Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Tom Buskey
Sorry, POC == Proof of Concept.

In this case, it's an all in one (AIO) cloud with 1 physical box that's not
really suitable for production.

On Thu, Sep 28, 2017 at 3:02 PM, Ted Roche  wrote:

> On Thu, Sep 28, 2017 at 2:32 PM, Tom Buskey  wrote:
> >
> > I misspoke about LVM for Glance/Swift.  The backend for the images are on
> > top of a filesystem in the POC clouds.  LVM is used for Cinder, the block
> > image store.  Ceph is often used to drop in replace LVM for Cinder and
> files
> > for Swift objects.
> >
>
> I'm following a disturbing amount of this discussion, but the only POC
> acronym I know is not appropriate in this context. Define, please?
>
>
> --
> Ted Roche
> Ted Roche & Associates, LLC
> http://www.tedroche.com
> ___
> gnhlug-discuss mailing list
> gnhlug-discuss@mail.gnhlug.org
> http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
>
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Ted Roche
On Thu, Sep 28, 2017 at 2:32 PM, Tom Buskey  wrote:
>
> I misspoke about LVM for Glance/Swift.  The backend for the images are on
> top of a filesystem in the POC clouds.  LVM is used for Cinder, the block
> image store.  Ceph is often used to drop in replace LVM for Cinder and files
> for Swift objects.
>

I'm following a disturbing amount of this discussion, but the only POC
acronym I know is not appropriate in this context. Define, please?


-- 
Ted Roche
Ted Roche & Associates, LLC
http://www.tedroche.com
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Tom Buskey
On Thu, Sep 28, 2017 at 2:00 PM, Joshua Judson Rosen  wrote:

> On 09/28/2017 01:46 PM, Tom Buskey wrote:
> > I work with OpenStack.  It manages images in Glance which sit above its
> object storage, Swift.
> >
> > On the POC clouds, you can use LVM as a backend for Glance.
> Snapshotting is *very* slow.  30 minutes for a snap of a
> > 80GB VM that's shutdown.
>
> OK..., that surprises me. A lot.
>


>
> For comparison, I just made an LVM snapshot of a volume 50% larger than
> that, that's *in use*
> (and mostly not in cache, if that even makes a difference, since my
> buffer+cache shows as only 17GB *total*),
> and the whole operation took only a fraction of a second:
>
> rozzin@zuul:~ $ time sudo lvcreate --name home_snap --size 128G
> --snapshot zuul-vg/home
>   Using default stripesize 64.00 KiB.
>   Logical volume "home_snap" created.
>
> real0m0.349s
> user0m0.028s
> sys 0m0.060s
>
>
> How in the world does that translate to 30-minutes (*5 thousand* x time)
> for a volume only 0.63x as big?
>
> When you say "snapshotting on top of LVM", does that entail actually
> making a full copy
> after the LVM snapshot is made--or something like that?
>
>
I'm not exactly sure what OpenStack VM snapshots are doing under the
covers.  I'm sure it's not an LVM snapshot.  Openstack VMs use .qcow2
files, which are COW.  There is a copy/convert process in the Openstack
process so the image the VM uses can be resumed and you can redeploy the
snapshot as a new VM or even copy/launch it on another system.

I misspoke about LVM for Glance/Swift.  The backend for the images are on
top of a filesystem in the POC clouds.  LVM is used for Cinder, the block
image store.  Ceph is often used to drop in replace LVM for Cinder and
files for Swift objects.
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Joshua Judson Rosen
On 09/28/2017 02:14 PM, mark wrote:
> AWS/EBS is not LVM under the covers, it's more like NFS; and snapshots are 
> more like VMware & how it does snapshots.

I have never used VMWare and have no idea how it does anything. Can you provide 
more insight on what that means?


> The OS cache exclusion refers to read-ahead and write caching going on in RAM.

Yes, I got that. The reason I included that in the citation was actually that I 
took it
as supporting my "this looks like atomic COW snapshotting" conclusion, because 
that's
exactly what I'm accustomed to getting through LVM (snapshotting a block device
captures all of the blocks that *have actually been written* at the time of the 
snapshot).

