Tag: shared-storage

Is Backup Software Dead?

Is Backup Software Dead?
Image courtesy of Simon Howden at FreeDigitalPhotos.net

Is backup software dead?  Everywhere I look, I see increased functionality within storage appliances and operating systems.  Appliances will backup themselves, Operating Systems now have quiescing and basic back up support, and the cloud is making backup targets stupid easy.  Should I buy dedicated backup software, or does my hardware, hypervisor or Operating System handle this?

As a storage professional, I will never discourage anyone from taking backups. As a matter of fact, I personally believe that more is better.  I am sure that many of have heard the popular saying ‘Two is one and one is none.’  Anyone who has mounted a blank backup that “worked last time” understands the wisdom of multiple backups.  Balancing this wisdom against the cost of additional methods of backup – what should I do?  While there is no one answer that will work for everyone, discussions help us formulate plans.

Many hardware solutions provide backup

I’m a big fan of taking administrative tasks off-line.  As an axiom to that, the closer I can get backup to where the data lives, the faster the backup will occur and the SMALLER the impact to production systems.  It stands to reason – if a system snapshot takes place on the storage appliance and it takes milliseconds to execute, isn’t that better than a full backup through system agents that may take hours over the weekend?

To take advantage of this, many storage vendors have added support within their hardware for snapshots and replication.  In essence, this makes a copy of your data volume and moves it within your environment.  Yes, this usually only works on products within the same manufacturing family.  Yes, vendors must support quiescing.  But many OS vendors are now building the functionality within their operating system to quiesce resident data painlessly.  Well, painlessly once you get it set up.  But what was once the realm of large, intense database houses, or financial trading houses now ships with many OSes.

This seems easy enough, right?  Your storage appliance and OS will do most of the difficult work.  But what about support for your hypervisor?  Maybe those legacy apps don’t support some sort of OS quiescing?  Or what about those that don’t even have a dedicated storage appliance?

Backup Software

While it will never be as fast as dedicated storage appliance backup, backup software does have a place.  Many places in fact.

Backup Software’s arguably most important function is as a broker.  The software acts as the middleman between your data (the source) and where ever you would like a copy of the data (the target).  And it provides a greater amount of flexibility than traditional “baked-in” solutions from hardware manufacturers.  Of course, this is a simplistic approach, and many backup packages have lots of gizmos and what-nots to make a backup administrator’s life easier.  But the main function is moving data.

Software works well with dissimilar hardware.  Want to backup data across many different manufacturers?  Software can do it. Want to move it between disk, tape, and the cloud?  Removable media?  Software shines here.  Want to work with legacy applications or operating systems that may not support data integrity?  Software does this and gives you the flexibility to customize it to your environment.

What works for you

I see a place for both hardware and software in a backup strategy.  Of course, I’m also the guy that still sees tape as the most economical means to store and archive large amounts of data.  The key point is to do what works for you.  I’ve worked with large organizations that had more data than could be reasonably backed up through software.  In this case, snaps and replication were a great fit.  But those same organizations had legacy apps that needed databases backed up hot and live, then log files backed up as well to insure transactional integrity.  Software to the rescue.

My point is that there are many tools in your toolbelt to use.  But, technology always changes. Does your hardware provide all the things you need to recover your data in an emergency? With the amazing growth of data, do you see software still being a viable backup method into the future?  How do budgets and cost affect your decision?  Please share your thoughts!

 

Share this:

SSD – How much is too much?

Speed, SSD, fast, SAN, NAS
Image courtesy of pixabay

I wrote an article just a few days ago entitled “Where did the 15k disk drive go?”  It was a short piece, quickly done and meant to draw fairly obvious conclusions.  When given a choice between faster and fastest, for the same or close money, people will always choose fastest.  Little did I suspect the sheer amount of comments and emails that I would get from that article.  It appears that everyone has an opinion on storage technology and how storage vendors build out their appliances.  So, in the spirit of keeping the discussion going, I’ve decided to ask the flip side of most of the comment and email subjects.  “If 15k drives are dead, then how much SSD is too much?”  Let the games begin!

