Mailing List Archive

>>>>> "Sami" == Sami Kapanen <sami.kapanen@hamk.fi> writes:

Sami> Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
Sami> Considering these, looking for experiences,

In my experience with backups, no one cares. It's only about
restores. And being able to quickly and easily search through backups
for the file(s) to restore. So test that side of things first.

In personal experience, Networker (EMC, was Legato) was awesome for
doing restores, the file indexes stayed online and made searching
trivial.

CommVault... sucks for this. The indexes get purged so damn fast that
you're always pulling back tapes just for the indexes, etc. Painful.

So test this part of your backup scheme first, or find the user forums
for Cohesity/Rubrik and ask that question there. And of course try to
ask people who have the same size environment as you want to backup as
well.

Cheers,
John
_______________________________________________
Toasters mailing list
Toasters@teaparty.net
https://www.teaparty.net/mailman/listinfo/toasters

RE: SnapDiff API, be sure to ask each vendor about that. It's important, because it shows where the tighter integration is.

You're likely to get very different answers about their roadmaps.

-----Original Message-----
From: Michael Bergman <michael.bergman@ericsson.com>
Sent: Tuesday, October 27, 2020 2:54 PM
To: Toasters@teaparty.net
Subject: Re: Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?

On 2020-10-27 19:28, John Stoffel wrote:
> Sami> Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
> Sami> Considering these, looking for experiences,
>
> In my experience with backups, no one cares. It's only about
> restores. And being able to quickly and easily search through backups
> for the file(s) to restore. So test that side of things first.

Exactly. Since years, I always write this "Restore/Backup". *Not* Backup/Restore.
That's not to say that RPO (aka ADL, Acceptable Data Loss) is unimportant.
But the end users, the data owners, do not care one bit about *how* that works. And until the need a restore, they couldn't care less so you can do whatever you want including not taking the backups except THAT ONE some user wants a restore from ;-)

Sorry. Couldn't resist.

The backup taking problem w Rubrik is that it's host based. SMB ("NAS") based backups taken by a Rubrik server, will traverse the file tress, scan them over and over and over again and then pull the data out (night time
perhaps...) via the front end traffic intrefaces the same as where your user workload is all the time.
In a Hi File Count environment (HFC) with billions of files and lots of "churn" (WRITE-DEL-WRITE-DEL-WRITE-DEL,...) this is not going to work well for you. The backups will just never finish; the backup window problem revisited. Back to the 1990's.

So with ONTAP gear you have the SnapDiff API. And then it can work quite well. No problem -- should be OK even in a HFC environment

John Stoffel wrote:
> In personal experience, Networker (EMC, was Legato) was awesome for
> doing restores, the file indexes stayed online and made searching
> trivial.

I concur. Like John says: was Legato. It was good in this respect.
I have no historical hands on experience w CommVault but what I know about it in theory has made me/us stay away from it. Every assessment, w a few years in between has made me go "... no."

My options expressed here of course.

If you follow Johns good advice here, not that much can go wrong for you IMHO.

/M

John Stoffel wrote:
> CommVault... sucks for this. The indexes get purged so damn fast that
> you're always pulling back tapes just for the indexes, etc. Painful.
>
> So test this part of your backup scheme first, or find the user forums
> for Cohesity/Rubrik and ask that question there. And of course try to
> ask people who have the same size environment as you want to backup as
> well.
>
> Cheers,
> John

_______________________________________________
Toasters mailing list
Toasters@teaparty.net
https://www.teaparty.net/mailman/listinfo/toasters

I will check these,

Rubrik states it is using SnapDiff. Cohesity says it is based on NAS snapshots as well.

