performance

You are currently browsing articles tagged performance.

In a previous post I discussed the new backup environment I’ve been deploying, what solutions we picked, and how they apply to the datacenter.  But I also mentioned that we had remote sites with systems we need to back up but I didn’t explain how we addressed them.  Frankly, the previous post was getting long and backing up remote offices is tricky so it deserved it’s own discussion.

Now that we had Symantec NetBackup running in the datacenter, backup up the bulk of our systems to disk by way of DataDomain, we need to look at remote sites.  For this we deployed Symantec NetBackup PureDisk.  Despite the fact that it has NetBackup in the name, PureDisk is an entirely different product with it’s own servers, clients, and management interfaces.  There are some integration points that are not-obvious at first but become important later.  Essentially PureDisk is two solutions in a single product — 1:) a “source-dedupe” backup solution that can be deployed independent of any other solution, and 2:) a “target-dedupe” backup storage appliance specifically integrated with the core NetBackup product via an option called PDDO.

As previously discussed, backing up a remote site across a WAN is best accomplished with a source-dedupe solution like PureDisk or Avamar.  This is exactly what we intended to do.  Most of our remote site clients are some flavor of UNIX or Windows and installing PureDisk clients was easily accomplished.  Backup policies were created in PureDisk and a little over a day later we had the first full backup complete.  All subsequent nightly backups transfer very small amounts of data across the WAN because they are incremental backups AND because the PureDisk client deduplicates the data before sending it to the PureDisk server.  The downside to this is that the PureDisk jobs have to scheduled, managed, and monitored from the PureDisk interface, completely separate from the NetBackup administration console.  Backups are sent to the primary datacenter and stored on the local PureDisk server, then the backed up data is replicated to the PureDisk server in the DR datacenter using PureDisk native replication.  Restores can be run from either of the PureDisk servers but must un-deduplicate the data before sending across the WAN making restores much slower than backups.  This was a known issue and still meets our SLAs for these systems.

Our biggest hurdle with PureDisk was the client OS support.  Since we have a very diverse environment we ran into a couple clients which had operating systems that PureDisk does not support.  Both Netware and x86 versions of Solaris are currently not supported, both of which were running in our remote sites.

We had a few options:

1.) Use the standard NetBackup client at the remote site and push all of the data across the WAN

2.) Deploy a NetBackup media server in the remote site with a tape library and send the tapes offsite

3.) Deploy a NetBackup media server in the remote site with a small DataDomain appliance and replicate

4.) Deploy a NetBackup media server and ALSO use PureDisk via the PDDO option (PureDisk Deduplication Option)

Option 1 is not feasible for any serious amount of data, Option 2 requires a costly tape library and some level of media handling every day, and Option 3 just plain costs too much money for a small remote site.

Option 4, using PDDO, leverages PureDisk’s “target-dedupe” persona and ends up being a very elegant solution with several benefits.

PDDO is a plug-in that installs on a Netbackup media server.  The PDDO plug-in deduplicates data that is being backed up by that media server and sends it across the network to a PureDisk server for storage.  The beauty of this option is that we were able to put a Netbackup media server in our remote site without any tape or other storage.  The data is copied from the client to the media server over the LAN, de-duplicated by PDDO, then sent over the WAN to the datacenter’s PureDisk server.  We get the bandwidth and storage efficiencies of PureDisk while using standard NetBackup clients.  A byproduct of this is that you get these PureDisk benefits without having to manage the backups in PureDisk’s separate management console.  To reduce the effects of the WAN on the performance of the backup jobs themselves, and to make the majority of restores faster, we put some internal disk on the media server that the backup jobs write to first.  After the backup job completes to the local disk, NetBackup duplicates the backup data to the PureDisk storage server, then duplicates another copy to the DR datacenter.  This is all handled by NetBackup lifecycle policies which became about 1000X more powerful with the 6.5.4 release.  I’ll discuss the power of lifecycle policies, specifically with the 6.5.4 release, when I talk about OST later.

