 WB2
RR
Controller
![Page 29: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/29.jpg)
Pisces: Place Partitions By Fairness Constraints
16
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
Results in feasible partition placement
RS
FQ
WA WA2 WB2
RR
Controller
![Page 30: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/30.jpg)
Controller
Strawman: Allocate Local Weights Evenly
17
WA1 = WB1 WA2 = WB2
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
WA WA2 WB2
RR RS
FQ
![Page 31: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/31.jpg)
Controller
Strawman: Allocate Local Weights Evenly
17
WA1 = WB1 WA2 = WB2
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
WA WA2 WB2
RR RS
FQ
![Page 32: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/32.jpg)
Controller
Strawman: Allocate Local Weights Evenly
17
WA1 = WB1 WA2 = WB2
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
WA WA2 WB2
RR RS
FQ
Overloaded
![Page 33: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/33.jpg)
Pisces: Allocate Local Weights By Tenant Demand
18
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
PP
WA WA2 WB2Controller
WA1 = WB1 WA2 = WB2
RR RS
FQ
![Page 34: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/34.jpg)
Pisces: Allocate Local Weights By Tenant Demand
18
maxmismatch
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
Compute per-tenant+/- mismatch
PP
WA WA2 WB2Controller
WA1 = WB1 WA2 = WB2
RR RS
FQ
![Page 35: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/35.jpg)
Pisces: Allocate Local Weights By Tenant Demand
18 A←B
WA1 > WB1
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
Compute per-tenant+/- mismatch
PP
WA WA2 WB2Controller
WA2 = WB2
RR RS
FQ
![Page 36: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/36.jpg)
Pisces: Allocate Local Weights By Tenant Demand
18 A←B A→B
WA1 > WB1 WA2 < WB2
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
Compute per-tenant+/- mismatch
PP
WA WA2 WB2Controller
Reciprocal weight swap
RR RS
FQ
![Page 37: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/37.jpg)
Strawman: Select Replicas Evenly
19
50% 50%RS
PP
WA WA2 WB2Controller
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
GET 1101100
RR
WA1 > WB1 WA2 < WB2
FQ
![Page 38: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/38.jpg)
Strawman: Select Replicas Evenly
19
50% 50%
RS
PP
WA WA2 WB2Controller
Tenant A Tenant BVM VM VM VM VM VM
WeightA WeightB
GET 1101100
RR
WA1 > WB1 WA2 < WB2
FQ
![Page 39: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/39.jpg)
Tenant A
Pisces: Select Replicas By Local Weight
20
Tenant BVM VM VM
WeightB
Controller50% 50%
RS
PP
WA WA2 WB2
VM VM VM
WeightA
GET 1101100
WA1 > WB1 WA2 < WB2
FQ
RR
![Page 40: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/40.jpg)
Tenant A
Pisces: Select Replicas By Local Weight
20
detect weightmismatch by
request latency
Tenant BVM VM VM
WeightB
Controller50% 50%
RS
PP
WA WA2 WB2
VM VM VM
WeightA
GET 1101100
WA1 > WB1 WA2 < WB2
FQ
RR
![Page 41: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/41.jpg)
Tenant A
Pisces: Select Replicas By Local Weight
20
60% 40%detect weightmismatch by
request latency
Tenant BVM VM VM
WeightB
Controller
RS
PP
WA WA2 WB2
VM VM VM
WeightA
GET 1101100
WA1 > WB1 WA2 < WB2
FQ
RR
![Page 42: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/42.jpg)
Tenant A
Pisces: Select Replicas By Local Weight
20
60% 40%detect weightmismatch by
request latency
Tenant BVM VM VM
WeightB
Controller
RS
PP
WA WA2 WB2
VM VM VM
WeightA
GET 1101100
WA1 > WB1 WA2 < WB2
FQ
RR
![Page 43: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/43.jpg)
Strawman: Queue Tenants By Single Resource
21
Tenant A Tenant BVM VM VM VM VM VM
Controller
RS
PP
WA WA2 WB2
FQ
WA2 < WB2WA1 > WB1
RR
GET 1101100GET 0100111
![Page 44: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/44.jpg)
Strawman: Queue Tenants By Single Resource
21
Tenant A Tenant BVM VM VM VM VM VM
Controller
RS
PP
WA WA2 WB2
FQ
WA2 < WB2WA1 > WB1
RR
![Page 45: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/45.jpg)
Strawman: Queue Tenants By Single Resource
Bandwidth limited Request Limited
21
out req out req
Tenant A Tenant BVM VM VM VM VM VM
Controller
RS
PP
WA WA2 WB2
FQ
WA2 < WB2
RR
![Page 46: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/46.jpg)
