NOTE: The grade of your report mainly depends on your contents (both quality and quantity).
DO NOT copy sentences of the paper into your report. You need to write them in your own words.
/***** the following is an example of "A" *****/
Paper Title: File System Design for an NFS File
Server Appliance
Student Name: 王小明
Department/Year: 資工研一
Student ID: 491234567
Date: March 10, 2006
Q1. What is the problem the authors are trying to solve?
A: The authors want to design a file system for NFS file server appliance. The
file system can provide fast NFS service, support dynamically growing storage
capacity, support RAID efficiently, and provide the fast consistency check
ability after an unclean shutdown. Moreover, the file system also provides
the file-system backup ability for system roll-back.
Q2. What other approaches or solutions existed at the time that this work was
done?
A: The related works are listed in the following:
1. Episode stored meta-data in files.
2. Lyon and Sandberg describe the NVRAM write cache technique.
Q3. What was wrong with the other approaches or solutions?
A: Episode replicated the entire inode file and all the indirect blocks,
therefore it generated considerable disk I/O and consumed lots of disk space.
Lyon and Sandberg's method wasted more space of NVRAM.
Q4. What is the authors' approach or solution?
A: The authors' approach is to design the Snapshot mechanism for WAFL file
system.
The Snapshots are read-only copies of the entire file system and use a
copy-on-write technique to save disk space. The WAFL stores three meta-data, the
inode file, the block-map, and the inode-map in files, and then we can write
meta-
data blocks anywhere on disk, increase the size of the file system on the fly,
and
enable copy-on-write technique. The WAFL simply creates a snapshot by just
duplicate the root inode due to the tree-of-block design. The WAFL uses
non-volatile RAM to keep a log of NFS requests between two consistency points.
Q5. Why is it better than the other approaches or solutions?
A: By duplicating just the root inode, snapshots can be created very quickly.
Besides, the WAFL also creates a special snapshot called a consistency point.
Hence, we can save space of duplication and provide fast recovery after an
unclean
shutdown. NVRAM can protect file system from any NFS request being lost when
system shutdowns uncleanly. Logging NFS requests to NVRAM can store more
operations per megabyte of NVRAM than recording the data required to reply the
requests.
Q6. How does it perform?
A: Based on SPEC SFS benchmark, WAFL outperforms than other system file server
at average response time.
Q7. Why is this work important?
A: It provides efficiently WAFL file system for NFS file service appliance.
In addition, the utilization of Snapshot and NVRAM techniques can improve
system performance even after unclean system shutdowns.
Q8. Can any improvement be done?
A: The experiments should be more reasonable and considering the real user
workload. Besides, the time period between two consistency points should be
considered more specifically. Current time period is 10 seconds, and I
think it will become a bottleneck of system performance.
/***** the following is an example of "B" *****/
Paper Title: File System Design for an NFS File
Server Appliance
Student Name: 王小明
Department/Year: 資工研一
Student ID: 491234567
Date: March 10, 2006
1.What is the problem the authors are trying to solve?
有些功能,並不適用於一般目地的電腦,它於具有特別需求,使檔案存取佳,且獨立出來的
設備可以讓使用者易於使用,如cisco等,而Network Appliance針對網路需求去特別設計出來的網路
sever - NFS file system sever.並搭配WAFL的技術,來讓我們對網路上的檔案存取更加有效率
2.What other approaches or solutions existed at the time that this work was
done?
存在一些本來就既有的裝置 分成一般目地 如網路印表機或是傳真機,解決一般需求;
而特殊的像Cisco 和 Bay Networks 是用來處理packet routing.
3.What was wrong with the other approaches or solutions?
上面的裝置 未提供networking service的需求:
1. It should provide fast NFS service.如下
2. It should support large file systems (tens of GB) that grow dynamically as
disks are added.
3. It should provide high performance while supporting RAID (Redundant Array
of Independent Disks).