> On Sep 28, 2017 1:17 PM, "Joshua Judson Rosen"  > wrote:
> 
> I'm working on a project that uses Amazon AWS-provided VPS instances,
> and the other guy on the project is telling me that "snapshotting hourly 
> may degrade performance",
> and I'm trying to determine where that's actually true. My gut feeling is 
> that it sounds kind of bogus.
> 
> >From the information I've been able to find about how Amazon's stuff 
> works (either in terms
> of how it's _implemented_ [for which I'm finding basically no insight] or 
> how it's _characterized_
> [in the engineering sense, not the literary sense]...), it really sounds 
> a _lot_ like Amazon
> is just using LVM snapshots, e.g. from  >:
> 
>         "snapshots can be done in real time while the volume is attached 
> and in use.
>          However, snapshots only capture data that has been written to 
> your Amazon EBS volume,
>          which might exclude any data that has been locally cached by 
> your application or OS."
> 
>         "By design, an EBS Snapshot of an entire 16 TB volume should take 
> no longer than the time
>          it takes to snapshot an entire 1 TB volume. However, the actual 
> time taken to create
>          a snapshot depends on several factors including the amount of 
> data that has changed
>          since the last snapshot of the EBS volume."
> 
> ... though I'm not entirely sure how to interpret that last bit about 
> "time taken to create a snapshot
> depends on... the amount of data that has changed since the last 
> snapshot";
> the _first half of that statement_ reads as "creating a snapshot is 
> constant time",
> which basically screams to me "copy-on-write just like LVM, and they're 
> probably implemented
> in terms of LVM".
> 
> Any insight here as to whether my gut is correct on this, or whether I'm 
> actually likely
> to notice an impact from hourly snapshots of, say, a 200-GB volume? How 
> about a 1-TB volume?
> 
> The only thing I'm seeing from Amazon that seems to _vaguely_ support 
> (maybe) the notion
> that `snapshotting too often' would be something to worry about is this 
> bit from elsewhere
> in that same FAQ page (under the heading of "performance", whereas the 
> others were
> under the heading of "snapshots" and a subheading of "performance 
> consistency of my HDD-backed volumes":
> 
>         Another factor is taking a snapshot which will decrease expected 
> write performance
>         down to the baseline rate, until the snapshot completes.
> 
> ... and, taken in the context of the previously-cited notes about 
> snapshots being
> `not base on volume-size but maybe influenced by 
> changed-since-last-snapshot set size'
> (and in the context of the explanations they give for HDD-backed vs. 
> SSD-backed storage),
> I'm basically reading that as:
> 
>         `if you're using HDD-backed storage then it's because you care 
> about *throughput*
>          more than *response time* and are likely to be monitoring 
> throughput,
>          and if you're monitoring throughput you may notice a *momentary 
> dip in throughput*
>          as the *HDDs* need to seek around to find the volume boundaries 
> and set up the COW records.'
> 
> Even if you don't have any insight into what's actually happening under 
> the covers at Amazon,
> does my reading of all of this sound right to you?
> 
> And, perhaps more interestingly, are these same caveats from Amazon 
> generally applicable to LVM?
> 
> --
> Connect with me on GNU social network: 
>  >
> Not on the network? Ask me for an invitation to the nhcrossing.com 
>  social hub
> ___
> gnhlug-discuss mailing list
> gnhlug-discuss@mail.gnhlug.org 
> http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/ 
> 
> 

-- 
"Don't be afraid to ask (λf.((λx.xx) 

Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread mark
AWS/EBS is not LVM under the covers, it's more like NFS; and snapshots are
more like VMware & how it does snapshots. The OS cache exclusion refers to
read-ahead and write caching going on in RAM.

Mark
On Sep 28, 2017 1:17 PM, "Joshua Judson Rosen" 
wrote:

> I'm working on a project that uses Amazon AWS-provided VPS instances,
> and the other guy on the project is telling me that "snapshotting hourly
> may degrade performance",
> and I'm trying to determine where that's actually true. My gut feeling is
> that it sounds kind of bogus.
>
> >From the information I've been able to find about how Amazon's stuff
> works (either in terms
> of how it's _implemented_ [for which I'm finding basically no insight] or
> how it's _characterized_
> [in the engineering sense, not the literary sense]...), it really sounds a
> _lot_ like Amazon
> is just using LVM snapshots, e.g. from :
>
> "snapshots can be done in real time while the volume is attached
> and in use.
>  However, snapshots only capture data that has been written to
> your Amazon EBS volume,
>  which might exclude any data that has been locally cached by your
> application or OS."
>
> "By design, an EBS Snapshot of an entire 16 TB volume should take
> no longer than the time
>  it takes to snapshot an entire 1 TB volume. However, the actual
> time taken to create
>  a snapshot depends on several factors including the amount of
> data that has changed
>  since the last snapshot of the EBS volume."
>
> ... though I'm not entirely sure how to interpret that last bit about
> "time taken to create a snapshot
> depends on... the amount of data that has changed since the last snapshot";
> the _first half of that statement_ reads as "creating a snapshot is
> constant time",
> which basically screams to me "copy-on-write just like LVM, and they're
> probably implemented
> in terms of LVM".
>
> Any insight here as to whether my gut is correct on this, or whether I'm
> actually likely
> to notice an impact from hourly snapshots of, say, a 200-GB volume? How
> about a 1-TB volume?
>
> The only thing I'm seeing from Amazon that seems to _vaguely_ support
> (maybe) the notion
> that `snapshotting too often' would be something to worry about is this
> bit from elsewhere
> in that same FAQ page (under the heading of "performance", whereas the
> others were
> under the heading of "snapshots" and a subheading of "performance
> consistency of my HDD-backed volumes":
>
> Another factor is taking a snapshot which will decrease expected
> write performance
> down to the baseline rate, until the snapshot completes.
>
> ... and, taken in the context of the previously-cited notes about
> snapshots being
> `not base on volume-size but maybe influenced by
> changed-since-last-snapshot set size'
> (and in the context of the explanations they give for HDD-backed vs.
> SSD-backed storage),
> I'm basically reading that as:
>
> `if you're using HDD-backed storage then it's because you care
> about *throughput*
>  more than *response time* and are likely to be monitoring
> throughput,
>  and if you're monitoring throughput you may notice a *momentary
> dip in throughput*
>  as the *HDDs* need to seek around to find the volume boundaries
> and set up the COW records.'
>
> Even if you don't have any insight into what's actually happening under
> the covers at Amazon,
> does my reading of all of this sound right to you?
>
> And, perhaps more interestingly, are these same caveats from Amazon
> generally applicable to LVM?
>
> --
> Connect with me on GNU social network:  com/rozzin>
> Not on the network? Ask me for an invitation to the nhcrossing.com social
> hub
> ___
> gnhlug-discuss mailing list
> gnhlug-discuss@mail.gnhlug.org
> http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
>
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Joshua Judson Rosen
On 09/28/2017 01:32 PM, Ken D'Ambrosio wrote:
> I would say it's unlikely to be LVM, because LVM is content-ignorant; it 
> snapshots the entire volume, which is
> inefficient, and when you're Amazon, you care a LOT about being efficient.  
> Instead, I imagine they're using some
> content-aware CoW solution such as ZFS.  But, whatever mechanism, I agree 
> with your opinion: I doubt that their solution
> -- almost certainly CoW of some sort -- stands a chance of being more than 
> even slightly impactful.

Oh--yeah, ZFS is another good candidate. Actually, there are a few others that 
I can think of as well

But it is basically `screaming "COW"' at me, and my gut is telling me that this 
`fear of over-snapshotting'
is basically (generally) the same as when people talk about how they `need to 
do multithreading [for EVERYTHING]
because it's so expensive to fork a new process' (there are some corner cases 
where fork() is actually
`too expensive' in at least some sense [and I've actually run into some of 
those cases],
 but *most* of those claims always seemed to be from people who didn't even 
know that COW was a thing...).