So the result of using PureDisk/PDDO/NetBackup together is a seamless solution, completely managed from within NetBackup, with all the client OS support the core NetBackup product has, the WAN efficiencies of source-dedupe, the storage efficiencies of target-dedupe, and the restore performance of local storage, but with very little storage in the remote site.

Remote Site Backup…  Done!!

For the near future, I’m considering putting NetBackup media servers with PDDO on VMWare in all of the remote sites so I can manage all of the backups in NetBackup without buying any new hardware at all.  This is not technically supported by Symantec but there is no tape/scsi involved so it should work fine.  Did I mention we wanted to avoid tape as much as possible?

Incidentally, despite my love for Avamar, I don’t believe they have anything like PDDO available in the Networker/Avamar integration and Avamar’s client OS support, while better than PureDisk’s, is still not quite as good as Netbackup and Networker.

Okay, so how does OST play into NetBackup, PureDisk, PDDO, and DataDomain?  What do the lifecycle policies have to do with it? And what is so damned special about lifecycle policies in NetBackup 6.5.4?  All that is next…

Tags: , , , , , , , ,

This is the 3rd part of a multi-part discussion on capacity vs performance in SAN environments.  My previous post discussed the use of thin provisioning to increase storage utilization.  Today we are going to focus on a newer technology called Data De-Duplication.

Data De-Duplication can be likened to an advanced form of compression.  It is a way to store large amounts of data with the least amount of physical disk possible.

De-duplication technology was originally targeted at lowering the cost of disk-based backup.  DataDomain (recently acquired by EMC Corp) was a pioneer in this space.  Each vendor has their own implementation of de-duplication technology but they are generally similar in that they take raw data, look for similarities in relatively small chunks of that data and remove the duplicates.  The diagram below is the simplest one I could find on the web.  You can see that where there were multiple C, D, B, etc blocks in the original data, the final “de-duplicated” data has only one of each.  The system then stores metadata (essentially pointers) to track what the original data looked like for later reconstruction.

Image from eWeek Article

The first and most widely used implementations of de-dupe technology were in the backup space where much of the same data is being backed up during each backup cycle and many days of history (retention) must be maintained.  Compression ratios using de-duplication alone can easily exceed 10:1 in backup systems.  The neat thing here is that when the de-duplication technology works at the block-level (rather than file-level) duplicate data is found across completely disparate data-sets.  There is commonality between Exchange email, Microsoft Word, and SQL data for example.  In a 24 hour period, backing up 5.7TB of data to disk, the de-dupe ratio in my own backup environment is 19.2X plus an additional 1.7X of standard compression on the post de-dupe’d data, consuming only 173.9GB of physical disk space.  The entire set of backed up data, totaling 106TB currently stored on disk, consumes only 7.5TB of physical disk.  The benefits are pretty obvious as you can see how we can store large amounts of data in much less physical disk space.

There are numerous de-duplication systems available for backup applications — DataDomain, Quantum DXi, EMC DL3D, NetApp VTL, IBM Diligent, and several others.  Most of these are “target-based” de-duplication systems because they do all the work at the storage layer with the primary benefit being better use of disk space.  They also integrate easily into most traditional backup environments.  There are also “source-based” de-duplication systems — EMC Avamar and Symantec PureDisk are two primary examples.  These systems actually replace your existing backup application entirely and perform their work on the client machine that is being backed up.  They save disk space just like the other systems but also reduce bandwidth usage during the backup which is extremely useful when trying to get backups of data across a slow network connection like a WAN.

So now you know why de-duplication is good, and how it helps in a backup environment.. But what about using it for primary storage like NAS/SAN environments?  Well it turns out several vendors are playing in that space as well.  NetApp was the first major vendor to add de-duplication to primary storage with their A-SIS (Advanced Single Instance Storage) product.  EMC followed with their own implementation of de-duplication on Celerra NAS.  They are entirely different in their implementation but attempt to address the same problem of ballooning storage requirements.