4. It should restart quickly, even after an unclean shutdown due to power
failure
or system crash.
而本來存在的一些裝置,也需解決該效能的問題: "Large file systems make RAID
desirable because the probability of disk failure increases with the number of
disks.
Large file systems require special techniques for fast restart because the file
system
consistency checks for normal UNIX file systems become unacceptably slow as file
systems grow."
4.What is the authors' approach or solution?
使用Snapshots的技巧,當block在即時的檔案系統上做修改或者移動時,就做snapshots花一些
空間去包含這些block的資料,並允許使用者可以在NFS使用指令存取snapshots的內容,或是讓系統管理
者可以新增或刪除snapshots.
5.Why is it better than the other approaches or solutions?
"A Snapshots is useful even if it is kept for just a few hours, because users
usually notice immediately when they have removed an important file. For file
systems
that change slowly, it may make sense to keep Snapshots on-line for longer."
6.How does it perform?
在WAFL上的實作上 利用三種型式 :
1.WAFL can write any file system block (except the one containing the root inode)
to any
location on disk.
2.WAFL can write blocks to disk in any order.
3.WAFL can allocate disk space for many NFS operations at once in a
single write episode.
7.Why is this work important?
The ability to schedule writes for many requests at once enables more
intelligent
allocation policies, and the fact that blocks can be written to any location and
in any
order allows a wide variety of strategies. It is easy to try new block
allocation
strategies without any change to WAFL's on-disk data structures.
8.Can any improvement be done?
The Bolck-Map File ->This technique does not work for WAFL because many
snapshots
can reference a block at the same time.
Creating a Snapshot ->TO aovid locking out incoming NFS requests in writing a
Snapshot
/***** the following is an example of "C" *****/
Paper Title: File System Design for an NFS File
Server Appliance
Student Name: 王小明
Department/Year: 資工研一
Student ID: 491234567
Date: March 10, 2006
1.What is the problem the authors are trying to
solve?
作者認為應該要有一個網路環境，專用的檔案系統。因為這樣可以讓 NFS 對網路環境有
最佳化存取，及這是一個趨勢。並且作者能達到下面四個主要的功能：
a.It should provide fast NFS service.
b.It should support large file systems that grow dynamically as disks are added.
c.It should provide high performance while supporting RAID.
d.It should restart quick,even after an unclean shutdown due to power failure or
system crash.
2.What other approaches or solutions existed at the time that this work was
done?
當時已經存在了許多 NFS。但是這些 NFS 是屬於較一般性，並不是像 WAFL 是專為網路環境而設計的
檔案系統。
3.What was wrong with the other approaches or solutions?
無法對網路有最佳化的運作。及像 FFS 其存放 indoe及data有其固定的位置，但是 WAFL 允許將 data
存放在任何一個地方，除了 root node 之外。如此將使 Disk 更有效率。
4.What is the authors' approach or solution?
作者主要是使用 snapshots 來指向舊的資料，及使用 copy on write 方式，而 root node 永遠指
向最新的資料。並且 WAFL inode 所指向的 block 都在同一層，不像 FFS 有不同層的 indoe。但是
如果資料真的太小的時候，資料就會被直接存在 point 的地方。另外 WAFL 將許多 operation 存在
NVRAM 中的 Log。將許多小的 Disk I/O，轉為一次較大的 Disk write。
5.Why is it better than the other approaches or solutions?
在 Episode 中跟作者的方法差不多，但是作者的 Snapshot 並不會像 Episode 真正拷貝一份 inode
的實體，而只是建立 point。如此將可節省許多 Disk I/O及空間。
6.How does it perform?
從圖六中，可發現 WAFL 的 Average response time ，都較其他檔案系統快二倍以上。
7.Why is this work important?
以作者的假設，在一些專用的應用，就應有專用的設計，才能使其功能最佳化。如 routing 使用
router一般。
8.Can any improvement be done?
No.