Backups

I need to copy about a terabyte of data a day from my datacenter to our DR
site. Does anyone know of a good solution for this? I am copying CIFS acroos
a 3mbps WAN link. I am running windows 2003 servers using a BlueARC SAN

 

Suppose you can use maximum bandwidth and don't do anything else but backup,
you can use simple math:
3 Mbps = 300 kBps
300 kBps * 3600 seconds (1 hour) = 1.08 GBph
1.08 GBph * 24 hours (1 day) = 25.92 GB per day
1000 GB (1TB) / 25.92 = 38.5802 days

This math does not take into account compression. Even at the best rate of
compression (let's say 50%), you would need 19 days.

I would suggest you start thinking about two stage backup.
1st stage disk based.
2nd stage tape based.
With fast arrays/SANS and link speeds 2GBps and above.

You've got steps above, just substitute the numbers.

Good luck


"Ed" <Ed@discussions.microsoft.com> wrote in message
news:F9D81F7A-78B4-447E-9C88-72A4401C29A1@microsoft.com...<!--coloro:blue--><span style="color:blue <!--/coloro-->
>I need to copy about a terabyte of data a day from my datacenter to our DR
> site. Does anyone know of a good solution for this? I am copying CIFS
> acroos
> a 3mbps WAN link. I am running windows 2003 servers using a BlueARC SAN <!--colorc--><!--/colorc-->

 

On Sep 21, 5:44 pm, Ed <E...@discussions.microsoft.com> wrote:<!--coloro:blue--><span style="color:blue <!--/coloro-->
> I need to copy about a terabyte of data a day from my datacenter to our DR
> site. Does anyone know of a good solution for this? I am copying CIFS acroos
> a 3mbps WAN link. I am running windows 2003 servers using a BlueARC SAN<!--colorc--><!--/colorc-->

Because, as Dusko pointed out in his post, you can't get a terabyte
down a 3mbps link, you need to look at what your data differential is
per day.

You want to only send the change in data, rather than all the data.

So look for some kind of a disk block method, where only changed
blocks are sent. This takes very little bandwidth in comparison,
especially if done in psudo-real-time, meaning throughout the day and
night, changed blocks are immediately transmitted to the mirrored
(remote) end.

Microsoft's Data Protection Manager is supposed to do this, and
Microsoft hired away one of Double-Take Software's VPs, as I read
somewhere about a year ago. But Microsoft's product only targets
their own data types, so if your needs go beyond MS networks and apps,
you need to look further.

Double-Take is still in business and offers an excellent stand-alone
product you can buy and configure yourself - and it's multi and
possibly cross platform, meaning it has ports for Linux as well as
Windows.

Double-Take also performs byte-change transfer as well as block-
change. This allows it to keep small files completely up-to-date at
both ends of a mirror.

I've use Double-Take at one site. It's an expensive (all this
software is right now) and a brilliant product that duplicates data
between two (or more) sites (at around $2,500 per site). With Windows
Server 2003 it's rock solid, and keeps gigs of data fully synchronized
within minutes. It can sync both ways, or one way.

While $5,000 for two endpoints from Double-Take seems high,
Microsoft's DPM is less to start, but climbs right up there too.
Anybody that can do this well is going to charge a lot for it, because
frankly, there's no other way as you can see by the math, to pull this
kind of thing off. It's difficult enough just to get 1TB of data
backed up on a LAN in an 8-hour window.

Recently, one of my Dell servers crashed, but the replicated data on
the mirror was used to restore it. This mirror was remotely located,
using a DSL-to-Cable modem based VPN capable of no more than 1.5 Mbps
transfer rate.

There's a lot of players in this field now, including big names like
Cisco who are targeting more than just data backup. Specifically,
they're applying data de-duplication, block and byte differential
analysis and application specific traffic demand analysis on MS SQL,
Oracle, DB2, MS Exchange, CIFS (MS File Systems) and many other apps
and file structures to accelerate data availability over remote or
slow networks.