How much SSD?

I heard a Texan once say “How much is too much?  Well, if it’s money in the bank or cows on the ranch, you can never have too much!”  He was talking about things that directly affected his performance as a cattleman and his ability to perform his job or company function.  The same can be said for SSD disk in the ever-changing storage arena of business.  How much is too much SSD in a storage array or on a server?  I’m not talking about the sheer amount of physical space – that depends on the applications and data depositories that the application will require.  Plus a little bit for growth.  What I am talking about is a percentage.  Of 100% of the storage space on a given server or storage appliance, just how what percentage should be SSD – fast but expensive?

In my opinion, much will depend on a storage study.  How many IOPs does your environment need so that storage is not the bottleneck in your environment?  Is there too much latency in your SAN or NAS?  If you don’t know the answers to these questions, then a storage study should be your next step.  Check out my article here.  SSD tends to be the most expensive option in GB/$, but that ratio is coming down as manufacturing processes change and get more efficient.  But we all work in the here-and-now, so as of today, how much SSD is too much in your SAN, NAS, or hyperconverged appliance?

All Flash, or no Flash?

I have seen several examples of SSD ratios, all aided by software in one form or another.  These fall into two camps at either end of the spectrum.

To start, there is the storage appliances with no SSD.  These are fairly simple, and I don’t see them around much. If all you need is an array of large disks spinning merrily along, and your storage goals are met, do you really need SSD?  I have been in proof-of-concept trials where SSD would not make any difference is system performance, until the programmers changed the application code to make it more parallel.

Then there is the “all flash all-the-time” argument.  I am familiar with one storage array vendor that sells an all flash array with compression and de-duplication and claims that across the environment, the cost per used Gigabyte is cheaper than their hybrid array (which does not offer any compression type functionality).  Of course with de-duplication your milage may vary, but that makes a compelling argument for all flash.  There are certain industries where milliseconds matter, like stock market trading, or transaction processing.  Those industries go all flash without a second thought.

The middle ground?

So now we reach the middle ground, where the argument get heated.  Hybrid arrays replace the fastest tier of storage with SSD, or use large amounts of SSD as caches to buffer regular hard drives.  Manufacturers use SSD to take the place of those good ol’ 15K drives, as well as some of the 10k drives, too.  The larger and slower SATA drives remain the workhorses for storage.  Older, slower data goes there to die.  Or at least drive your backup software crazy.

So, where does all this leave us?  Should we go ahead and use all flash since it is the wave of the future?  Since I will be replacing my array as I outgrow it, should I buy affordable now, and look to put in all-flash when it is the standard?  Assuming that I am not a government agency with black-budget trillions to spend, how much SSD is too much SSD?  Looking forward to your comments.

Share this:

Where did the 15K disk drive go?

15k disk
Image courtesy of Suriya Kankliang at FreeDigitalPhotos.net

Just a few years ago, everyone wanted disk drives that spun at 15,000 rpm, commonly known as “15k disk”.  Why did people want these?  Well, the faster the spindle turned, the shorter the seek times, the less latency and the faster the writes to that disk.  Since I never worked at any of the drive manufacturers, I can’t really speak to the truth of this, but I do take it on faith.  So when everything on a storage array was spinning disk, why did people want “15k spindles” in the line up?  And since SSD has become so popular, why don’t I really see them anymore?

Why do I want expensive, small disks?

The reason that everyone wanted 15k disk drives was pretty straightforward.  The disk themselves were fairly small in capacity (600GB being a standard size) and expensive on a GB/$ ratio.  But they were FAST.  If there was a target IOPs for a storage array, it was easier to balance out size and speed with a ratio of 15k disk, 10k disk, and standard 7.2k SATA drives.  Speed from the smaller drives and space from the slower drives.  While everything was acceptable ON AVERAGE, the laws of physics still applied to the different speeds of disk. There was a bit of balance that had to happen. You could put your fast access volumes on 15k, but you still needed the SATA drives for the larger storage requirements.  This solution worked, but was expensive – and a bit “hands-on”.

