Update

2 views
Skip to first unread message

Kyungyong Lee

unread,
Dec 8, 2009, 11:28:17 PM12/8/09
to aci...@googlegroups.com

In last week, I worked on analytically comparing MapReduce, DeeToo, and Sword’s latency and bandwidth consumption performance. I also worked on adding geographical location information at the MapReduce-condor resource finder. The work flow is as follows :  When a node receives a query, such as (other.latitude>0.0&&other.longitude>0.0), the node sends local xml rpc(Information.Info) to get the geographical information. After getting the information, the node inserts the information into the classad for further matching check.

Currently, MapReduce-condor resource finder system supports two kinds of location-based resource finding method. One method uses geographical location as described above, and the other method uses IP address information. Sample IP address query string : requirements = regexp( "<127.123.[0-9]*.[0-9]*:[0-9]*>", MyAddress)

 

Best,

Kyungyong Lee

Pierre St Juste

unread,
Dec 8, 2009, 11:30:58 PM12/8/09
to aci...@googlegroups.com
No update from me.

--

You received this message because you are subscribed to the Google Groups "acis.p2p" group.
To post to this group, send email to aci...@googlegroups.com.
To unsubscribe from this group, send email to acisp2p+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/acisp2p?hl=en.



--
Pierre St Juste

Yonggang Liu

unread,
Dec 9, 2009, 12:40:37 AM12/9/09
to aci...@googlegroups.com
This week I mainly focused on the NMI "build and test" system from U. Wisconsin. It is a system based on Condor and provides flexibility for users to build their applications without doing the software-specific modifications. (Reason: It is not tied to a specific developer tool, revision control system, or testing framework). "build and test" system isolates the execution environment for the test programs, and it also has a set of principles that makes it good for testing programs, such as light weight, central results repository and fault tolerance.
Since we have IPOP running well with Condor, I am looking for ways to build IPOP into their system. I see it will be very helpful for their system because the system scale will have the potential to be greatly enlarged if IPOP is merged. I have run some simple tasks in this system, and I'm still learning it.

Thank you,
Yonggang Liu
Advanced Computing and Information Systems Laboratory
University of Florida

XU, Jiangyan

unread,
Dec 9, 2009, 10:12:30 AM12/9/09
to aci...@googlegroups.com
I have been conducting ec2 experiment to compare bittorrent and nfs transfer time and bandwidth usage pattern with limited bandwidth environment created by tc, as planned last week. The current status is that the scripts were almost ready with some minor bugs. I am fixing them.

Jiangyan

twcho...@gmail.com

unread,
Dec 9, 2009, 10:40:00 AM12/9/09
to acis.p2p
I'm about to test Deetoo on Planet-lab.
One small problem;
When I need to set xmlrpc server for each 'cache' or 'query' node,
each node's address should be acquired.
That makes it to use of "URL = "http://127.0.0.1:10001/brunet:node:
5DPDBOL5JMLFSHERS7N33FDIJRUE3JXM.rem"".

I think this is time and effort consumption.
Thus, I'd like to add another method for Add.

public void Add(Node node, bool isCacheNode) {
string type;
if (isCacheNode) {
type = "cache";
}
else {
type = "query";
}
Add(Node, type);
}

The above method, I believe, make my work a lot easier.
What do you think?

P. Oscar Boykin

unread,
Dec 15, 2009, 3:23:11 PM12/15/09
to aci...@googlegroups.com
Did you settle this?

This question is not very clear to me. Where do you propose to add
the method "Add"?

It seems to clearly know a lot about Deetoo, so it seems like it can
only make sense in the Deetoo specific code.

By the way, the ternary operator (in C, C++, Java and C#) might help:

Add(Node n, bool iscache) {
Add(n, iscache ? "cache" : "query");
}

is the same as your code.

The ternary operator is nice:

(boolean expression) ? (true case) : (false case)

Also, you can define a constant:

string _CACHE = "cache"
string _QUERY = "query"

so the compiler can check if you mistype something, which is a common
bug with strings.

Best,
> --
>
> You received this message because you are subscribed to the Google Groups "acis.p2p" group.
> To post to this group, send email to aci...@googlegroups.com.
> To unsubscribe from this group, send email to acisp2p+u...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/acisp2p?hl=en.
>
>
>



--
P. Oscar Boykin http://boykin.acis.ufl.edu
Assistant Professor, Department of Electrical and Computer Engineering
University of Florida

Kyungyong Lee

unread,
Dec 15, 2009, 11:29:10 PM12/15/09
to aci...@googlegroups.com

Last week, I mainly worked on simulating Sword on top of Brunet while debugging Sword simulator. While doing simulation, I got dht value size problem. At current Dht implementation, it seems that an entry’s value size is limited to 1024 byte.

http://github.com/johnynek/brunet/blob/master/src/dht/TableServer.cs#L64

 

Sword saves a node’s all attributes’ information at dht. Assuming an average attribute size is 20 bytes, and 100 attributes exist in a node.  The total attribute size would be 2000 bytes, which exceeds the dht value size restriction. I manually increased the dht value maximum size as needed to solve the issue. What problem can occur if we do not limit the maximum dht entry value size?

