global filesystem?

There is a problem the users were warned about dealing with
sharing files on windows boxes over a T1 between North
America and Europe. Now I'm trying to find possible solutions
on how to fix this problem. One idea I think I've heard of is
a filesystem that can be shared, like AFS, that exchanges or
replicates only the changes to a file, like JFS logs changes
to the filesystem. Such that only the changes are replicated
over the T1 and not the full file as is being done now.

Anyone know about this or have some ideas?

Mike

In article ,
Mike writes:
| There is a problem the users were warned about dealing with
| sharing files on windows boxes over a T1 between North
| America and Europe. Now I'm trying to find possible solutions
| on how to fix this problem. One idea I think I've heard of is
| a filesystem that can be shared, like AFS, that exchanges or
| replicates only the changes to a file, like JFS logs changes
| to the filesystem. Such that only the changes are replicated
| over the T1 and not the full file as is being done now.
|
| Anyone know about this or have some ideas?
|
| Mike

You can do volume replication to accomplish this. It can be
done at the host level (like Veritas VVR), at the network
level (like FalconStor or IBM SVC), or at the storage array
level (like EMC SDRC). All of these just move the changes over
the wire.

Keith Michaels wrote:
In article ,
Mike writes:
There is a problem the users were warned about dealing with
sharing files on windows boxes over a T1 between North
America and Europe. Now I'm trying to find possible solutions
on how to fix this problem. One idea I think I've heard of is
a filesystem that can be shared, like AFS, that exchanges or
replicates only the changes to a file, like JFS logs changes
to the filesystem. Such that only the changes are replicated
over the T1 and not the full file as is being done now.

Anyone know about this or have some ideas?

Mike

You can do volume replication to accomplish this. It can be
done at the host level (like Veritas VVR), at the network
level (like FalconStor or IBM SVC), or at the storage array
level (like EMC SDRC). All of these just move the changes over
the wire.

Yes, but unlike the the shared filesystem they don't allow simultaneous
access to the data, i.e. two people can't be working on the same file (or
even filesystem) at the different locations.


--
Nik Simpson

On Fri, 25 Jun 2004 18:04:44 -0400, "Nik Simpson"
wrote:

Keith Michaels wrote:
In article ,
Mike writes:
There is a problem the users were warned about dealing with
sharing files on windows boxes over a T1 between North
America and Europe. Now I'm trying to find possible solutions
on how to fix this problem. One idea I think I've heard of is
a filesystem that can be shared, like AFS, that exchanges or
replicates only the changes to a file, like JFS logs changes
to the filesystem. Such that only the changes are replicated
over the T1 and not the full file as is being done now.

Anyone know about this or have some ideas?

Mike

You can do volume replication to accomplish this. It can be
done at the host level (like Veritas VVR), at the network
level (like FalconStor or IBM SVC), or at the storage array
level (like EMC SDRC). All of these just move the changes over
the wire.

Yes, but unlike the the shared filesystem they don't allow simultaneous
access to the data, i.e. two people can't be working on the same file (or
even filesystem) at the different locations.

True enough. But I don't think anyone is really expecting to be able
to do simulataneous data access across continents. Hell, across NA is
too much still and have any sort of performance.

AFS is great for what it does, which is proved read-only replicas
across any geographic spread. But as mentioned several other
applications will do this also. I happen to think AFS is better in
most cases but hey, that's just me.

No matter what though, the latency will be a killer. So if you have
any real churn rate on the data you will still be hurting with a T1.

~F

Faeandar wrote:
On Fri, 25 Jun 2004 18:04:44 -0400, "Nik Simpson"
wrote:

Keith Michaels wrote:
In article ,
Mike writes:
There is a problem the users were warned about dealing with
sharing files on windows boxes over a T1 between North
America and Europe. Now I'm trying to find possible solutions
on how to fix this problem. One idea I think I've heard of is
a filesystem that can be shared, like AFS, that exchanges or
replicates only the changes to a file, like JFS logs changes
to the filesystem. Such that only the changes are replicated
over the T1 and not the full file as is being done now.

Anyone know about this or have some ideas?

Mike

You can do volume replication to accomplish this. It can be
done at the host level (like Veritas VVR), at the network
level (like FalconStor or IBM SVC), or at the storage array
level (like EMC SDRC). All of these just move the changes over
the wire.

Yes, but unlike the the shared filesystem they don't allow
simultaneous access to the data, i.e. two people can't be working on
the same file (or even filesystem) at the different locations.

True enough. But I don't think anyone is really expecting to be able
to do simulataneous data access across continents. Hell, across NA is
too much still and have any sort of performance.

I think we need the original poster to clarify that requirement, otherwise
its just conjecture.



No matter what though, the latency will be a killer. So if you have
any real churn rate on the data you will still be hurting with a T1.

Agreed, T1 is pretty wimpy from a bandwidth perspective these days, heck, my
home connection is faster.

--
Nik Simpson

Yes, but unlike the the shared filesystem they don't allow
simultaneous access to the data, i.e. two people can't be working on
the same file (or even filesystem) at the different locations.