EMC Celerra de-dupe performs file-level single-instancing to eliminate duplicate files in a filesystem, and then uses a proprietary compression engine to reduce the size of the files themselves.  Celerra does not deal with portions of files.  In practice, this feature can significantly reduce the storage requirements for a NAS volume.  In a test I performed recently for storing large amounts of Syslog data, Celerra de-dupe easily saved 90% of the disk space consumed by the logs and it hadn’t actually touched all of the files yet.

NetApp’s A-SIS works at a 4KB block size and compares all data within a filesystem regardless of the data type.  Besides NAS shares, A-SIS also works on block volumes (ie: FiberChannel and iSCSI LUNs) where EMC’s implementation does not.  Celerra has an advantage when working with files which contain high amounts of duplication in very small block sizes (like 50 bytes) since NetApp looks at 4KB chunks.  Celerra’s use of a more traditional compression engine saves more space in the syslog scenario but NetApp’s block level approach could save more space than Celerra when dealing with lots of large files.

The ability to work on traditional LUNs presents some interesting opportunities, especially in a VMWare/Hyper-V environment.  As I mentioned in my previous post, virtual environments have lots of redundant data since there are many systems potentially running the same operating system sharing the disk subsystem.  If you put 10 Windows virtual machines on the same LUN, de-duplication will likely save you tons of disk space on that LUN.  There are limitations to this that prevent the full benefits from being realized.  VMWare best practices require you to limit the number of virtual machine disks sharing the same SAN lun for performance reasons (VMFS locking issues) and A-SIS can only de-dupe data within a LUN but not across multiple LUNs.  So in a large environment your savings are limited.  NetApp’s recommendation is to use NFS NAS volumes for VMWare instead of FC or iSCSI LUNs because you can eliminate the VMFS locking issue and place many VMs on a single very large NFS volume which can then be de-duplicated.  Unfortunately there are limits on certain VMWare features when using NFS so this may not be an option for some applications or environments.  Specifically, VMWare Site Recovery Manager which coordinates site-to-site replication and failover of entire VMWare environments does not currently support NFS as of this writing.

When it comes to de-duplication’s impact on performance it’s kind of all over the map.  In backup applications, most target based systems either perform the work in memory while the data is coming in or as a post-process job that runs when the backups for that day have completed.  In either case, the hardware is designed for high throughput and performance is not really a problem.  For primary data, both EMC and NetApp’s implementations are post-process and do not generally impact write performance.  However, EMC has limitations on the size of files that can be de-duplicated before a modification to a compressed file causes a significant delay.  Since they also limit de-duplication to files that have not been accessed or modified for some period of time, the problem is minimal in most environments.  NetApp claims to have little performance impact to either read or writes when using A-SIS.  This has much to do with the architecture of the NetApp WAFL filesystem and how A-SIS interacts with it but it would take an entirely new post to describe how that all works.  Suffice it to say that NetApp A-SIS is useful in more situations than EMC’s Celerra De-duplication.

Where I do see potential problems with performance regardless of the vendor is in the same situation as thin provisioning.  If your application requires 1000 IOPS but you’ve only got 2 disks in the array because of the disk space savings from thin-provisioning and/or de-duplication, the application performance will suffer.  You still need to service the IOPS and each disk has a finite number of IOPS (100-200 generally for FC/SCSI).  Flash/SSD changes the situation dramatically however.

Right now I believe that de-duplication is extremely useful for backups, but not quite ready for prime-time when it comes to primary storage.  There are just too many caveats to make any general recommendations.  If you happen to purchase an EMC Celerra or NetApp FAS/IBM nSeries that supports de-duplication, make sure to read all the best-practices documentation from the vendor and make a decision on whether your environment can use de-duplication effectively, then experiment with it in a lab or dev/test environment.  It could save you tons of disk space and money or it could be more trouble than it’s worth.  The good thing is that it’s pretty much a free option from EMC and NetApp depending on the hardware you own/purchase and your maintenance agreements.

Tags: , , , , , , , , , , , , , ,

In my previous post, where I discussed the problem of unusable (or slack) disk space on a SAN, I promised a follow-up with techniques on how to increase storage utilization.  I realized that I should discuss some related technologies first and then follow that up with how to put it all together.  So today I start by talking about Thin Provisioning.  I will then follow up with an explanation of De-Duplication and finally talk about how to use multiple technologies together to get the most use out of your storage.