 

Best,

Kyungyong Lee

 

Pierre St Juste

unread,
Dec 16, 2009, 12:19:56 AM12/16/09
to aci...@googlegroups.com
I plan on doing a quick summery of Brunet and MS P2P Framework at the meeting.

--

You received this message because you are subscribed to the Google Groups "acis.p2p" group.
To post to this group, send email to aci...@googlegroups.com.
To unsubscribe from this group, send email to acisp2p+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/acisp2p?hl=en.



--
Pierre St Juste

P. Oscar Boykin

unread,
Dec 16, 2009, 10:10:11 AM12/16/09
to aci...@googlegroups.com

In a real system using UDP, packets larger than 1500 bytes may not be transmitted (due to poor support for fragmentation).

TCP doesn't have this problem.

We could do fragmentation at the Brunet layer, or we would fragment at the dht layer.  This problem needs to be solved eventually, but there has not yet been a clear winner in terms of how to solve it.

David has started a tcp implementation that could run on top of brunet.  That might be the best way to go because a tcp implementation could allow us to do the dht on tcp on brunet, and any other application could use it too.

Best

--

Kyungyong Lee

unread,
Dec 16, 2009, 10:31:41 AM12/16/09
to aci...@googlegroups.com

OK. It makes sense. 1500 bytes are the maximum allowed packet size at the Ethernet. Due to header(link layer, network layer, transport layer), actual allowed application data size may be less than 1500 bytes. I also think that supporting TCP on top of Brunet is the best method. I will graciously help implementing TCP method on Brunet, if you need more hands.

 

Best,

Kyungyong Lee

Jimmy Lipham

unread,
Dec 16, 2009, 11:06:56 AM12/16/09
to aci...@googlegroups.com
I too would love to see TCP handled at the Brunet layer. I have
discussed this with David in the past and feel that it would be great
all around to have this sort of functionality that way we wouldn't be
bound to using 3rd party TCP stacks like those offered in the tun/tap
drivers. Now, in the case of IPOP/SocialVPN, the data would have to use
the stack provided by the VPN adapter, but it would be leaps and bounds
for other non-VPN applications to have the same reliability layer. I
will be willing to test to the best of my abilities when this comes to
fruition!

I look forward to it, too!

Jimmy Lipham
Information Technology Professional
UGA Food Services
Phone: (706) 542-7134

"If at first you don't succeed, so much for skydiving..."



Kyungyong Lee wrote:
>
> OK. It makes sense. 1500 bytes are the maximum allowed packet size at
> the Ethernet. Due to header(link layer, network layer, transport
> layer), actual allowed application data size may be less than 1500
> bytes. I also think that supporting TCP on top of Brunet is the best
> method. I will graciously help implementing TCP method on Brunet, if
> you need more hands.
>
> Best,
>
> Kyungyong Lee
>
> *From:* aci...@googlegroups.com [mailto:aci...@googlegroups.com] *On
> Behalf Of *P. Oscar Boykin
> *Sent:* Wednesday, December 16, 2009 10:10 AM
> *To:* aci...@googlegroups.com
> *Subject:* Re: [P2P] Update
>
> In a real system using UDP, packets larger than 1500 bytes may not be
> transmitted (due to poor support for fragmentation).
>
> TCP doesn't have this problem.
>
> We could do fragmentation at the Brunet layer, or we would fragment at
> the dht layer. This problem needs to be solved eventually, but there
> has not yet been a clear winner in terms of how to solve it.
>
> David has started a tcp implementation that could run on top of
> brunet. That might be the best way to go because a tcp implementation
> could allow us to do the dht on tcp on brunet, and any other
> application could use it too.
>
> Best
>
> On Dec 15, 2009 11:29 PM, "Kyungyong Lee" <iamkyu...@gmail.com
> <mailto:iamkyu...@gmail.com>> wrote:
>
> Last week, I mainly worked on simulating Sword on top of Brunet
> while debugging Sword simulator. While doing simulation, I got dht
> value size problem. At current Dht implementation, it seems that
> an entry�s value size is limited to 1024 byte.
>
> http://github.com/johnynek/brunet/blob/master/src/dht/TableServer.cs#L64
>
> Sword saves a node�s all attributes� information at dht. Assuming
> an average attribute size is 20 bytes, and 100 attributes exist in
> a node. The total attribute size would be 2000 bytes, which
> exceeds the dht value size restriction. I manually increased the
> dht value maximum size as needed to solve the issue. What problem
> can occur if we do not limit the maximum dht entry value size?
>
> Best,
>
> Kyungyong Lee
>
> --
>
> You received this message because you are subscribed to the Google
> Groups "acis.p2p" group.
> To post to this group, send email to aci...@googlegroups.com
> <mailto:aci...@googlegroups.com>.
> To unsubscribe from this group, send email to
> acisp2p+u...@googlegroups.com
> <mailto:acisp2p%2Bunsu...@googlegroups.com>.