Just google "ip based data duplication remote mirror sync" and other
stuff like "double-take" and go from there, and get educated. There's
a lot of solutions and an awful lot to learn about what's good, and
what's not. Be weary of supposed "White Papers" as they are often not
independent and are very, very biased toward their own products.

'gluck

 

is it possible to do compression at the server level- with third party
software than do compression across the WAN with a hardware appliance like a
Cisco WAAS

"Dusko Savatovic" wrote:
<!--coloro:blue--><span style="color:blue <!--/coloro-->
> Suppose you can use maximum bandwidth and don't do anything else but backup,
> you can use simple math:
> 3 Mbps = 300 kBps
> 300 kBps * 3600 seconds (1 hour) = 1.08 GBph
> 1.08 GBph * 24 hours (1 day) = 25.92 GB per day
> 1000 GB (1TB) / 25.92 = 38.5802 days
>
> This math does not take into account compression. Even at the best rate of
> compression (let's say 50%), you would need 19 days.
>
> I would suggest you start thinking about two stage backup.
> 1st stage disk based.
> 2nd stage tape based.
> With fast arrays/SANS and link speeds 2GBps and above.
>
> You've got steps above, just substitute the numbers.
>
> Good luck
>
>
> "Ed" <Ed@discussions.microsoft.com> wrote in message
> news:F9D81F7A-78B4-447E-9C88-72A4401C29A1@microsoft.com...<!--coloro:green--><span style="color:green <!--/coloro-->
> >I need to copy about a terabyte of data a day from my datacenter to our DR
> > site. Does anyone know of a good solution for this? I am copying CIFS
> > acroos
> > a 3mbps WAN link. I am running windows 2003 servers using a BlueARC SAN <!--colorc--><!--/colorc-->
>
>
> <!--colorc--><!--/colorc-->

 

is it possible to do compression at the server level- with third party
software than do compression across the WAN with a hardware appliance like a
Cisco WAAS

"Dull Nouvelle" wrote:
<!--coloro:blue--><span style="color:blue <!--/coloro-->
> On Sep 21, 5:44 pm, Ed <E...@discussions.microsoft.com> wrote:<!--coloro:green--><span style="color:green <!--/coloro-->
> > I need to copy about a terabyte of data a day from my datacenter to our DR
> > site. Does anyone know of a good solution for this? I am copying CIFS acroos
> > a 3mbps WAN link. I am running windows 2003 servers using a BlueARC SAN<!--colorc--><!--/colorc-->
>
> Because, as Dusko pointed out in his post, you can't get a terabyte
> down a 3mbps link, you need to look at what your data differential is
> per day.
>
> You want to only send the change in data, rather than all the data.
>
> So look for some kind of a disk block method, where only changed
> blocks are sent. This takes very little bandwidth in comparison,
> especially if done in psudo-real-time, meaning throughout the day and
> night, changed blocks are immediately transmitted to the mirrored
> (remote) end.
>
> Microsoft's Data Protection Manager is supposed to do this, and
> Microsoft hired away one of Double-Take Software's VPs, as I read
> somewhere about a year ago. But Microsoft's product only targets
> their own data types, so if your needs go beyond MS networks and apps,
> you need to look further.
>
> Double-Take is still in business and offers an excellent stand-alone
> product you can buy and configure yourself - and it's multi and
> possibly cross platform, meaning it has ports for Linux as well as
> Windows.
>
> Double-Take also performs byte-change transfer as well as block-
> change. This allows it to keep small files completely up-to-date at
> both ends of a mirror.
>
> I've use Double-Take at one site. It's an expensive (all this
> software is right now) and a brilliant product that duplicates data
> between two (or more) sites (at around $2,500 per site). With Windows
> Server 2003 it's rock solid, and keeps gigs of data fully synchronized
> within minutes. It can sync both ways, or one way.
>
> While $5,000 for two endpoints from Double-Take seems high,
> Microsoft's DPM is less to start, but climbs right up there too.
> Anybody that can do this well is going to charge a lot for it, because
> frankly, there's no other way as you can see by the math, to pull this
> kind of thing off. It's difficult enough just to get 1TB of data
> backed up on a LAN in an 8-hour window.
>
> Recently, one of my Dell servers crashed, but the replicated data on
> the mirror was used to restore it. This mirror was remotely located,
> using a DSL-to-Cable modem based VPN capable of no more than 1.5 Mbps
> transfer rate.
>
> There's a lot of players in this field now, including big names like
> Cisco who are targeting more than just data backup. Specifically,
> they're applying data de-duplication, block and byte differential
> analysis and application specific traffic demand analysis on MS SQL,
> Oracle, DB2, MS Exchange, CIFS (MS File Systems) and many other apps
> and file structures to accelerate data availability over remote or
> slow networks.
>
> Just google "ip based data duplication remote mirror sync" and other
> stuff like "double-take" and go from there, and get educated. There's
> a lot of solutions and an awful lot to learn about what's good, and
> what's not. Be weary of supposed "White Papers" as they are often not
> independent and are very, very biased toward their own products.
>
> 'gluck
> <!--colorc--><!--/colorc-->

