• 沒有找到結果。

Data Storage

5.3.3 Communication Costs

In the third part, we calculate the communication costs (network workload) between one server and two clients.

• Exp3 : communication costs (network workload)

Since the wireless module (chip and antenna) of the smartphone has poor quality, the con-nect bandwidth is between 300KBps to 500KBps. The smartphone use an AP to concon-nect to the Internet. We use iftop at file server to calculate the network communication of specific connection. At client, we execute a sequence of file operations including create, remove, read, write, and list, then record the total network usage.

S = {(hi, fi, oi, ti)+}

– hi:host – fi:file

– oi:file operation – ti:time

Here is two clients (hosts) and one server. First, host 1 reads file 1, so that host 1 down-loads file 1. After reading file 1, host 1 modifies some data then closes file 1, so that host 1 uploads file 1 to the file server. Since host 1 accesses file 1, file server records the information of host 1. Second, host 2 reads the same file, file 1, so that host 2 also down-loads file 1. After reading file 1, host 2 modifies some data then close file 1. Then, host 2 uploads new file 1 to the file server. The file server gets the data of file 1 and records the information of host 2. The file server finds that host 1 also accesses file 1, so file server sends a notification automatically to host 1 to announce host 1 updating file 1.

In Table 5.5, NFS is a traditional network file system, AshFS, Coda, and MoFS are file systems supporting mobility. Since this experiment is the calculation of communication costs, we just compare with the network file systems. The connection bandwidth is be-tween 800KBps to 1MBps on laptop. All the file system use the same testing sequence to measure the communication cost (like ). NFS executes all the file operations during the network so that it has the most network usage. AshFS caches file in local cache but it need to check the file status frequently, this may exhaust the communication costs. Using

``rsync'' is the other factor cause AshFS exhaust the communication. MoFS does not fre-quently check the file status, and resolve the network. Since MoFS does not implement

Client 1

require file1

Server

open file1

download file1

write file1

write file1

close file1

upload file1

sequence

Client 2

sequenc server notification

Figure 5.6:Experiment flow for Communication Cost

differential upload, the communication cost of AshFS is similar to the communication cost of MoFS. If we use differential upload instead of whole file upload, the communication cost of MoFS can be more less.

Table 5.5: Communication Costs of MoFS/Laptop (KB)

NFS Coda AshFS MoFS

Network Usage 386 273 233 225

Chapter 6 Conclusion

In this research, we design and implement a mobile file system, named MoFS. The new proposed file system, MoFS supports disconnected operations with a persistent cache and de-creases the communication costs at file status notification from the MoFS server. Taking FUSE as it underlying file system library, MoFS requires no kernel patch in its deployment. Also, by integrating with the SSH protocol, MoFS can provide user authentication and file encryption to protect communication sessions between file servers and clients.

The MoFS client program has been implemented on both x86- and ARM-based platforms (Openmoko Freerunner GTA02). We made several experiments to compare MoFS and other file systems (Coda, AshFS, NFS, Ext3) in terms of processing latencies, read/write throughputs, communication costs. From the experiment results, we conclude that MoFS/x86 and MoFS/

ARM have the best performance in retrieving file stats and removing files. For read/write throughputs, MoFS/x86 and MoFS/ARM have the best performance. Also, comparatively, we observe that less network traffic (communication costs) is required when running the MoFS client on Laptop. In short, MoFS has faster processing time, better read/write throughputs and lower communication costs.

References

[1] M. Satyanarayanan, J. Kistler, P. Kumar, M. Okasaki, E. Siegel, and D. Steere, ''Coda: A Highly Available File System for a Distributed Workstation Environment,'' IEEE Transac-tions on Computers, vol. 39, no. 4, pp. 447--459, 1990.

[2] P. J. Braam, ''The Coda Distributed File System,'' Linux Journal, p. 6, 1998.

[3] J. Lui, O. So, and T. Tam, ''NFS/M: An Open Platform Mobile File System,'' in Proc.

18th International Conference on Distributed Computing Systems, 26--29 May 1998, pp.

488--495.

[4] ''FUSE: Filesystem in Userspace.'' [Online]. Available: http://fuse.sourceforge.net/

[5] I. Voras and M. Zagar, ''Network Distributed File System in User Space,'' in Proc. 28th International Conference on Information Technology Interfaces, 2006, pp. 669--674.