So what is Thin Provisioning?  It is a technology that allows you to create LUNs or Volumes on a storage device such that the LUN/Volume(s) appear to the host or client to be larger than they actually are.   In general, NAS clients and SAN attached hosts see “Thin Provisioned” LUNs just as they see any other LUN but the actual amount of disk space used on the storage device can be significantly smaller than the provisioned size.  How does this help increase storage utilization?  Well, with thin provisioning you provide applications with exactly the storage they want and/or need but you don’t have to purchase all of the disk capacity up front.

Let’s start with a comparison of using standard LUNs vs thin LUNs with a theoretical application set:

Say we have 3 servers, each running Windows Server.  The operating system partition is on local disk and application data drives are on SAN.  Each server runs an application that collects and stores data over time and the application owner expects that over the next year or so the data will grow to 1TB on each server.  In this particular case we also know that the application’s performance requirements are relatively low.

With traditional provisioning we might create 3 LUNs that are 1TB each and present them to the servers.  This provides the application with room for the expected growth.  Using 300GB FC disks we can carve out three 4+1 RAID5 sets, create one LUN in each and it would work fine.  Alternatively we could use wide striping (ie: a MetaLUN on EMC Clariion) and put all three LUNs on the same 15 disks.  Either way we’ve just burned 15 disks on the storage array based on uncertain future requirements.  If we were stingier with storage we could create smaller LUNs (500GB for example) and use LUN expansion technology to increase the size when the application data fills the disk to that capacity.

In the Thin Provisioning world we still create three 1TB LUNs but they would start out by taking no space.  The pool of disk that the LUNs get provisioned from doesn’t even need to have 3TB of capacity.  As the application data grows over the next 12 months or longer the pool size only needs to grow to accommodate the actual amount of data stored.  Depending on the storage array, we can add disks to the pool one at a time.  So on day one we start with 3 disks in the pool, and then add additional disks one by one throughout the year.  We can then create additional LUNs for other applications without adding disks.  As we add disks to the pool, we expand the capacity available for all of the LUNs to grow (up to each LUN’s maximum size) and we increase performance for ALL of the LUNs in the pool since we are adding spindles.  The real-world benefits come as we consolidate numerous LUNs into a single disk pool.

The nice thing about this approach is that we stop managing the size of individual LUNs and just manage the underlying disk pool as a whole.   And the cost-per-GB for SAN disk constantly goes down so we can spend only what we have to today, and when we add more later it will likely be a little cheaper.  Disk capacity utilization will be much higher in a thin model compared with the traditional/thick model.

The story gets even better in a virtual server environment such as with MS Hyper-V or VMWare ESX.  First, the virtual server OS drives are on the SAN in addition to the application data, and there can be multiple virtual disks on the same LUN.  Whether physical or virtual, we need to maintain some free space in the disks to keep applications running, plus with virtual systems we need some free space on the LUN for features of the virtualization technology like snapshots.  The net effect is that in a virtualized environment, disk utilization never gets much above 50% when slack space at both the virtual layer and inside the virtual servers is considered.  With thin provisioning we could potentially store twice the number of virtual servers on the same physical disks.

There are caveats of course.  Maintaining performance is the primary concern.  Whether used in a thick LUN or thin LUN, each disk has a specific amount of performance.   Thin provisioning has no effect on the amount of IOPS or bandwidth the application requires nor the amount of IOPS the physical disk can handle.  So even if thin provisioning saves 50% disk space in your environment, you may not be able to use all of that reclaimed space before running into performance bottlenecks.  If the storage array has QOS features (ie: EMC Clariion NQM) it is possible to prioritize the more important applications in your disk pool to maintain performance where it matters.

Other problems that you may encounter have to do with interoperability.  For starters, some applications are not “thin-friendly”; ie: they write data in such a way as to negate any benefit that thin provisioning provides.  Also, while many storage arrays support thin provisioning, each has different rules about the use of thin LUNs.  For example, in some scenarios you can’t replicate thin LUNs using native array tools.  It pays to do your homework before choosing a new storage array or implementing thin provisioning.

