Resumen del contenido incluido en la página 1
Integrating Portable and
Distributed Storage
N. Tolia, J. Harkes, M. Kozuch, M. Satyanarayanan
IRP-TR-03-10
INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL®
PRODUCTS. NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO
ANY INTELLECTUAL PROPERTY RIGHTS IS GRANTED BY THIS DOCUMENT. EXCEPT
AS PROVIDED IN INTEL'S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS,
INTEL ASSUMES NO LIABILITY WHATSOEVER, AND INTEL DISCLAIMS ANY EXPRESS
OR
Resumen del contenido incluido en la página 2
Integrating Portable and Distributed Storage Niraj Tolia , Jan Harkes , Michael Kozuch , M. Satyanarayanan Carnegie Mellon University, Intel Research Pittsburgh Abstract that combines the strengths of distributed le sys- We describe a technique called lookaside caching that combines the tems and portable storage devices, while negating their strengths of distributed le systems and portable storage devices, weaknesses. In spite of its simplicity, this technique while negating thei
Resumen del contenido incluido en la página 3
2 Background devices, a system such as PersonalRAID can be valu- able in managing complexity. To understand the continuing popularity of portable Consistency: Without explicit user effort, a dis- storage, it is useful to review the strengths and weak- tributed le system presents the latest version of a le nesses of portable storage and distributed le systems. when it is accessed. In contrast, a portable device has While there is considerable variation in the designs of to be explicitly kept u
Resumen del contenido incluido en la página 4
there should be no compromise of robustness, consis- al. [22]. In that work, a recipe is an XML description of tency or security. There should also be no added com- le content that enables block-level reassembly of the plexity in sharing and collaboration. Finally, the design le from content-addressable storage. One can view should be tolerant of human error: improper use of the the hash of a le as the smallest possible recipe for it. portable storage device (such as using the wrong de- The i
Resumen del contenido incluido en la página 5
cfs lka --clear exclude all indexes Kernel Size Files Bytes Release Days cfs lka +db1 include index db1 Version (MB) Same Same Date Stale cfs lka -db1 exclude index db1 2.4.18 118.0 100% 100% 02/25/02 0 cfs lka --list print lookaside statistics 2.4.17 116.2 90% 79% 12/21/01 66 2.4.13 112.6 74% 52% 10/23/01 125 Figure 1. Lookaside Commands on Client 2.4.9 108.0 53% 30% 08/16/01 193 2.4.0 95.3 28% 13% 01/04/01 417 at a specied pathname. It computes the SHA-1 hash This table shows key characterist
Resumen del contenido incluido en la página 6
Measured Data Rate 100% File Size Read (Mb/s) Write (Mb/s) 80% 4 KB 6.3 7.4 16 KB 6.3 12.5 60% 64 KB 16.7 25.0 256 KB 25.0 22.2 40% 1 MB 28.6 25.8 10 MB 29.3 26.4 20% 100 MB 29.4 26.5 0% This tables displays the measured read and write bandwidths for different le sizes on the portable storage device used in our experiments. To discount caching effects, we unmounted and remounted the device before each trial. For the same rea- 2.4.0 2.4.1 2.4.2 2.4.3 2.4.4 2.4.5 2.4.6 2.4.7 son, all writes were
Resumen del contenido incluido en la página 7
Lookaside Device State Bandwidth No Device 2.4.18 2.4.17 2.4.13 2.4.9 2.4.0 100 Mb/s 287.7 (5.6) 292.7 (6.4) 324.7 (16.4) 346.4 (6.9) 362.7 (3.4) 358.1 (7.7) [-1.7%] [-12.9%] [-20.4%] [-26.1%] [-24.5%] 10 Mb/s 388.4 (12.9) 282.9 (8.3) 364.8 (12.4) 402.7 (2.3) 410.9 (2.1) 421.1 (12.8) [27.1%] [6.1%] [-3.7%] [-5.8%] [-8.4%] 1 Mb/s 1148.3 (6.9) 424.8 (3.1) 543.6 (11.5) 835.8 (3.7) 1012.4 (12.0) 1094.3 (5.4) [63.0%] [52.7%] [27.2%] [11.8%] [4.7%] 100 Kb/s 9348.8 (84.3) 884.9 (12.0) 3011.2 (167.6) 58