> $.02, YMMV and other assorted disclaimers,
> 
> -Ken
> 
> 
> On 2017-09-28 13:16, Joshua Judson Rosen wrote:
>> I'm working on a project that uses Amazon AWS-provided VPS instances,
>> and the other guy on the project is telling me that "snapshotting
>> hourly may degrade performance",
>> and I'm trying to determine where that's actually true. My gut feeling
>> is that it sounds kind of bogus.
>>
>>> From the information I've been able to find about how Amazon's stuff works 
>>> (either in terms
>> of how it's _implemented_ [for which I'm finding basically no insight]
>> or how it's _characterized_
>> [in the engineering sense, not the literary sense]...), it really
>> sounds a _lot_ like Amazon
>> is just using LVM snapshots, e.g. from :
>>
>> "snapshots can be done in real time while the volume is attached and in 
>> use.
>>  However, snapshots only capture data that has been written to your
>> Amazon EBS volume,
>>  which might exclude any data that has been locally cached by your
>> application or OS."
>>
>> "By design, an EBS Snapshot of an entire 16 TB volume should take no
>> longer than the time
>>  it takes to snapshot an entire 1 TB volume. However, the actual time
>> taken to create
>>  a snapshot depends on several factors including the amount of data
>> that has changed
>>  since the last snapshot of the EBS volume."
>>
>> ... though I'm not entirely sure how to interpret that last bit about
>> "time taken to create a snapshot
>> depends on... the amount of data that has changed since the last snapshot";
>> the _first half of that statement_ reads as "creating a snapshot is
>> constant time",
>> which basically screams to me "copy-on-write just like LVM, and
>> they're probably implemented
>> in terms of LVM".
>>
>> Any insight here as to whether my gut is correct on this, or whether
>> I'm actually likely
>> to notice an impact from hourly snapshots of, say, a 200-GB volume?
>> How about a 1-TB volume?
>>
>> The only thing I'm seeing from Amazon that seems to _vaguely_ support
>> (maybe) the notion
>> that `snapshotting too often' would be something to worry about is
>> this bit from elsewhere
>> in that same FAQ page (under the heading of "performance", whereas the
>> others were
>> under the heading of "snapshots" and a subheading of "performance
>> consistency of my HDD-backed volumes":
>>
>> Another factor is taking a snapshot which will decrease expected
>> write performance
>> down to the baseline rate, until the snapshot completes.
>>
>> ... and, taken in the context of the previously-cited notes about
>> snapshots being
>> `not base on volume-size but maybe influenced by
>> changed-since-last-snapshot set size'
>> (and in the context of the explanations they give for HDD-backed vs.
>> SSD-backed storage),
>> I'm basically reading that as:
>>
>> `if you're using HDD-backed storage then it's because you care about
>> *throughput*
>>  more than *response time* and are likely to be monitoring throughput,
>>  and if you're monitoring throughput you may notice a *momentary dip
>> in throughput*
>>  as the *HDDs* need to seek around to find the volume boundaries and
>> set up the COW records.'
>>
>> Even if you don't have any insight into what's actually happening
>> under the covers at Amazon,
>> does my reading of all of this sound right to you?
>>
>> And, perhaps more interestingly, are these same caveats from Amazon
>> generally applicable to LVM?
> 

-- 
"Don't be afraid to ask (λf.((λx.xx) (λr.f(rr."
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Joshua Judson Rosen
On 09/28/2017 01:48 PM, Bill Ricker wrote:
> The lack of coherence due to OS cave not being flushed should still be a 
> concern. 

In the general case, yes. In my particular case I'm specifically concerned only
with data that's stored transactionally to the extent that (and I really hope
that I'm not grossly mistaken on this...) I'd expect to survive an unexpected 
power-loss,
like PostgreSQL and git.

(and in the case of git, I'm only talking about its internal object-store,
 *not working trees*, which I know from experience *cannot* be expected to 
survive that--
 in other words, always sync between "git pull" or "git checkout" and a 
power-cut!)

> OTOH I saw a storage level replication system propagate corruption to the 
> remote site's copy of the Production DBMS ...
> So it perfectly replicated the primary's failure. Oops. Easiest recovery was 
> restoring a nightly backup to the test
> system since both Prod nodes were so hosed.
> This is bit one reason I like best-effort (asynchronous) dbms level 
> transaction replication. The remote is 30s behind
> but is in a consistent state. Most users can handle checking if their last 
> txn before crash survived; most will even if
> not instructed to! (Even if asked not to!)

☺

-- 
Connect with me on GNU social network: 
Not on the network? Ask me for an invitation to the nhcrossing.com social hub
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Joshua Judson Rosen
On 09/28/2017 01:46 PM, Tom Buskey wrote:
> I work with OpenStack.  It manages images in Glance which sit above its 
> object storage, Swift.
> 
> On the POC clouds, you can use LVM as a backend for Glance.  Snapshotting is 
> *very* slow.  30 minutes for a snap of a
> 80GB VM that's shutdown.