 

Yes, both are possible, but your initial copy, or backup, will take the same
amount of time that Dusko pointed out. There are several third party
products. Cisco WAAS will cache CIFS and compress it and I hear it does a
good job. I don't know that it fits your situation unless they've changed
the way CIFS is handled with it. Still in the same boat though, initial copy
of data will take forever.

You need to be thinking, instead of compression for backup across WAN, of
transmitting block level changes and synching the data. That will be much
less data going across your WAN. As Dull points out MS DPM will do the trick
as will several third party products.

If your Windows 2003 Servers are R2, you can take advantage of the new DFS.
I've used it very successfully and have been happy with the results.
Although I did not have 1TB of data to replicate, I did have several GBs of
data.

So in short, if you get a hard copy via tape, DVDs, hard drives or whatever
media to your DR site, you can then use a MS product, or a third party
product that will synch the data through block changes in data rather than
the complete files.


hth,
Brooke


"Ed" <Ed@discussions.microsoft.com> wrote in message
news:2A73E45F-E3E9-41C7-A77E-DA2723BA3197@microsoft.com...<!--coloro:blue--><span style="color:blue <!--/coloro-->
> is it possible to do compression at the server level- with third party
> software than do compression across the WAN with a hardware appliance like
> a
> Cisco WAAS
>
> "Dusko Savatovic" wrote:
><!--coloro:green--><span style="color:green <!--/coloro-->
>> Suppose you can use maximum bandwidth and don't do anything else but
>> backup,
>> you can use simple math:
>> 3 Mbps = 300 kBps
>> 300 kBps * 3600 seconds (1 hour) = 1.08 GBph
>> 1.08 GBph * 24 hours (1 day) = 25.92 GB per day
>> 1000 GB (1TB) / 25.92 = 38.5802 days
>>
>> This math does not take into account compression. Even at the best rate
>> of
>> compression (let's say 50%), you would need 19 days.
>>
>> I would suggest you start thinking about two stage backup.
>> 1st stage disk based.
>> 2nd stage tape based.
>> With fast arrays/SANS and link speeds 2GBps and above.
>>
>> You've got steps above, just substitute the numbers.
>>
>> Good luck
>>
>>
>> "Ed" <Ed@discussions.microsoft.com> wrote in message
>> news:F9D81F7A-78B4-447E-9C88-72A4401C29A1@microsoft.com...<!--coloro:darkred--><span style="color:darkred <!--/coloro-->
>> >I need to copy about a terabyte of data a day from my datacenter to our
>> >DR
>> > site. Does anyone know of a good solution for this? I am copying CIFS
>> > acroos
>> > a 3mbps WAN link. I am running windows 2003 servers using a BlueARC SAN<!--colorc--><!--/colorc-->
>>
>>
>> <!--colorc--><!--/colorc--><!--colorc--><!--/colorc-->