Resumen del contenido incluido en la página 8
1 suspend. Lookaside caching can then reduce the per- How slow is the resume step? formance overhead of cache misses at the resume site. This speed is determined by the time to fetch and decompress the physical memory image of the A different use of lookaside caching for ISR is VM that was saved at suspend. This is the smallest based on the observation that there is often substantial part of total VM state that must be present to begin commonality in VM state across users. For example, executi
Resumen del contenido incluido en la página 9
Number of Length Update Working Trace Operations (Hours) Ops. Set (MB) No With Lookaside Lookaside Win purcell 87739 27.66 6% 252 messiaen 44027 21.27 2% 227 100 Mb/s 14 (0.5) 13 (2.2) 7.1% robin 37504 15.46 7% 85 10 Mb/s 39 (0.4) 12 (0.5) 69.2% berlioz 17917 7.85 8% 57 1 Mb/s 317 (0.3) 12 (0.3) 96.2% 100 Kb/s 4301 (0.6) 12 (0.1) 99.7% This table summarizes the le system traces used for the benchmark described in Section 5.3. Update Ops. only refer This table shows the resume latency (in seco
Resumen del contenido incluido en la página 10
Lookaside Device State Trace Bandwidth No Device 100% 66% 33% 100 Mb/s 50.1 (2.6) 53.1 (2.4) 50.5 (3.1) 48.8 (1.9) Purcell 10 Mb/s 61.2 (2.0) 55.0 (6.5) 56.5 (2.9) 56.6 (4.6) 1 Mb/s 292.8 (4.1) 178.4 (3.1) 223.5 (1.8) 254.2 (2.0) 100 Kb/s 2828.7 (28.0) 1343.0 (0.7) 2072.1 (30.8) 2404.6 (16.3) 100 Mb/s 26.4 (1.6) 31.8 (0.9) 29.8 (0.9) 27.9 (0.8) Messiaen 10 Mb/s 36.3 (0.5) 34.1 (0.7) 36.7 (1.5) 37.8 (0.5) 1 Mb/s 218.9 (1.2) 117.8 (0.9) 157.0 (0.6) 184.8 (1.3) 100 Kb/s 2327.3 (14.8) 903.8 (1.4) 14
Resumen del contenido incluido en la página 11
side performance improvement. This form of cooper- ative caching can be especially valuable in situations No With where the clients have LAN connectivity to each other, Lookaside Lookaside Win but poor connectivity to a distant le server. The heavy 100 Mb/s 173 (9) 103 (3.9) 40.1% price of a cache miss on a large le is then borne only 10 Mb/s 370 (14) 163 (2.9) 55.9% by the rst client to access the le. Misses elsewhere 1 Mb/s 2688 (39) 899 (26.4) 66.6% are serviced at LAN speeds, provided th
Resumen del contenido incluido en la página 12
possessed by portable devices, while simultaneously [8] HUGHES, J.F., THOMAS, B.W. Novell's Guide to NetWare 6 Net- works. John Wiley & Sons, 2002. preserving the consistency, robustness and ease of shar- [9] KOZUCH, M., SATYANARAYANAN, M. Internet Suspend/Resume. ing/collaboration provided by distributed le systems. In Proceedings of the Fourth IEEE Workshop on Mobile Computing Systems and Applications (Calicoon, NY, 2002). One can envision many extensions to lookaside [10] MENEZES, A.J., VAN
Resumen del contenido incluido en la página 13
No Lookaside With Lookaside 1000% 1000% 900% 900% 800% 800% 700% 700% 600% 600% 500% 500% 400% 400% 300% 300% 200% 200% 100% 100% 0% 0% Operations Sorted by Slowdown Operations Sorted by Slowdown (a) 100 Mb/s 1000% 1000% 900% 900% 800% 800% 700% 700% 600% 600% 500% 500% 400% 400% 300% 300% 200% 200% 100% 100% 0% 0% Operations Sorted by Slowdown Operations Sorted by Slowdown (b) 10 Mb/s 1000% 1000% 900% 900% 800% 800% 700% 700% 600% 600% 500% 500% 400% 400% 300% 300% 200% 200% 100% 100% 0% 0% Ope