Strawman: Queue Tenants By Single Resource
Bandwidth limited Request Limited
bottleneck resource (out bytes) fair share
21
out req out req
Tenant A Tenant BVM VM VM VM VM VM
Controller
RS
PP
WA WA2 WB2
FQ
WA2 < WB2
RR
![Page 47: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/47.jpg)
Pisces: Queue Tenants By Dominant Resource
Bandwidth limited Request Limited
22
out req out req
Tenant A Tenant BVM VM VM VM VM VM
Track per-tenantresource vector
Controller
RS
PP
WA WA2 WB2
FQ
WA2 < WB2
RR
![Page 48: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/48.jpg)
Pisces: Queue Tenants By Dominant Resource
Bandwidth limited Request Limited
22
out req out req
Tenant A Tenant BVM VM VM VM VM VM
Track per-tenantresource vector
dominant resource fair share
Controller
RS
PP
WA WA2 WB2
FQ
WA2 < WB2
RR
![Page 49: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/49.jpg)
Pisces: Queue Tenants By Dominant Resource
Bandwidth limited Request Limited
22
out req out req
Tenant A Tenant BVM VM VM VM VM VM
Track per-tenantresource vector
dominant resource fair share
Controller
RS
PP
WA WA2 WB2
FQ
WA2 < WB2
RR
![Page 50: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/50.jpg)
Pisces Mechanisms Solve For Global Fairness
23Timescale
Syst
em
Vis
ibil
ity
RS
dominant resourcefair shares
PP
WA WA2 WB2
FQ
![Page 51: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/51.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutesTimescale
Syst
em
Vis
ibil
ity glob
al
RS
Con
trol
ler
dominant resourcefair shares
PP
WA WA2 WB2
FQ
![Page 52: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/52.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsTimescale
Syst
em
Vis
ibil
ity glob
al
RS
Con
trol
ler
dominant resourcefair shares
PP
WA WA2 WB2
FQ
![Page 53: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/53.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
PP
WA WA2 WB2
FQ
![Page 54: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/54.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
PP
WA WA2 WB2
FQ
![Page 55: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/55.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
PP
WA WA2 WB2
FQ
![Page 56: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/56.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
PP
WA WA2 WB2
FQ
![Page 57: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/57.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 58: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/58.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 59: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/59.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 60: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/60.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 61: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/61.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 62: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/62.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 63: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/63.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 64: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/64.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Maximum bottleneck flow weight exchange
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 65: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/65.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Maximum bottleneck flow weight exchange
FAST-TCP basedreplica selection
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 66: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/66.jpg)
Pisces Mechanisms Solve For Global Fairness
23
minutessecondsmicrosecondsTimescale
Syst
em
Vis
ibil
ity
loca
lgl
obal
RS
RR
RR
...
SS
SS
...C
ontr
olle
r
dominant resourcefair shares
Maximum bottleneck flow weight exchange
FAST-TCP basedreplica selection
DRR token-basedDRFQ scheduler
Replica Selection Policies
Wei
ght A
lloca
tions
fairness and capacity constraints
PP
WA WA2 WB2
FQ
![Page 67: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/67.jpg)
Evaluation
24
![Page 68: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/68.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
![Page 69: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/69.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?
