Difference between revisions of "Partitioning Cache Data"

From Adpnwiki
Jump to navigation Jump to search
Line 51: Line 51:
 
ADAH2 518 3.94 3.95
 
ADAH2 518 3.94 3.95
 
BPL2 519 3.94 3.95
 
BPL2 519 3.94 3.95
35.76 35.91 -12.52% -12.49%</nowiki>
+
35.76 35.91 -12.52% -12.49%</nowiki>
  
 
''1+6'' with 4 additional nodes
 
''1+6'' with 4 additional nodes
  <nowiki>au_host AUCount ContentSize in TB DiskCost in TB
+
  <nowiki>au_host AUCount ContentSize in TB DiskCost in TB
ADAH 503 2.60 2.61 -35.35% -49.11%
+
ADAH 503 2.60 2.61 -35.35% -49.11%
ADAH2 407 2.52 2.54 -47.69% -50.61%
+
AUB 425 3.16 3.17 -45.37% -38.13%
AUB 425 3.16 3.17 -45.37% -38.13%
+
BPL 405 2.47 2.48 -47.94% -51.57%
AUB2 501 3.25 3.26 -35.60% -36.42%
+
SHC 408 2.83 2.84 -47.56% -44.66%
BPL 405 2.47 2.48 -47.94% -51.57%
+
TROY 401 3.34 3.35 -48.46% -34.68%
BPL2 457 2.86 2.87 -41.26% -43.98%
+
UAB 413 2.50 2.51 -46.92% -51.13%
SHC 408 2.83 2.84 -47.56% -44.66%
+
UAT 740 4.79 4.81 -4.88% -6.23%
TROY 401 3.34 3.35 -48.46% -34.68%
+
UNA 357 2.97 2.98 -54.11% -41.96%
UAB 413 2.50 2.51 -46.92% -51.13%
+
ADAH2 407 2.52 2.54
UAT 740 4.79 4.81 -4.88% -6.23%
+
AUB2 501 3.25 3.26
UAT2 428 2.48 2.49 -44.99% -51.52%
+
BPL2 457 2.86 2.87
UNA 357 2.97 2.98 -54.11% -41.96% -41.68% -41.67%
+
UAT2 428 2.48 2.49
35.76 35.91</nowiki>
+
35.76 35.91 -41.32% -39.68%</nowiki>
  
 
Looking at ''1+5'' which is 6 discrete nodes in the network (double quorum), base storage decrease is 25% with no additional nodes. ''1+5'' with 4 additional nodes achieves a staggering 50% on average per node storage reduction.  
 
Looking at ''1+5'' which is 6 discrete nodes in the network (double quorum), base storage decrease is 25% with no additional nodes. ''1+5'' with 4 additional nodes achieves a staggering 50% on average per node storage reduction.  
Line 74: Line 74:
 
  <nowiki>
 
  <nowiki>
 
au_host AUCount ContentSize in TB DiskCost in TB
 
au_host AUCount ContentSize in TB DiskCost in TB
ADAH 581 3.8384 3.854856396 -25.32% -24.87%
+
ADAH 581 3.83 3.85 -25.32% -24.87%
AUB 539 3.9789 3.996972466 -30.72% -22.12%
+
AUB 539 3.97 3.99 -30.72% -22.12%
BPL 537 3.7228 3.737751961 -30.98% -27.13%
+
BPL 537 3.72 3.73 -30.98% -27.13%
SHC 590 3.3681 3.383465672 -24.16% -34.07%
+
SHC 590 3.36 3.38 -24.16% -34.07%
TROY 555 3.7953 3.812593842 -28.66% -25.71%
+
TROY 555 3.79 3.81 -28.66% -25.71%
UAB 572 3.6851 3.701880264 -26.48% -27.87%
+
UAB 572 3.68 3.70 -26.48% -27.87%
UAT 757 4.9206 4.939711952 -2.70% -3.68%
+
UAT 757 4.92 4.93 -2.70% -3.68%
UNA 536 3.3432 3.356886768 -31.11% -34.56%
+
UNA 536 3.34 3.35 -31.11% -34.56%
30.6524 30.78411932 -25.02% -25.00%</nowiki>
+
30.65 30.78 -25.02% -25.00%</nowiki>
  