David Isaac Wolinsky

unread,
Dec 16, 2009, 11:14:06 AM12/16/09
to aci...@googlegroups.com
Working on IPTPS submission.

XU, Jiangyan

unread,
Dec 16, 2009, 11:36:04 AM12/16/09
to aci...@googlegroups.com
Hi

I have done some experiments with ec2 BitTorrent last week. I wasn't
able to gather all the experiment data earlier so... sorry about the
last minute update.

More log parsing and analysis scripts are needed but here are some numbers:

In a scenario where we have 1 data publisher and 10 downloaders; File
size: 25MB; Bandwidth limit on publisher: 1Mbps both upload and
download. I compared the file transfer time of two mechanisms:

NFS:
min/avg/max = 916.243/2227.9469/3212.739 seconds

BitTorrent:
min/avg/max = 137.13/141.94/145.2 seconds

Best,
Jiangyan

Renato Figueiredo

unread,
Dec 16, 2009, 11:36:55 AM12/16/09
to acisp2p
That's good, Pierre, unfortunately I won't be able to make it to the meeting, but you can fill me in sometime later.
Dr. Renato J. Figueiredo
Associate Professor
ACIS Lab - ECE - University of Florida
UF Site Director, Center for Autonomic Computing
http://byron.acis.ufl.edu
ph: 352-392-6430

Pierre St Juste

unread,
Dec 16, 2009, 1:52:02 PM12/16/09
to aci...@googlegroups.com
Attached is the pdf comparing MSP2P with Brunet.
msnp2p_brunet_comp.pdf

JustDoIt Ly

unread,
Dec 17, 2009, 8:31:12 AM12/17/09
to acis.p2p
As far as I know, MS-P2P is available to Windows XP also if you have a
patch installed. And it was available since 2006.

You may check here, http://technet.microsoft.com/en-us/library/bb457079.aspx
.

So, what's your conclusion and suggestions?

On Dec 17, 2:52 am, Pierre St Juste <pton...@gmail.com> wrote:
> Attached is the pdf comparing MSP2P with Brunet.
>
> On Wed, Dec 16, 2009 at 11:36 AM, Renato Figueiredo <rjo...@gmail.com>wrote:
>
>
>
>
>
> > That's good, Pierre, unfortunately I won't be able to make it to the
> > meeting, but you can fill me in sometime later.
>

> > On Wed, Dec 16, 2009 at 12:19 AM, Pierre St Juste <pton...@gmail.com>wrote:
>
> >> I plan on doing a quick summery of Brunet and MS P2P Framework at the
> >> meeting.
>

> >> On Tue, Dec 15, 2009 at 11:29 PM, Kyungyong Lee <iamkyungy...@gmail.com>wrote:
>
> >>>  Last week, I mainly worked on simulating Sword on top of Brunet while
> >>> debugging Sword simulator. While doing simulation, I got dht value size
> >>> problem. At current Dht implementation, it seems that an entry’s value size
> >>> is limited to 1024 byte.
>
> >>>http://github.com/johnynek/brunet/blob/master/src/dht/TableServer.cs#L64
>
> >>> Sword saves a node’s all attributes’ information at dht. Assuming an
> >>> average attribute size is 20 bytes, and 100 attributes exist in a node.  The
> >>> total attribute size would be 2000 bytes, which exceeds the dht value size
> >>> restriction. I manually increased the dht value maximum size as needed to
> >>> solve the issue. What problem can occur if we do not limit the maximum dht
> >>> entry value size?
>
> >>> Best,
>
> >>> Kyungyong Lee
>
> >>> --
> >>> You received this message because you are subscribed to the Google Groups
> >>> "acis.p2p" group.
> >>> To post to this group, send email to aci...@googlegroups.com.
> >>> To unsubscribe from this group, send email to

> >>> acisp2p+u...@googlegroups.com<acisp2p%2Bunsu...@googlegroups.com >


> >>> .
> >>> For more options, visit this group at
> >>>http://groups.google.com/group/acisp2p?hl=en.
>
> >> --
> >> Pierre St Juste
>
> >>  --
> >> You received this message because you are subscribed to the Google Groups
> >> "acis.p2p" group.
> >> To post to this group, send email to aci...@googlegroups.com.
> >> To unsubscribe from this group, send email to