![Page 70: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/70.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?- What is the overhead of using Pisces?
![Page 71: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/71.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?- What is the overhead of using Pisces?
•Does Pisces handle mixed workloads?
![Page 72: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/72.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?- What is the overhead of using Pisces?
•Does Pisces handle mixed workloads?
•Does Pisces provide weighted system-wide fairness?
![Page 73: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/73.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?- What is the overhead of using Pisces?
•Does Pisces handle mixed workloads?
•Does Pisces provide weighted system-wide fairness?
•Does Pisces provide local dominant resource fairness?
![Page 74: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/74.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?- What is the overhead of using Pisces?
•Does Pisces handle mixed workloads?
•Does Pisces provide weighted system-wide fairness?
•Does Pisces provide local dominant resource fairness?
•Does Pisces handle dynamic demand?
![Page 75: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/75.jpg)
Evaluation
24
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?- What is the overhead of using Pisces?
•Does Pisces handle mixed workloads?
•Does Pisces provide weighted system-wide fairness?
•Does Pisces provide local dominant resource fairness?
•Does Pisces handle dynamic demand?
•Does Pisces adapt to changes in object popularity?
![Page 76: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/76.jpg)
Evaluation
25
•Does Pisces achieve (even) system-wide fairness?
- Is each Pisces mechanism necessary for fairness?- What is the overhead of using Pisces?
•Does Pisces handle mixed workloads?
•Does Pisces provide weighted system-wide fairness?
•Does Pisces provide local dominant resource fairness?
•Does Pisces handle dynamic demand?
•Does Pisces adapt to changes in object popularity?
![Page 77: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/77.jpg)
Pisces Achieves System-wide Per-tenant Fairness
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
Unmodified Membase
Ideal fair share: 110 kreq/s (1kB requests)
0.57 MMR
Min-Max Ratio: min rate/max rate (0,1]
8 Tenants - 8 Client - 8 Storage NodesZipfian object popularity distribution
![Page 78: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/78.jpg)
Pisces Achieves System-wide Per-tenant Fairness
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80G
ET R
eque
sts
(kre
q/s)
Time (s)
Unmodified Membase
Ideal fair share: 110 kreq/s (1kB requests)
Pisces
0.57 MMR 0.98 MMR
Min-Max Ratio: min rate/max rate (0,1]
8 Tenants - 8 Client - 8 Storage NodesZipfian object popularity distribution
![Page 79: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/79.jpg)
Each Pisces Mechanism Contributes to System-wide Fairness and Isolation
27
Unmodified Membase
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0.57 MMR
![Page 80: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/80.jpg)
Each Pisces Mechanism Contributes to System-wide Fairness and Isolation
27
Unmodified Membase
0.36 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0.57 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
2x vs 1x demand
![Page 81: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/81.jpg)
Each Pisces Mechanism Contributes to System-wide Fairness and Isolation
27
Unmodified Membase
0.59 MMR
0.36 MMR 0.58 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0.57 MMR
FQ
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
2x vs 1x demand
![Page 82: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/82.jpg)
Each Pisces Mechanism Contributes to System-wide Fairness and Isolation
27
Unmodified Membase
0.59 MMR
0.36 MMR 0.58 MMR 0.74 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