 
''1+5'' with 4 additional nodes
 
''1+5'' with 4 additional nodes
  
  <nowiki>au_host AUCount ContentSize in TB DiskCost in TB
+
  <nowiki>au_host AUCount ContentSize in TB DiskCost in TB
ADAH 395 1.9515 1.961049938 -49.23% -61.80%
+
ADAH 395 1.95 1.96 -49.23% -61.80%
ADAH2 332 2.0561 2.066130066 -57.33% -59.75%
+
AUB 349 2.90 2.91 -55.14% -43.30%
AUB 349 2.8968 2.908515167 -55.14% -43.30%
+
BPL 463 2.77 2.78 -40.49% -45.74%
AUB2 328 2.5209 2.531873703 -57.84% -50.65%
+
SHC 350 2.94 2.95 -55.01% -42.46%
BPL 463 2.7718 2.782540035 -40.49% -45.74%
+
TROY 327 1.83 1.84 -57.97% -64.18%
BPL2 367 2.5277 2.538369083 -52.83% -50.52%
+
UAB 331 2.10 2.11 -57.46% -58.88%
SHC 350 2.9395 2.95199995 -55.01% -42.46%
+
UAT 736 4.85 4.87 -5.40% -5.02%
TROY 327 1.83 1.839459991 -57.97% -64.18%
+
UNA 312 2.37 2.38 -59.90% -53.55%
UAB 331 2.1005 2.109290314 -57.46% -58.88%
+
ADAH2 332 2.06 2.07
UAT 736 4.852 4.871892166 -5.40% -5.02%
+
AUB2 328 2.52 2.53
UAT2 377 1.833 1.840956402 -51.54% -64.12%
+
BPL2 367 2.53 2.54
UNA 312 2.3728 2.382042503 -59.90% -53.55% -50.01% -50.00%
+
UAT2 377 1.83 1.84
30.65 30.78 </nowiki>
+
30.65 30.78 -47.57% -46.87%</nowiki>

Revision as of 09:43, 17 June 2013

Partitioning Cache Data

Quorum in the network is 3. (This is down from the previous value of 4... find out why).


Any sort of partitioning strategy would need to be implemented at the titledb.xml level. Title AUs can be assigned to peers centrally, and each peer should receive a custom titledb.xml file. If LOCKSS is unwilling to support that then there is alternative using local parameters.

org.lockss.titleDbs = http://bpldb.bplonline.org/etc/adpn/titledb-local.xml

Implementing a 1 + 6 partitioning strategy can save 12% on average for each network node. 1 + 6 indicates AU owner + 6 additional network nodes. Adding 2 additional nodes to the network can decrease per node storage by an average of 30%. Adding 4 additional nodes and partitioning cache data can save per node storage on average of 41%. This means we could store up to 18 TB of data on 10.6 TB nodes.

Implementing a 1 + 5 which is 6 discrete nodes in the network (double quorum), base storage decrease is 25% with no additional nodes. 1 + 5 with 4 additional nodes achieves a staggering 50% on average per node storage reduction.

All nodes (default configuration)

			
au_host	AUCount	ContentSize in TB	DiskCost in TB
ADAH	778	5.11	5.13
AUB	778	5.11	5.13
BPL	778	5.11	5.13
SHC	778	5.11	5.13
TROY	778	5.11	5.13
UAB	778	5.11	5.13
UAT	778	5.11	5.13
UNA	778	5.11	5.13
		40.88	41.04

Doesn’t include vacated publisher AUs (which is between 500 and 600 GB).