Restore times are important, but not crucial to us (University), we don't loose millions per day if/when restoring.

sk


> -----Original Message-----
> From: Toasters <toasters-bounces@teaparty.net> On Behalf Of Parisi, Justin
> Sent: tiistai 27. lokakuuta 2020 21:04
> To: NGC-michael.bergman-ericsson.com <michael.bergman@ericsson.com>;
> Toasters@teaparty.net
> Subject: RE: Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
>
> RE: SnapDiff API, be sure to ask each vendor about that. It's important, because
> it shows where the tighter integration is.
>
> You're likely to get very different answers about their roadmaps.
>
> -----Original Message-----
> From: Michael Bergman <michael.bergman@ericsson.com>
> Sent: Tuesday, October 27, 2020 2:54 PM
> To: Toasters@teaparty.net
> Subject: Re: Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
>
> On 2020-10-27 19:28, John Stoffel wrote:
> > Sami> Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
> > Sami> Considering these, looking for experiences,
> >
> > In my experience with backups, no one cares. It's only about
> > restores. And being able to quickly and easily search through backups
> > for the file(s) to restore. So test that side of things first.
>
> Exactly. Since years, I always write this "Restore/Backup". *Not*
> Backup/Restore.
> That's not to say that RPO (aka ADL, Acceptable Data Loss) is unimportant.
> But the end users, the data owners, do not care one bit about *how* that
> works. And until the need a restore, they couldn't care less so you can do
> whatever you want including not taking the backups except THAT ONE some
> user wants a restore from ;-)
>
> Sorry. Couldn't resist.
>
> The backup taking problem w Rubrik is that it's host based. SMB ("NAS") based
> backups taken by a Rubrik server, will traverse the file tress, scan them over and
> over and over again and then pull the data out (night time
> perhaps...) via the front end traffic intrefaces the same as where your user
> workload is all the time.
> In a Hi File Count environment (HFC) with billions of files and lots of "churn"
> (WRITE-DEL-WRITE-DEL-WRITE-DEL,...) this is not going to work well for you.
> The backups will just never finish; the backup window problem revisited. Back to
> the 1990's.
>
> So with ONTAP gear you have the SnapDiff API. And then it can work quite well.
> No problem -- should be OK even in a HFC environment
>
> John Stoffel wrote:
> > In personal experience, Networker (EMC, was Legato) was awesome for
> > doing restores, the file indexes stayed online and made searching
> > trivial.
>
> I concur. Like John says: was Legato. It was good in this respect.
> I have no historical hands on experience w CommVault but what I know about it
> in theory has made me/us stay away from it. Every assessment, w a few years in
> between has made me go "... no."
>
> My options expressed here of course.
>
> If you follow Johns good advice here, not that much can go wrong for you
> IMHO.
>
> /M
>
> John Stoffel wrote:
> > CommVault... sucks for this. The indexes get purged so damn fast that
> > you're always pulling back tapes just for the indexes, etc. Painful.
> >
> > So test this part of your backup scheme first, or find the user forums
> > for Cohesity/Rubrik and ask that question there. And of course try to
> > ask people who have the same size environment as you want to backup as
> > well.
> >
> > Cheers,
> > John
>
> _______________________________________________
> Toasters mailing list
> Toasters@teaparty.net
> https://www.teaparty.net/mailman/listinfo/toasters

_______________________________________________
Toasters mailing list
Toasters@teaparty.net
https://www.teaparty.net/mailman/listinfo/toasters

Veeam is adding SnapDiff in v11 for NAS backup too, it's all about the
restore capabilities and Veeam is fully focused on all the (granular)
recovery options.

Op wo 28 okt. 2020 om 07:33 schreef Sami Kapanen <sami.kapanen@hamk.fi>:

> I will check these,
>
> Rubrik states it is using SnapDiff. Cohesity says it is based on NAS
> snapshots as well.
>
> Restore times are important, but not crucial to us (University), we don't
> loose millions per day if/when restoring.
>
> sk
>
>
> > -----Original Message-----
> > From: Toasters <toasters-bounces@teaparty.net> On Behalf Of Parisi,
> Justin
> > Sent: tiistai 27. lokakuuta 2020 21:04
> > To: NGC-michael.bergman-ericsson.com <michael.bergman@ericsson.com>;
> > Toasters@teaparty.net
> > Subject: RE: Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
> >
> > RE: SnapDiff API, be sure to ask each vendor about that. It's important,
> because
> > it shows where the tighter integration is.
> >
> > You're likely to get very different answers about their roadmaps.
> >
> > -----Original Message-----
> > From: Michael Bergman <michael.bergman@ericsson.com>
> > Sent: Tuesday, October 27, 2020 2:54 PM
> > To: Toasters@teaparty.net
> > Subject: Re: Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
> >
> > On 2020-10-27 19:28, John Stoffel wrote:
> > > Sami> Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
> > > Sami> Considering these, looking for experiences,
> > >
> > > In my experience with backups, no one cares. It's only about
> > > restores. And being able to quickly and easily search through backups
> > > for the file(s) to restore. So test that side of things first.
> >
> > Exactly. Since years, I always write this "Restore/Backup". *Not*
> > Backup/Restore.
> > That's not to say that RPO (aka ADL, Acceptable Data Loss) is
> unimportant.
> > But the end users, the data owners, do not care one bit about *how* that
> > works. And until the need a restore, they couldn't care less so you can
> do
> > whatever you want including not taking the backups except THAT ONE some
> > user wants a restore from ;-)
> >
> > Sorry. Couldn't resist.
> >
> > The backup taking problem w Rubrik is that it's host based. SMB ("NAS")
> based
> > backups taken by a Rubrik server, will traverse the file tress, scan
> them over and
> > over and over again and then pull the data out (night time
> > perhaps...) via the front end traffic intrefaces the same as where your
> user
> > workload is all the time.
> > In a Hi File Count environment (HFC) with billions of files and lots of
> "churn"
> > (WRITE-DEL-WRITE-DEL-WRITE-DEL,...) this is not going to work well for
> you.
> > The backups will just never finish; the backup window problem revisited.
> Back to
> > the 1990's.
> >
> > So with ONTAP gear you have the SnapDiff API. And then it can work quite
> well.
> > No problem -- should be OK even in a HFC environment
> >
> > John Stoffel wrote:
> > > In personal experience, Networker (EMC, was Legato) was awesome for
> > > doing restores, the file indexes stayed online and made searching
> > > trivial.
> >
> > I concur. Like John says: was Legato. It was good in this respect.
> > I have no historical hands on experience w CommVault but what I know
> about it
> > in theory has made me/us stay away from it. Every assessment, w a few
> years in
> > between has made me go "... no."
> >
> > My options expressed here of course.
> >
> > If you follow Johns good advice here, not that much can go wrong for you
> > IMHO.
> >
> > /M
> >
> > John Stoffel wrote:
> > > CommVault... sucks for this. The indexes get purged so damn fast that
> > > you're always pulling back tapes just for the indexes, etc. Painful.
> > >
> > > So test this part of your backup scheme first, or find the user forums
> > > for Cohesity/Rubrik and ask that question there. And of course try to
> > > ask people who have the same size environment as you want to backup as
> > > well.
> > >
> > > Cheers,
> > > John
> >
> > _______________________________________________
> > Toasters mailing list
> > Toasters@teaparty.net
> > https://www.teaparty.net/mailman/listinfo/toasters
>
> _______________________________________________
> Toasters mailing list
> Toasters@teaparty.net
> https://www.teaparty.net/mailman/listinfo/toasters
>

A customer of mine was exploring both Rubrick and Cohesity to backup *only*
VMware.

They said they went with Rubrick after Rubrick explained that they use
SnapDiff and Cohesity does not.

I informed them, they were essentially lied to. Sure Rubrick does use
SnapDiff but only for NAS backups and not for VMware.
VMware backups go through vCenter and are directed to the ESXi hosts
themselves. No direct NetApp involvement.

Yeah...they were not too happy about that one. They probably would have
gone a different route knowing the truth.

Unfortunately, they did not challenge any FUD either way.