OK..., that surprises me. A lot.

For comparison, I just made an LVM snapshot of a volume 50% larger than that, 
that's *in use*
(and mostly not in cache, if that even makes a difference, since my 
buffer+cache shows as only 17GB *total*),
and the whole operation took only a fraction of a second:

rozzin@zuul:~ $ time sudo lvcreate --name home_snap --size 128G 
--snapshot zuul-vg/home
  Using default stripesize 64.00 KiB.
  Logical volume "home_snap" created.

real0m0.349s
user0m0.028s
sys 0m0.060s


How in the world does that translate to 30-minutes (*5 thousand* x time)
for a volume only 0.63x as big?

When you say "snapshotting on top of LVM", does that entail actually making a 
full copy
after the LVM snapshot is made--or something like that?

> You can use other storage backends in OpenStack that are faster.  A full non 
> LVM Swift.  Ceph and glusterfs are common
> choices where performance matters.  They wouldn't be using ZFS but probably 
> something using their S3 object store.
> 
> 
> 
> 
> On Thu, Sep 28, 2017 at 1:32 PM, Ken D'Ambrosio  > wrote:
> 
> I would say it's unlikely to be LVM, because LVM is content-ignorant; it
> snapshots the entire volume, which is inefficient, and when you're
> Amazon, you care a LOT about being efficient.  Instead, I imagine
> they're using some content-aware CoW solution such as ZFS.  But,
> whatever mechanism, I agree with your opinion: I doubt that their
> solution -- almost certainly CoW of some sort -- stands a chance of
> being more than even slightly impactful.
> 
> $.02, YMMV and other assorted disclaimers,
> 
> -Ken
> 
> 
> On 2017-09-28 13:16, Joshua Judson Rosen wrote:
> > I'm working on a project that uses Amazon AWS-provided VPS instances,
> > and the other guy on the project is telling me that "snapshotting
> > hourly may degrade performance",
> > and I'm trying to determine where that's actually true. My gut feeling
> > is that it sounds kind of bogus.
> >
> >> From the information I've been able to find about how Amazon's stuff
> >> works (either in terms
> > of how it's _implemented_ [for which I'm finding basically no insight]
> > or how it's _characterized_
> > [in the engineering sense, not the literary sense]...), it really
> > sounds a _lot_ like Amazon
> > is just using LVM snapshots, e.g. from
> > >:
> >
> >       "snapshots can be done in real time while the volume is attached 
> and
> > in use.
> >        However, snapshots only capture data that has been written to 
> your
> > Amazon EBS volume,
> >        which might exclude any data that has been locally cached by your
> > application or OS."
> >
> >       "By design, an EBS Snapshot of an entire 16 TB volume should take 
> no
> > longer than the time
> >        it takes to snapshot an entire 1 TB volume. However, the actual 
> time
> > taken to create
> >        a snapshot depends on several factors including the amount of 
> data
> > that has changed
> >        since the last snapshot of the EBS volume."
> >
> > ... though I'm not entirely sure how to interpret that last bit about
> > "time taken to create a snapshot
> > depends on... the amount of data that has changed since the last
> > snapshot";
> > the _first half of that statement_ reads as "creating a snapshot is
> > constant time",
> > which basically screams to me "copy-on-write just like LVM, and
> > they're probably implemented
> > in terms of LVM".
> >
> > Any insight here as to whether my gut is correct on this, or whether
> > I'm actually likely
> > to notice an impact from hourly snapshots of, say, a 200-GB volume?
> > How about a 1-TB volume?
> >
> > The only thing I'm seeing from Amazon that seems to _vaguely_ support
> > (maybe) the notion
> > that `snapshotting too often' would be something to worry about is
> > this bit from elsewhere
> > in that same FAQ page (under the heading of "performance", whereas the
> > others were
> > under the heading of "snapshots" and a subheading of "performance
> > consistency of my HDD-backed volumes":
> >
> >       Another factor is taking a snapshot which will decrease expected
> > write performance
> >       down to the baseline rate, until the snapshot completes.
> >
> > ... and, taken in the context of the previously-cited notes about
> > snapshots being
>

Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Bill Ricker
The lack of coherence due to OS cave not being flushed should still be a
concern.