1 + 6 no additional nodes

au_host	AUCount	ContentSize	DiskCost	count	size	cost
ADAH	667	4.60	4.62		-14.27%	-9.99%	-9.97%
AUB	676	3.83	3.85		-13.11%	-25.00%	-24.94%
BPL	669	4.49	4.51		-14.01%	-12.19%	-12.15%
SHC	666	4.48	4.49		-14.40%	-12.41%	-12.41%
TROY	666	4.29	4.30		-14.40%	-16.13%	-16.09%
UAB	667	4.58	4.60		-14.27%	-10.29%	-10.26%
UAT	767	5.06	5.09		-1.41%	-0.89%	-0.86%
UNA	667	4.43	4.45		-14.27%	-13.27%	-13.23%
		35.76	35.91			-12.52%	-12.49%

1 + 6 with 2 additional nodes

au_host	AUCount	ContentSize	DiskCost	count	size	cost
ADAH	519	3.93	3.95		-33.29%	-23.00%	-22.98%
AUB	540	3.21	3.23		-30.59%	-37.15%	-37.09%
BPL	522	3.09	3.11		-32.90%	-39.49%	-39.44%
SHC	519	3.22	3.23		-33.29%	-36.97%	-36.96%
TROY	519	3.26	3.28		-33.29%	-36.15%	-36.09%
UAB	519	3.02	3.03		-33.29%	-40.93%	-40.89%
UAT	751	4.89	4.91		-3.47%	-4.40%	-4.38%
UNA	519	3.26	3.28		-33.29%	-36.15%	-36.09%
ADAH2	518	3.94	3.95				
BPL2	519	3.94	3.95				
		35.76	35.91		-12.52%	-12.49%

1+6 with 4 additional nodes

au_host	AUCount	ContentSize in TB	DiskCost in TB		
ADAH	503	2.60	2.61	-35.35%	-49.11%
AUB	425	3.16	3.17	-45.37%	-38.13%
BPL	405	2.47	2.48	-47.94%	-51.57%
SHC	408	2.83	2.84	-47.56%	-44.66%
TROY	401	3.34	3.35	-48.46%	-34.68%
UAB	413	2.50	2.51	-46.92%	-51.13%
UAT	740	4.79	4.81	-4.88%	-6.23%
UNA	357	2.97	2.98	-54.11%	-41.96%
ADAH2	407	2.52	2.54		
AUB2	501	3.25	3.26		
BPL2	457	2.86	2.87		
UAT2	428	2.48	2.49		
		35.76	35.91	-41.32%	-39.68%

Looking at 1+5 which is 6 discrete nodes in the network (double quorum), base storage decrease is 25% with no additional nodes. 1+5 with 4 additional nodes achieves a staggering 50% on average per node storage reduction.

1+5 with no additional nodes

							
au_host	AUCount	ContentSize in TB	DiskCost in TB				
ADAH	581	3.83	3.85	-25.32%	-24.87%		
AUB	539	3.97	3.99	-30.72%	-22.12%		
BPL	537	3.72	3.73	-30.98%	-27.13%		
SHC	590	3.36	3.38	-24.16%	-34.07%		
TROY	555	3.79	3.81	-28.66%	-25.71%		
UAB	572	3.68	3.70	-26.48%	-27.87%		
UAT	757	4.92	4.93	-2.70%	-3.68%		
UNA	536	3.34	3.35	-31.11%	-34.56%		
		30.65	30.78	-25.02%	-25.00%

1+5 with 4 additional nodes

au_host	AUCount	ContentSize in TB	DiskCost in TB		
ADAH	395	1.95	1.96	-49.23%	-61.80%
AUB	349	2.90	2.91	-55.14%	-43.30%
BPL	463	2.77	2.78	-40.49%	-45.74%
SHC	350	2.94	2.95	-55.01%	-42.46%
TROY	327	1.83	1.84	-57.97%	-64.18%
UAB	331	2.10	2.11	-57.46%	-58.88%
UAT	736	4.85	4.87	-5.40%	-5.02%
UNA	312	2.37	2.38	-59.90%	-53.55%
ADAH2	332	2.06	2.07		
AUB2	328	2.52	2.53		
BPL2	367	2.53	2.54		
UAT2	377	1.83	1.84		
		30.65	30.78	-47.57%	-46.87%