PPFQ
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0.57 MMR
FQ
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
2x vs 1x demand
0.64 MMR
![Page 83: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/83.jpg)
Each Pisces Mechanism Contributes to System-wide Fairness and Isolation
27
Unmodified Membase
0.59 MMR 0.93 MMR
0.36 MMR 0.58 MMR 0.74 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
WAPPFQ
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
0.96 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0.57 MMR
FQ
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
2x vs 1x demand
0.64 MMR
![Page 84: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/84.jpg)
Each Pisces Mechanism Contributes to System-wide Fairness and Isolation
27
Unmodified Membase
0.59 MMR 0.93 MMR
0.36 MMR 0.58 MMR 0.74 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
RSWAPPFQ PPFQ
0.90 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
0.96 MMR 0.89 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0.57 MMR
FQ
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
2x vs 1x demand
0.64 MMR
![Page 85: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/85.jpg)
Each Pisces Mechanism Contributes to System-wide Fairness and Isolation
27
Unmodified Membase
0.59 MMR 0.93 MMR 0.98 MMR
0.36 MMR 0.58 MMR 0.74 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
RSWAPPFQ WAPPFQ
0.90 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90 0
20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80 90
0.96 MMR 0.97 MMR0.89 MMR
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
0.57 MMR
FQ
0 20 40 60 80
100 120 140 160 180
10 20 30 40 50 60 70 80
GET
Req
uest
s (k
req/
s)
Time (s)
2x vs 1x demand
0.64 MMR
![Page 86: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/86.jpg)
Pisces Imposes Low-overhead
28
0
875
1750
2625
3500
1kB Requests 10B Requests
Aggregate System Throughput
GET
Req
uest
s (k
req/
s)
Unmodified Membase Pisces
< 5%
> 19%
![Page 87: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/87.jpg)
Pisces Achieves System-wide Weighted Fairness
29
4 heavy hitters 20 moderate demand 40 low demand
![Page 88: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/88.jpg)
Pisces Achieves System-wide Weighted Fairness
29
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (10)10x weight (40)
1x weight (50)
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (4)10x weight (20)
1x weight (40)
0
50
100
150
200
20 25 30 35 40 45 50 55 60
1x weight2x weight
3x weight4x weight
0
5
10
15
20
20 25 30 35 40 45 50 55 60
GE
T R
eque
sts
(kre
q/s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
SE
T R
eque
sts
(kre
q/s)
Time (s) Time (s)4 heavy hitters 20 moderate demand 40 low demand
![Page 89: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/89.jpg)
Pisces Achieves System-wide Weighted Fairness
29
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (10)10x weight (40)
1x weight (50)
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (4)10x weight (20)
1x weight (40)
0
50
100
150
200
20 25 30 35 40 45 50 55 60
1x weight2x weight
3x weight4x weight
0
5
10
15
20
20 25 30 35 40 45 50 55 60
GE
T R
eque
sts
(kre
q/s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
SE
T R
eque
sts
(kre
q/s)
Time (s) Time (s)0.98 MMR
4 heavy hitters 20 moderate demand 40 low demand0.89 MMR 0.91 MMR
![Page 90: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/90.jpg)
Pisces Achieves System-wide Weighted Fairness
29
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (10)10x weight (40)
1x weight (50)
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (4)10x weight (20)
1x weight (40)
0
50
100
150
200
20 25 30 35 40 45 50 55 60
1x weight2x weight
3x weight4x weight
0
5
10
15
20
20 25 30 35 40 45 50 55 60
GE
T R
eque
sts
(kre
q/s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
SE
T R
eque
sts
(kre
q/s)
Time (s) Time (s)0.98 MMR
4 heavy hitters 20 moderate demand 40 low demand0.89 MMR 0.91 MMR
0.91 MMR