There were even a few manufacturers that started to offer ILM with these systems.  This means that “hot” or active data writes to the 15k disk drives since theoretically the write speed on these is fastest.  Your storage appliance now writes more across the aggregate of your SAN environment.  Once this data is written to the fastest disk on your SAN or NAS, it stays there for a bit.  This logic being that it also has the fastest read times and therefore the best performance when you wanted to recall that data.  These ILM vendors then move the data off of the fastest tier of disk to a slower tier as that data becomes less active or “ages”.  This allowed you to store older, less accessed data on the slower and less expensive tiers of storage.  Because the database has to run quickly, but who cares if it takes accounting a week to get the data for their year-end reporting, right?  Remember that the next time you need that expense report reimbursed!

The next step

Then SSD entered the market.  At an affordable price, that is.  Not only could manufacturers use SSD as caching, but they were large enough that manufacturers could also use them as the fastest tier of data storage in an ILM strategy.  And the form factor of SSD disks allows them to be used in the existing storage appliance enclosures – JUST LIKE spinning disk.  Now, instead of expensive 15k disks, you could put in units in the same form factor that would read and write several hundred times faster than disk.  With the speed and storage capability of SSD, it became unnecessary to use 15K disk in storage appliances for speed.

But I still see some 15k disk out there…

You will still see 15k disk used in local solutions.  A 15K SAS disk RAID 5 array is quite speedy when used in a local physical server.  Virtualization solutions, or database servers will often use 15K spindles for disk targets.  They need sizable storage capacities and quick access.  However, the cost of SSD is coming down.  This allows the justification for installation of SSD disk or arrays in physical servers.  Seagate has stopped development of new models for their 15k disk.  Previously storage technology leapt from Tape to HDD for large data storage like disaster recovery.  Now storage acceptance from high speed disk to SSD will likely accelerate.  Technology to increase access speed, reduce manufacturer costs, and increase storage capacity will accelerate this change.  So long 15k disk, we hardly knew ya!

Share this:

The SMB IT Guy’s guide to ensuring your success from Hyperconvergence.

Image courtesy of Stuart Miles at FreeDigitalPhotos.net

By now, everyone realizes the advantages of server virtualization. Flexibility in the face of rapidly changing technology, reduction in administrative effort on busy IT staff, and cost savings from reducing physical machines is just the beginning. As you may have heard, hyperconverged infrastructure solutions offer all of these advantages, plus the added benefit of simplicity in your environment.

This article is targeted towards small to mid-sized business: 50 to 500 employees supported with 1-5 or so staffers in the IT department. These IT shops don’t rely on specialists, but a few really good “jack-of-all-trades”.  If you are looking for a way to bring this up with the boss, make sure to see the article written for the senior directors or owners in the business here.

So – there are a lot of different hyperconverged vendors out there and a lot of solutions. If you believe the literature and web demos, they will all do everything you need in your environment. How do you know which is the best and what to look out for?

As with everything in life, the answer is – It depends. No one can answer the question of which is best for you, without the intimate knowledge of your environment which probably only you have. What I can do is provide you with some questions that you might want to ask the various solutions providers. These may help you determine which solution works best for your organization, and that management will buy off on.

Here are 5 questions to help you in your inquiry.

What comes in the box?

Well, not literally, but what does this solution entail? How many servers of MINE will this solution cover, and how much extra capacity will I have? Are there any extras that might later cost me money or maintenance fees? Are installation services needed and possibly included in this solution? Is high availability between hardware units included in this quote? The answers to these questions may not make or break the solution for you, but you should know what you are getting for your money. You need to be able to present this effectively to management so no one gets any unpleasant surprises later. Maybe you only need a barebones system right now. That’s fine, but make sure that you know what is included and what everyone’s expectations are.