> >> acisp2p+u...@googlegroups.com<acisp2p%2Bunsu...@googlegroups.com >


> >> .
> >> For more options, visit this group at
> >>http://groups.google.com/group/acisp2p?hl=en.
>
> > --
> > Dr. Renato J. Figueiredo
> > Associate Professor
> > ACIS Lab - ECE - University of Florida
> > UF Site Director, Center for Autonomic Computing
> >http://byron.acis.ufl.edu
> > ph: 352-392-6430
>
> >  --
> > You received this message because you are subscribed to the Google Groups
> > "acis.p2p" group.
> > To post to this group, send email to aci...@googlegroups.com.
> > To unsubscribe from this group, send email to

> > acisp2p+u...@googlegroups.com<acisp2p%2Bunsu...@googlegroups.com >


> > .
> > For more options, visit this group at
> >http://groups.google.com/group/acisp2p?hl=en.
>
> --
> Pierre St Juste
>

>  msnp2p_brunet_comp.pdf
> 120KViewDownload

Pierre St Juste

unread,
Dec 17, 2009, 10:47:04 AM12/17/09
to aci...@googlegroups.com
Conclusion is that MSP2P is designed on IPv6 and (for now) depends on Teredo IPv6 Networking by MS to work. Also, MSP2P was designed as you typical P2P DHT which specializes in providing distributed lookup services in log(N) time. Brunet P2P was designed as a datagram transport framework where you can easily implement any P2P system of your liking (Chord, Bamboo, Viceroy, Pastry and so on). Currently we have a Symphony-based implementation. Basically, you can think of Brunet of the P2P version of the Internet Protocol (IP) and it specializes in routing datagrams of any type.
 
Bottom line: Brunet is a more general purpose framework for P2P research. MSP2P is not so general. Plus Brunet is open-source.

To unsubscribe from this group, send email to acisp2p+u...@googlegroups.com.

For more options, visit this group at http://groups.google.com/group/acisp2p?hl=en.





--
Pierre St Juste

JustDoIt Ly

unread,
Dec 21, 2009, 12:58:31 AM12/21/09
to acis.p2p
I found a page about WCF in Mono here, http://www.mono-project.com/WCF
.

So it is possible to use MS P2P Standard in Mono also.

Just for your information.

> > <acisp2p%2Bunsu...@googlegroups.com<acisp2p%252Bunsubscribe@googlegroup s.com>>


> > > >>> .
> > > >>> For more options, visit this group at
> > > >>>http://groups.google.com/group/acisp2p?hl=en.
>
> > > >> --
> > > >> Pierre St Juste
>
> > > >>  --
> > > >> You received this message because you are subscribed to the Google
> > Groups
> > > >> "acis.p2p" group.
> > > >> To post to this group, send email to aci...@googlegroups.com.
> > > >> To unsubscribe from this group, send email to
> > > >> acisp2p+u...@googlegroups.com<acisp2p%2Bunsu...@googlegroups.com >

> > <acisp2p%2Bunsu...@googlegroups.com<acisp2p%252Bunsubscribe@googlegroup s.com>>


> > > >> .
> > > >> For more options, visit this group at
> > > >>http://groups.google.com/group/acisp2p?hl=en.
>
> > > > --
> > > > Dr. Renato J. Figueiredo
> > > > Associate Professor
> > > > ACIS Lab - ECE - University of Florida
> > > > UF Site Director, Center for Autonomic Computing
> > > >http://byron.acis.ufl.edu
> > > > ph: 352-392-6430
>
> > > >  --
> > > > You received this message because you are subscribed to the Google
> > Groups
> > > > "acis.p2p" group.
> > > > To post to this group, send email to aci...@googlegroups.com.
> > > > To unsubscribe from this group, send email to
> > > > acisp2p+u...@googlegroups.com<acisp2p%2Bunsu...@googlegroups.com >

> > <acisp2p%2Bunsu...@googlegroups.com<acisp2p%252Bunsubscribe@googlegroup s.com>>


> > > > .
> > > > For more options, visit this group at
> > > >http://groups.google.com/group/acisp2p?hl=en.
>
> > > --
> > > Pierre St Juste
>
> > >  msnp2p_brunet_comp.pdf
> > > 120KViewDownload
>
> > --
>
> > You received this message because you are subscribed to the Google Groups
> > "acis.p2p" group.
> > To post to this group, send email to aci...@googlegroups.com.
> > To unsubscribe from this group, send email to

Reply all
Reply to author
Forward
0 new messages