![Page 91: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/91.jpg)
Pisces Achieves System-wide Weighted Fairness
29
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (10)10x weight (40)
1x weight (50)
0 20 40 60 80
100 120 140 160
25 30 35 40 45 50 55 60
100x weight (4)10x weight (20)
1x weight (40)
0
50
100
150
200
20 25 30 35 40 45 50 55 60
1x weight2x weight
3x weight4x weight
0
5
10
15
20
20 25 30 35 40 45 50 55 60
GE
T R
eque
sts
(kre
q/s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
GE
T R
eque
sts
(kre
q/s)
Time (s)
SE
T R
eque
sts
(kre
q/s)
Time (s) Time (s)0.98 MMR
4 heavy hitters 20 moderate demand 40 low demand
0.56 MMR
0.89 MMR 0.91 MMR
0.91 MMR
![Page 92: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/92.jpg)
Pisces Achieves Dominant Resource Fairness
30
1kB workloadbandwidth limited
10B workloadrequest limited
![Page 93: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/93.jpg)
0
100
200
300
400
500
600
700
800
20 25 30 35 40 45 50 55 60
1kB bandwidth limited10B request limited
0
50
100
150
200
250
300
20 25 30 35 40 45 50 55 60
Pisces Achieves Dominant Resource Fairness
30
Time (s)
Band
wid
th (
Mb/
s)
1kB workloadbandwidth limited
10B workloadrequest limited
![Page 94: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/94.jpg)
0
100
200
300
400
500
600
700
800
20 25 30 35 40 45 50 55 60
1kB bandwidth limited10B request limited
0
50
100
150
200
250
300
20 25 30 35 40 45 50 55 60
Pisces Achieves Dominant Resource Fairness
30
Time (s)
Band
wid
th (
Mb/
s)
GET
Req
uest
s (k
req/
s)
Time (s)
1kB workloadbandwidth limited
10B workloadrequest limited
![Page 95: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/95.jpg)
0
100
200
300
400
500
600
700
800
20 25 30 35 40 45 50 55 60
1kB bandwidth limited10B request limited
0
50
100
150
200
250
300
20 25 30 35 40 45 50 55 60
Pisces Achieves Dominant Resource Fairness
30
Time (s)
Band
wid
th (
Mb/
s)
GET
Req
uest
s (k
req/
s)
76% of bandwidth 76% of request rate
Time (s)
1kB workloadbandwidth limited
10B workloadrequest limited
![Page 96: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/96.jpg)
0
100
200
300
400
500
600
700
800
20 25 30 35 40 45 50 55 60
1kB bandwidth limited10B request limited
0
50
100
150
200
250
300
20 25 30 35 40 45 50 55 60
Pisces Achieves Dominant Resource Fairness
30
Time (s)
Band
wid
th (
Mb/
s)
GET
Req
uest
s (k
req/
s)
76% of bandwidth 76% of request rate
Time (s)
1kB workloadbandwidth limited
10B workloadrequest limited
24% of request rate
![Page 97: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/97.jpg)
Pisces Adapts to Dynamic Demand
31
Constant BurstyDiurnal (2x wt)
Tenant Demand
![Page 98: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/98.jpg)
Pisces Adapts to Dynamic Demand
31
Constant BurstyDiurnal (2x wt)
0
50
100
150
200
0 10 20 30 40 50 60 70 80 90
GET
Req
uest
s (k
req/
s)
Time (s)
Tenant Demand
![Page 99: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/99.jpg)
Pisces Adapts to Dynamic Demand
31
Constant BurstyDiurnal (2x wt)
~2x
0
50
100
150
200
0 10 20 30 40 50 60 70 80 90
GET
Req
uest
s (k
req/
s)
Time (s)
Tenant Demand
![Page 100: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/100.jpg)
Pisces Adapts to Dynamic Demand
31
Constant BurstyDiurnal (2x wt)
~2x
even
0
50
100
150
200
0 10 20 30 40 50 60 70 80 90
GET
Req
uest
s (k
req/
s)
Time (s)
Tenant Demand
![Page 101: Performance Isolation and Fairness for Multi-Tenant Cloud ...mfreed/docs/pisces-osdi12-slides.pdf · Performance Isolation and Fairness for Multi-Tenant Cloud Storage David Shue*,](https://reader036.vdocuments.mx/reader036/viewer/2022062601/5e02f540d9e2ea2f20411633/html5/thumbnails/101.jpg)
Conclusion
•Pisces Contributions- Per-tenant weighted max-min fair shares of system-wide
resources w/ high utilization
- Arbitrary object distributions
- Different resource bottlenecks
- Novel decomposition into 4 complementary mechanisms
32
PPPartition
Placement WA RS FQWeight
AllocationReplica
SelectionFair
Queuing