OTOH I saw a storage level replication system propagate corruption to the
remote site's copy of the Production DBMS ... So it perfectly replicated
the primary's failure. Oops. Easiest recovery was restoring a nightly
backup to the test system since both Prod nodes were so hosed.
This is bit one reason I like best-effort (asynchronous) dbms level
transaction replication. The remote is 30s behind but is in a consistent
state. Most users can handle checking if their last txn before crash
survived; most will even if not instructed to! (Even if asked not to!)
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Tom Buskey
I work with OpenStack.  It manages images in Glance which sit above its
object storage, Swift.

On the POC clouds, you can use LVM as a backend for Glance.  Snapshotting
is *very* slow.  30 minutes for a snap of a 80GB VM that's shutdown.

You can use other storage backends in OpenStack that are faster.  A full
non LVM Swift.  Ceph and glusterfs are common choices where performance
matters.  They wouldn't be using ZFS but probably something using their S3
object store.




On Thu, Sep 28, 2017 at 1:32 PM, Ken D'Ambrosio  wrote:

> I would say it's unlikely to be LVM, because LVM is content-ignorant; it
> snapshots the entire volume, which is inefficient, and when you're
> Amazon, you care a LOT about being efficient.  Instead, I imagine
> they're using some content-aware CoW solution such as ZFS.  But,
> whatever mechanism, I agree with your opinion: I doubt that their
> solution -- almost certainly CoW of some sort -- stands a chance of
> being more than even slightly impactful.
>
> $.02, YMMV and other assorted disclaimers,
>
> -Ken
>
>
> On 2017-09-28 13:16, Joshua Judson Rosen wrote:
> > I'm working on a project that uses Amazon AWS-provided VPS instances,
> > and the other guy on the project is telling me that "snapshotting
> > hourly may degrade performance",
> > and I'm trying to determine where that's actually true. My gut feeling
> > is that it sounds kind of bogus.
> >
> >> From the information I've been able to find about how Amazon's stuff
> >> works (either in terms
> > of how it's _implemented_ [for which I'm finding basically no insight]
> > or how it's _characterized_
> > [in the engineering sense, not the literary sense]...), it really
> > sounds a _lot_ like Amazon
> > is just using LVM snapshots, e.g. from
> > :
> >
> >   "snapshots can be done in real time while the volume is attached
> and
> > in use.
> >However, snapshots only capture data that has been written to your
> > Amazon EBS volume,
> >which might exclude any data that has been locally cached by your
> > application or OS."
> >
> >   "By design, an EBS Snapshot of an entire 16 TB volume should take
> no
> > longer than the time
> >it takes to snapshot an entire 1 TB volume. However, the actual
> time
> > taken to create
> >a snapshot depends on several factors including the amount of data
> > that has changed
> >since the last snapshot of the EBS volume."
> >
> > ... though I'm not entirely sure how to interpret that last bit about
> > "time taken to create a snapshot
> > depends on... the amount of data that has changed since the last
> > snapshot";
> > the _first half of that statement_ reads as "creating a snapshot is
> > constant time",
> > which basically screams to me "copy-on-write just like LVM, and
> > they're probably implemented
> > in terms of LVM".
> >
> > Any insight here as to whether my gut is correct on this, or whether
> > I'm actually likely
> > to notice an impact from hourly snapshots of, say, a 200-GB volume?
> > How about a 1-TB volume?
> >
> > The only thing I'm seeing from Amazon that seems to _vaguely_ support
> > (maybe) the notion
> > that `snapshotting too often' would be something to worry about is
> > this bit from elsewhere
> > in that same FAQ page (under the heading of "performance", whereas the
> > others were
> > under the heading of "snapshots" and a subheading of "performance
> > consistency of my HDD-backed volumes":
> >
> >   Another factor is taking a snapshot which will decrease expected
> > write performance
> >   down to the baseline rate, until the snapshot completes.
> >
> > ... and, taken in the context of the previously-cited notes about
> > snapshots being
> > `not base on volume-size but maybe influenced by
> > changed-since-last-snapshot set size'
> > (and in the context of the explanations they give for HDD-backed vs.
> > SSD-backed storage),
> > I'm basically reading that as:
> >
> >   `if you're using HDD-backed storage then it's because you care
> about
> > *throughput*
> >more than *response time* and are likely to be monitoring
> throughput,
> >and if you're monitoring throughput you may notice a *momentary
> dip
> > in throughput*
> >as the *HDDs* need to seek around to find the volume boundaries
> and
> > set up the COW records.'
> >
> > Even if you don't have any insight into what's actually happening
> > under the covers at Amazon,
> > does my reading of all of this sound right to you?
> >
> > And, perhaps more interestingly, are these same caveats from Amazon
> > generally applicable to LVM?
> ___
> gnhlug-discuss mailing list
> gnhlug-discuss@mail.gnhlug.org
> http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
>
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/