--tmac

*Tim McCarthy, **Principal Consultant*

*Proud Member of the #NetAppATeam <https://twitter.com/NetAppATeam>*

*I Blog at TMACsRack <https://tmacsrack.wordpress.com/>*

[image: Cisco and NetApp FlexPod Design Specialist]
<https://www.youracclaim.com/badges/58cf082d-acd8-4529-821a-bb7eb93a296c/public_url>[image:
NetApp Certified Implementation Engineer - SAN Specialist, ONTAP]
<https://www.youracclaim.com/badges/162b629e-b4f1-48af-a8f9-d2a9517ec100/public_url>[image:
NetApp Certified Data Administrator, ONTAP]
<https://www.youracclaim.com/badges/b41a5941-6885-4181-b984-21df36bc27a8/public_url>[image:
NetApp Certified Storage Installation Engineer, ONTAP]
<https://www.youracclaim.com/badges/367c462d-d58b-4cbf-9e8d-a5068b247cd6/public_url>[image:
NetApp Certified Support Engineer]
<https://www.youracclaim.com/badges/053a4154-47c0-41e8-899f-b2dc81b228a3/public_url>[image:
NetApp Accredited Hardware Support Engineer]
<https://www.youracclaim.com/badges/02a6dbc2-e7f4-4779-846c-3d0eabd48798/public_url>[image:
NetApp Accredited MetroCluster IP Implementation Engineer]
<https://www.youracclaim.com/badges/24bde044-b196-4b33-8fe7-603e533f2551/public_url>
[image: Cisco and NetApp FlexPod Implementation and Administration
Specialist]
<https://www.youracclaim.com/badges/53a73b2a-ca83-43b8-895e-3299735dd406/public_url>[image:
NetApp Certified Implementation Engineer - Data Protection Specialist]
<https://www.youracclaim.com/badges/51e81930-cad0-4e1f-b54d-dde7f181516c/public_url>[image:
NetApp Certified Technology Associate]
<https://www.youracclaim.com/badges/bb26a4bf-2449-4998-a374-74235bc6dec3/public_url>[image:
Subject Matter Expert]
<https://www.youracclaim.com/badges/fc9747ad-648b-4472-8557-72e70e2879c4/public_url>[image:
NetApp Certified Support Engineer - ONTAP Specialist]
<https://www.youracclaim.com/badges/c0815721-1751-4cc3-9bae-01d2778a5358/public_url>[image:
NetApp Accredited StorageGRID Implementation Engineer]
<https://www.youracclaim.com/badges/9613df3b-99b8-41ce-a62f-3a7443316b4e/public_url>[image:
NetApp Accredited MetroCluster Installation Engineer]
<https://www.youracclaim.com/badges/52ebdd24-6c9b-40a5-ba75-a60508b5f28b/public_url>


On Wed, Oct 28, 2020 at 5:11 AM Jan van Leuken <janvanleuken@gmail.com>
wrote:

> Veeam is adding SnapDiff in v11 for NAS backup too, it's all about the
> restore capabilities and Veeam is fully focused on all the (granular)
> recovery options.
>
> Op wo 28 okt. 2020 om 07:33 schreef Sami Kapanen <sami.kapanen@hamk.fi>:
>
>> I will check these,
>>
>> Rubrik states it is using SnapDiff. Cohesity says it is based on NAS
>> snapshots as well.
>>
>> Restore times are important, but not crucial to us (University), we don't
>> loose millions per day if/when restoring.
>>
>> sk
>>
>>
>> > -----Original Message-----
>> > From: Toasters <toasters-bounces@teaparty.net> On Behalf Of Parisi,
>> Justin
>> > Sent: tiistai 27. lokakuuta 2020 21:04
>> > To: NGC-michael.bergman-ericsson.com <michael.bergman@ericsson.com>;
>> > Toasters@teaparty.net
>> > Subject: RE: Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
>> >
>> > RE: SnapDiff API, be sure to ask each vendor about that. It's
>> important, because
>> > it shows where the tighter integration is.
>> >
>> > You're likely to get very different answers about their roadmaps.
>> >
>> > -----Original Message-----
>> > From: Michael Bergman <michael.bergman@ericsson.com>
>> > Sent: Tuesday, October 27, 2020 2:54 PM
>> > To: Toasters@teaparty.net
>> > Subject: Re: Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
>> >
>> > On 2020-10-27 19:28, John Stoffel wrote:
>> > > Sami> Anyone using Cohesity or Rubrik for Netapp CIFS NAS backup?
>> > > Sami> Considering these, looking for experiences,
>> > >
>> > > In my experience with backups, no one cares. It's only about
>> > > restores. And being able to quickly and easily search through backups
>> > > for the file(s) to restore. So test that side of things first.
>> >
>> > Exactly. Since years, I always write this "Restore/Backup". *Not*
>> > Backup/Restore.
>> > That's not to say that RPO (aka ADL, Acceptable Data Loss) is
>> unimportant.
>> > But the end users, the data owners, do not care one bit about *how* that
>> > works. And until the need a restore, they couldn't care less so you can
>> do
>> > whatever you want including not taking the backups except THAT ONE some
>> > user wants a restore from ;-)
>> >
>> > Sorry. Couldn't resist.
>> >
>> > The backup taking problem w Rubrik is that it's host based. SMB ("NAS")
>> based
>> > backups taken by a Rubrik server, will traverse the file tress, scan
>> them over and
>> > over and over again and then pull the data out (night time
>> > perhaps...) via the front end traffic intrefaces the same as where your
>> user
>> > workload is all the time.
>> > In a Hi File Count environment (HFC) with billions of files and lots of
>> "churn"
>> > (WRITE-DEL-WRITE-DEL-WRITE-DEL,...) this is not going to work well for
>> you.
>> > The backups will just never finish; the backup window problem
>> revisited. Back to
>> > the 1990's.
>> >
>> > So with ONTAP gear you have the SnapDiff API. And then it can work
>> quite well.
>> > No problem -- should be OK even in a HFC environment
>> >
>> > John Stoffel wrote:
>> > > In personal experience, Networker (EMC, was Legato) was awesome for
>> > > doing restores, the file indexes stayed online and made searching
>> > > trivial.
>> >
>> > I concur. Like John says: was Legato. It was good in this respect.
>> > I have no historical hands on experience w CommVault but what I know
>> about it
>> > in theory has made me/us stay away from it. Every assessment, w a few
>> years in
>> > between has made me go "... no."
>> >
>> > My options expressed here of course.
>> >
>> > If you follow Johns good advice here, not that much can go wrong for you
>> > IMHO.
>> >
>> > /M
>> >
>> > John Stoffel wrote:
>> > > CommVault... sucks for this. The indexes get purged so damn fast that
>> > > you're always pulling back tapes just for the indexes, etc. Painful.
>> > >
>> > > So test this part of your backup scheme first, or find the user forums
>> > > for Cohesity/Rubrik and ask that question there. And of course try to
>> > > ask people who have the same size environment as you want to backup as
>> > > well.
>> > >
>> > > Cheers,
>> > > John
>> >
>> > _______________________________________________
>> > Toasters mailing list
>> > Toasters@teaparty.net
>> > https://www.teaparty.net/mailman/listinfo/toasters
>>
>> _______________________________________________
>> Toasters mailing list
>> Toasters@teaparty.net
>> https://www.teaparty.net/mailman/listinfo/toasters
>>
> _______________________________________________
> Toasters mailing list
> Toasters@teaparty.net
> https://www.teaparty.net/mailman/listinfo/toasters

>>>>> "Sami" == Sami Kapanen <sami.kapanen@hamk.fi> writes:

Sami> I will check these,

Sami> Rubrik states it is using SnapDiff. Cohesity says it is based on
Sami> NAS snapshots as well.

Be careful and look at the details of how those snapshots are used.
When doing NDMP backups, yes, the backup is based on a snapshot of the
volume. And if you have lots of files, then the restores will be
*painful* if you only need one or two files in a 10 million file
volume. NDMP isn't really quick on that type of restore in my
experience.

Sami> Restore times are important, but not crucial to us (University),
Sami> we don't loose millions per day if/when restoring.

So if you keep your students and faculty and staff completely
segregated in terms of the volumes they keep their data and home dirs
on, then you can provide different levels of backup guarantees.
Faculty/Staff get better ones of course.

So if you use snapshots as a way to let users recover files from
*oops* moments where they deleted a file an hour ago or so and want it
back, then you might run into the issue of someone duming a large new
files, or deleting alot of data which all goes into the snapshots, and
you run out of disk space... then you have a tough call to make.

Do you delete snapshots to make room? Do you grow the volume?

Same with backups to a degree, if you have a full backup that has been
running for a day or two, and you run out of space, do you kill the
backup?

The worse case scenario is the tough one. It's been 20+ years since I
worked in a Uni, and disk space was always a problem even then, since
some users would try to game the system.

John
_______________________________________________
Toasters mailing list
Toasters@teaparty.net
https://www.teaparty.net/mailman/listinfo/toasters