[6] M. E. Hoskins, ''SSHFS: super easy file access over SSH,'' Linux Journal, vol. 2006, no.

146, p. 4, 2006.

[7] M. Szeredi, ''SSH Filesystem,'' 2005. [Online]. Available: http://fuse.sourceforge.net/

sshfs.html

[8] L.-Y. Chang, ''AshFS: A Lightweight Mobile File System Supporting Disconnected Oper-ations,'' Master's thesis, Department of Electrical and Control Engineering, National Chiao Tung University, 2008.

[9] Marionraven and T. M, ''HTTPFS,'' 2006. [Online]. Available: http://httpfs.sourceforge.

net/

[10] B. Callaghan, B. Pawlowski, and P. Staubach, ''NFS Version 3 Protocol Specification,'' RFC 1813 (Informational), Internet Engineering Task Force, Jun. 1995. [Online].

Available: http://www.ietf.org/rfc/rfc1813.txt

[11] B. Atkin and K. Birman, ''Network-Aware Adaptation Techniques for Mobile File Sys-tems,'' in Proc. Fifth IEEE International Symposium on Network Computing and Applica-tions NCA 2006, 2006, pp. 181--188.

[12] S. Lehtinen and C. Lonvick, ''The Secure Shell (SSH) Protocol Assigned Numbers,'' RFC 4250 (Proposed Standard), Internet Engineering Task Force, Jan. 2006. [Online].

Available: http://www.ietf.org/rfc/rfc4250.txt

[13] T. Ylonen and C. Lonvick, ''The Secure Shell (SSH) Protocol Architecture,'' RFC 4251 (Proposed Standard), Internet Engineering Task Force, Jan. 2006. [Online]. Available:

http://www.ietf.org/rfc/rfc4251.txt

[14] T. Ylonen and C. Lonvick, ''The Secure Shell (SSH) Authentication Protocol,'' RFC 4252 (Proposed Standard), Internet Engineering Task Force, Jan. 2006. [Online]. Available:

http://www.ietf.org/rfc/rfc4252.txt

[15] T. Ylonen and C. Lonvick, ''The Secure Shell (SSH) Transport Layer Protocol,'' RFC 4253 (Proposed Standard), Internet Engineering Task Force, Jan. 2006. [Online]. Available:

http://www.ietf.org/rfc/rfc4253.txt

[16] T. Ylonen and C. Lonvick, ''The Secure Shell (SSH) Connection Protocol,'' RFC 4254 (Proposed Standard), Internet Engineering Task Force, Jan. 2006. [Online]. Available:

http://www.ietf.org/rfc/rfc4254.txt

[17] B. Cornell, P. Dinda, and F. Bustamante, ''Wayback: A User-level Versioning File

Sys-tem for Linux,'' in Proceedings of Usenix Annual Technical Conference, FREENIX Track, 2004, pp. 19--28.

[18] A. Boukerche and R. Al-Shaikh, ''Towards Building a Fault Tolerant and Conflict-Free Distributed File System for Mobile Clients,'' vol. 2, April 2006, pp. 6 pp.--.

[19] A. Helal, A. Khushraj, and J. Zhang, ''Incremental Hoarding and Reintegration in Mobile Environments,'' 2002, pp. 8--11.

[20] D. Dwyer and V. Bharghavan, ''A Mobility-Aware File System for Partially Connected Operation,'' SIGOPS Oper. Syst. Rev., vol. 31, no. 1, pp. 24--30, 1997.

[21] R. Coker, ''The Bonnie++ Benchmark,'' 1999. [Online]. Available: http://www.coker.

com.au/bonnie++/

[22] I. Dowse and D. Malone, ''Recent Filesystem Optimisations on FreeBSD,'' in Proceedings of the FREENIX Track: 2002 USENIX Annual Technical Conference. Berkeley, CA, USA: USENIX Association, 2002, pp. 245--258.

[23] P. Warren and C. Lightfoot, ''iftop: Display bandwidth usage on an interface,'' Feb 12th 2006. [Online]. Available: http://www.ex-parrot.com/pdw/iftop/

[24] D. A. Bandel, ''Focus on Software,'' Linux Journal, vol. 2002, no. 99, p. 12, 2002.

相關文件