Re: Is Amazon AWS/EBS snapshotting just LVM, or what?

2017-09-28 Thread Ken D'Ambrosio
I would say it's unlikely to be LVM, because LVM is content-ignorant; it 
snapshots the entire volume, which is inefficient, and when you're 
Amazon, you care a LOT about being efficient.  Instead, I imagine 
they're using some content-aware CoW solution such as ZFS.  But, 
whatever mechanism, I agree with your opinion: I doubt that their 
solution -- almost certainly CoW of some sort -- stands a chance of 
being more than even slightly impactful.

$.02, YMMV and other assorted disclaimers,

-Ken


On 2017-09-28 13:16, Joshua Judson Rosen wrote:
> I'm working on a project that uses Amazon AWS-provided VPS instances,
> and the other guy on the project is telling me that "snapshotting
> hourly may degrade performance",
> and I'm trying to determine where that's actually true. My gut feeling
> is that it sounds kind of bogus.
> 
>> From the information I've been able to find about how Amazon's stuff 
>> works (either in terms
> of how it's _implemented_ [for which I'm finding basically no insight]
> or how it's _characterized_
> [in the engineering sense, not the literary sense]...), it really
> sounds a _lot_ like Amazon
> is just using LVM snapshots, e.g. from 
> :
> 
>   "snapshots can be done in real time while the volume is attached and 
> in use.
>However, snapshots only capture data that has been written to your
> Amazon EBS volume,
>which might exclude any data that has been locally cached by your
> application or OS."
> 
>   "By design, an EBS Snapshot of an entire 16 TB volume should take no
> longer than the time
>it takes to snapshot an entire 1 TB volume. However, the actual time
> taken to create
>a snapshot depends on several factors including the amount of data
> that has changed
>since the last snapshot of the EBS volume."
> 
> ... though I'm not entirely sure how to interpret that last bit about
> "time taken to create a snapshot
> depends on... the amount of data that has changed since the last 
> snapshot";
> the _first half of that statement_ reads as "creating a snapshot is
> constant time",
> which basically screams to me "copy-on-write just like LVM, and
> they're probably implemented
> in terms of LVM".
> 
> Any insight here as to whether my gut is correct on this, or whether
> I'm actually likely
> to notice an impact from hourly snapshots of, say, a 200-GB volume?
> How about a 1-TB volume?
> 
> The only thing I'm seeing from Amazon that seems to _vaguely_ support
> (maybe) the notion
> that `snapshotting too often' would be something to worry about is
> this bit from elsewhere
> in that same FAQ page (under the heading of "performance", whereas the
> others were
> under the heading of "snapshots" and a subheading of "performance
> consistency of my HDD-backed volumes":
> 
>   Another factor is taking a snapshot which will decrease expected
> write performance
>   down to the baseline rate, until the snapshot completes.
> 
> ... and, taken in the context of the previously-cited notes about
> snapshots being
> `not base on volume-size but maybe influenced by
> changed-since-last-snapshot set size'
> (and in the context of the explanations they give for HDD-backed vs.
> SSD-backed storage),
> I'm basically reading that as:
> 
>   `if you're using HDD-backed storage then it's because you care about
> *throughput*
>more than *response time* and are likely to be monitoring throughput,
>and if you're monitoring throughput you may notice a *momentary dip
> in throughput*
>as the *HDDs* need to seek around to find the volume boundaries and
> set up the COW records.'
> 
> Even if you don't have any insight into what's actually happening
> under the covers at Amazon,
> does my reading of all of this sound right to you?
> 
> And, perhaps more interestingly, are these same caveats from Amazon
> generally applicable to LVM?
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/