True enough. But I don't think anyone is really expecting to be able
to do simulataneous data access across continents. Hell, across NA is
too much still and have any sort of performance.

I think we need the original poster to clarify that requirement, otherwise
its just conjecture.

I think if the OP is really looking for that good luck. Let me know
if you find it. Hell I can't even hope to get my block changes from
NA to India in 24 hours, forget about simultaneous access.



No matter what though, the latency will be a killer. So if you have
any real churn rate on the data you will still be hurting with a T1.

Agreed, T1 is pretty wimpy from a bandwidth perspective these days, heck, my
home connection is faster.

The bandwidth is less of an issue because my guess is these apps are
all single stream, so wider pip won't make a diff. But even if it's
mult-stream capable, the latency is still huge.

~F

In article , Faeandar wrote:


Yes, but unlike the the shared filesystem they don't allow
simultaneous access to the data, i.e. two people can't be working on
the same file (or even filesystem) at the different locations.

True enough. But I don't think anyone is really expecting to be able
to do simulataneous data access across continents. Hell, across NA is
too much still and have any sort of performance.

I think we need the original poster to clarify that requirement, otherwise
its just conjecture.

I think if the OP is really looking for that good luck. Let me know
if you find it. Hell I can't even hope to get my block changes from
NA to India in 24 hours, forget about simultaneous access.

There is an issue where one group of people from Europe using Citrix
are complaining of slowness when they access Windows servers here in
North America. The windows admin mentioned the slowness is probably
due to the replication of the windows user profiles.

I am researching what alternatives we have. One option I thought of is
a global, replicated filesystem such that Europe replicates with a
server in Europe, the files get replicated between the European server
and the NA server, then the local windows servers replicate with the
NA server.

I'm not looking for simultaneous(?) access and edits to a file. Just
some way to reduce the preceived(?) slowness of logging into the servers
and using the applications.

No matter what though, the latency will be a killer. So if you have
any real churn rate on the data you will still be hurting with a T1.

Agreed, T1 is pretty wimpy from a bandwidth perspective these days, heck, my
home connection is faster.

The bandwidth is less of an issue because my guess is these apps are
all single stream, so wider pip won't make a diff. But even if it's
mult-stream capable, the latency is still huge.

Not my choice on the bandwidth between NA and Europe. I just have to
make it work.

Mike

Mike wrote:
In article , Faeandar
wrote:


Yes, but unlike the the shared filesystem they don't allow
simultaneous access to the data, i.e. two people can't be working
on the same file (or even filesystem) at the different locations.

True enough. But I don't think anyone is really expecting to be
able to do simulataneous data access across continents. Hell,
across NA is too much still and have any sort of performance.

I think we need the original poster to clarify that requirement,
otherwise its just conjecture.

I think if the OP is really looking for that good luck. Let me know
if you find it. Hell I can't even hope to get my block changes from
NA to India in 24 hours, forget about simultaneous access.

There is an issue where one group of people from Europe using Citrix
are complaining of slowness when they access Windows servers here in
North America. The windows admin mentioned the slowness is probably
due to the replication of the windows user profiles.


Sounds more like a problem that should be discussed with Citrix, if the
users in Europe are logging into Citrix servers in the US, then I'm not sure
why the user profiles would be replicated to Europe since the "users" are
all in the US (either locally or remotely through Citrix)


--
Nik Simpson

"Faeandar" wrote in message
...
On Fri, 25 Jun 2004 18:04:44 -0400, "Nik Simpson"
wrote:

Keith Michaels wrote:
In article ,
Mike writes:
There is a problem the users were warned about dealing with
sharing files on windows boxes over a T1 between North
America and Europe. Now I'm trying to find possible solutions
on how to fix this problem. One idea I think I've heard of is
a filesystem that can be shared, like AFS, that exchanges or
replicates only the changes to a file, like JFS logs changes
to the filesystem. Such that only the changes are replicated
over the T1 and not the full file as is being done now.

Anyone know about this or have some ideas?

Mike

You can do volume replication to accomplish this. It can be
done at the host level (like Veritas VVR), at the network
level (like FalconStor or IBM SVC), or at the storage array
level (like EMC SDRC). All of these just move the changes over
the wire.

Yes, but unlike the the shared filesystem they don't allow simultaneous
access to the data, i.e. two people can't be working on the same file (or
even filesystem) at the different locations.

True enough. But I don't think anyone is really expecting to be able
to do simulataneous data access across continents. Hell, across NA is
too much still and have any sort of performance.

Well, that *does* depend on the file system you're using to some extent.

With a flexible cluster file system like VMS's, it's at least possible to
work cross-continent with only occasional latency problems (though you'll
always see added latency on file writes as the remote copy gets updated, if
you demand synchronous-level consistency and that mirror copies be widely
dispersed). Locking for each individual file occurs where the most activity
on that file exists, so as long as people in widely-separated locations are
accessing different files the read performance they perceive is that of
local access, even if their access allows write-sharing.

- bill