Licenses

There are a few main solutions out there and they all handle this differently. Many manufacturers of these solutions OEM hypervisors, so ask how that affects the cost of your unit(s). Is there the possibility of having to purchase additional software licenses in order to expand? Are all of the management consoles and utilities provided under the license of the hyperconverged product? If not, what isn’t included that I may want, and where can I get it? Do I need to deal with the hyperconverged manufacturer, or do I have to drag another vendor into this? How many vendors are involved in this solution and who do I call if I need support? Are there different tiers in the number of licenses? What do my maintenance costs look like 3 and 5 years out? If my server count grows by 20% per year, what additional costs will I encounter? Most solutions providers will be more than happy to work these numbers for you, and your management will love your forward thinking “strategic planning”.

Simplicity and Ease of Use

Hyperconverged infrastructure solutions are all about making things simple, right? Find out. Get to know how this particular solution works. You don’t need to see the actual code, but it might be nice to know conceptually how everything fits together. Does this solution come with any training? Is training required? Is training an extra cost? Are basic functions like setting up virtual machines, virtual disks, and virtual NICs intuitive? What about more advanced tasks? That pesky application that we have that demands VLAN tagging, how does this solution support that? Can I do every task I need to do from the management interface? How easy is this product to use for non-pre-sales-engineers-that-don’t-work-for-the-manufacturer?

Backup, Recovery, and Failover

OK – we are looking at this solution because recovery and business continuity are supposedly made much easier with this. Can I stop dropping by the office after hours and on weekends to do silly little server tasks, like rebooting crashed boxes… for payroll… at the end of the month? How does this solution help me with recovery tasks? How does it handle a crashed server? How does the solution handle network failures, disk failures, or whole server failures? Can I SEE it demonstrated live? How will this solution affect my existing backup strategy? Will my current backup solution work, or does this solution include something that replaces it? Does it do native snapshots? How many? Will it replicate those snapshots somewhere automagically? How can my existing DR plan be improved with this solution?

Scalability

Everyone has a constantly changing environment. How does this solution handle growth and changing needs? What does it take to add 20% capacity to this solution? How much does it cost, and how easy is it to do? Will I have to stop production or do it at 3am? Do I need additional chassis to do this, or can I upgrade the units internally? Will this require downtime? What if I want to start moving things to the edge of my infrastructure? How flexible is this product? Do I have the ability to add more memory, CPU, or plain disk to this solution independent of purchasing the next model? What is the roadmap for this product line – Flash disk, software, and NIC speeds?

Hyperconverged infrastructure promises to be an amazing step in the IT virtualization lifecycle. There are different capabilities and features in all of the various solutions. You just need to ask a few questions to figure out which one is right for you. Not just right for you right now, but right for you in 3 to 5 years. Only after you can answer the questions above will you be able to enjoy the REAL benefits of simplicity that hyperconvergence provides.

Share this:

Snapshots – Everyday Uses and Hacks

Storage Snapshot
Image courtesy of ddpavumba at FreeDigitalPhotos.net

Creating snapshots in a storage environment is an amazing technology.  The ability to take an instant “picture” of a data volume is a tool that is used in a variety of ways.  It makes your job easier and more manageable.  It can help secure your environment.

Different vendors implement snapshots in various ways, but the general theory remains the same. An almost instantaneous copy of data that may be moved and manipulated by a system administrator.  The theory of this is nice, but how can we USE this functionality.  Can it make their job easier and protect their systems from the everyday issues they see “in the wild”?

With organizations I work with, we see many innovative uses of snapshotting technology.  There are amazing examples of real world IT organizations making their jobs faster, easier, and much less stressful.  In other words, they used “business hacks” to make their snapshots work for them. We will discuss five real world ways to use snapshots that are relevant and guaranteed applicable to your everyday work load.

Snapshots in your DR strategy

The first things that pops into most people’s mind is backups and disaster recovery.  Snapshots produce an exact copy of virtual machines or data volumes that is stored within the storage appliances.  Most vendors allow these snapshots to be replicated or moved to another storage appliance.  This allows you to use an appliance in another location as a disaster recovery site.  Or, it is possible to mount these snapshots as volumes and allow your backup server to incorporate these exact replicas of data into your existing backup or Disaster Recovery plan.