I didn’t cover thin provisioning in NAS environments directly but the feature works in the same manner.  Thin volumes are provisioned from pools of storage and users/clients see a large amount of available disk space even if the disk pool itself is very small.  Since NAS is traditionally used for user home directories and departmental shares, absolute performance is usually not as much of a concern so thin provisioning is much easier to implement and in many cases is the default behavior or simply a check box on NAS appliances like EMC Celerra or NetApp FAS.

Thin provisioning is a powerful technology when used where it makes sense.  In my next post I’ll explain de-duplication technology and then talk about how these technologies can be used together plus some workarounds for the caveats that I’ve mentioned.

Tags: , , , , , , , , , , , , , ,

In the past–in the days of 2GB,4GB,9GB,18GB and even 36GB drives–when you were tasked with purchasing and configuring hard drives for an application, you were given the amount of storage space required for the application and that was pretty much good enough. If you or your company were more organized you’d do an analysis of the performance requirements for that application (ie: IOPS, read/write ratios, bandwidth, etc.) to make sure you had enough spindles to accommodate the application. More often than not, the capacity requirements necessitated more disk than the performance so you’d build your RAID group and fill it up all the way.

Fast forward a few years and 72GB drives are no longer available, 146GB drives are getting close to end-of-sale and there are 300, 400, 600GB drives, and terabyte SATA drives available for almost any storage system or server. The problem is that as these hard drives get bigger, they aren’t getting any faster. In fact, SATA drives are relatively new in the Enterprise space and are slower than traditional 10,000 and 15,000 RPM SCSI drives. But they hold terabytes of data. Today, performance is the primary requirement and capacity is second because in general you need more spindles for the performance of your application than you do to achieve the capacity requirement.

As an example, let’s take a 100GB SQL database that requires 800 IOPS at 50% Read/50% Write.

Back in the day with 18GB drives you’d need 12 disks to provide ~100GB of space in RAID10. Using SCSI-3 10K drives, you can expect about 140 IOPS per disk giving you 1680 IOPS available. Accounting for RAID10 write penalties, you’d have an effective 1100 IOPS, more than enough for your workload of 800 IOPS.

Today, a single 146GB 10K disk can provide all the capacity required for this database; but you still need at least 10 disks to achieve your 800 IOPS workload with RAID10, or 15 disks with RAID5. The capacity of a RAID10 group with ten 146GB drives is approximately 680GB, leaving you with 580GB of free (or slack) space in the RAID group. The trouble is that you can’t use that space for any of your other applications because the SQL database requires all of the performance available in that RAID group. Change it to RAID5, or use new larger disks, and it’s even worse. Switching to 15K RPM drives can help, but it’s only a 30% increase in performance.

If you are managing SAN storage for a large company, your management probably wants you to show them high disk capacity utilization on the SAN to help justify the cost of storage consolidation. But as the individual disk sizes get larger, it becomes increasingly difficult to keep the capacity utilization high, and for many companies it ends up dropping. Thin Provisioning and De-Duplication technologies are all the rage right now as storage companies push their wares, and customers everywhere are hoping that those buzzwords can somehow save them money on storage costs by increasing capacity utilization. But be aware, if you have slack space due to performance requirements, those technologies won’t do you any good and could hurt you. They are useful for certain types of applications, something I’ll discuss in a later post.

So what do you do? Well, there’s not a lot you can do except educate your management on the difference between sizing for performance and sizing for capacity. They should be aware that slack space is a byproduct of the ever increasing size of hard disk drives. Some vendors are selling high speed flash or SSD disks for their SAN storage systems which can be 30-50X faster than a 15K RPM drive and have similar capacities. But flash has a significant cost which only makes sense if you can leverage most of the IOPS available in each disk. In the next installment I’ll discuss tiered data techniques and how they can overcome some of these problems, increasing performance in some cases while also increasing utilization rates.

Tags: , , , ,

Newer entries »