There are several advantages to this approach.  The data in a snapshot is an exact replica in time, so it is easy to manage RPO and RTO.  Also, this approach takes the data backup “offline” of your production servers.  Sure, the network and storage are still involved in transferring this data, but the data transfers happen out-of-band.  This reduces slow systems and lag.  Many vendors now include APIs for cloud storage in their software and storage appliances.  Now, you may back up your snapshots directly to cloud storage.

Update “insurance” snapshots

We’ve all done it.  Installed that patch from our system or software vendor and it breaks the box.  Perhaps breaks is a strong word.  It temporarily overwhelms our system with new features and benefits. While snapshots can’t make the process of ironing out an ornery system update any easier, it can provide you with insurance.

By taking a snapshot before you update a system, you have an exact copy that you know works.  Suppose you cannot straighten out all the goodness that was Big-Name-Accounting-Package 5.0 before Monday 8am rolls around.  Now you have the ability to fail-back to your old system while you continue to straighten out the misbehaving upgrade.  Almost a form of version control for those of you familiar with the software development world.  This nifty trick also works on desktops.  If you are using VDI, make copies of your desktop images and use the same concept.  It may not save you time getting to the next version, but it will certainly save your bacon as far as system up-time and help-desk calls are concerned.

Gold copy snapshots

If you are making snapshots of servers before you upgrade, you are probably already doing this, but we will mention it anyway.  Snapshots are amazing tools for creating new servers, virtual machines, or desktops.

Once you have installed an operating system and all the various patches and utilities that you routinely use – take a snapshot.  Now this new, untouched system as-pure-as-the-driven-snow will be the starting point for all new servers or desktops that you implement.  This is often referred to as the “Gold copy“, a reference to software development and when code is ready to ship out to customers.

This “Gold copy” has standard system configurations already in place, various drive mappings, and config files.  It is all in there.  Sure you may edit some things like network and licensing, but you have a starting place that is pretty solid.  In the future, if you need to make changes then just make changes and save as a new snapshot.  This may not seem like much, but anyone who has built a new system from scratch will tell you that this is a genuine lifesaver.

This concept applies to both virtual machines and stand-alone servers or desktops.  Several customers we work with will use an application to “ghost” images from storage appliances to a new non-virtualized server or desktop.  Mount the snapshot you would like to use as your system image, then transfer it over to your new hardware using the disk image utility of your choice.  Of course, this works best in a virtualized environment, but it is also a valuable tool for the not-yet-virtualized.  By the way, why aren’t you virtualized yet?

Instant data set snapshots

We regularly hear from customers asking how to generate test data for new systems testing.  In several cases, systems administration is tasked with creating data sets that the consultants or systems specialists can use to ensure the systems are working as anticipated.

Instead of this being a problem, use the best test data that there is – an exact copy of your current, live data.  There is no need to create new data sets from your existing data. By creating a snapshot of your current databases, you may test systems with what was once hot and live data.  But, there is no negative impact if this data is corrupted or destroyed.  You can even create multiple copies of this data to use across multiple tests.

Getting around malware with snapshots

Today’s data environment can be a pretty scary place.  Look no further than the headlines to see

Malware, virus, spyware
Image courtesy of Stuart Miles at FreeDigitalPhotos.net

stories about malware and ransomware wrecking havoc on organizations.  If the recent exploits of the bad guys is any indication, things are getting much larger in scope.  The WannaCry attack is still fresh in everyone’s minds and is rumored to have effected over 230,000 machines world-wide. It is safe to say that there are external threats to your data that can be remediated with snapshots.

A schedule of snapshots  on your storage appliance is the solution.  Whether this is part of your disaster recovery planning or not, set up a schedule. This concept is similar to the “patch insurance” we discussed above.

By making a number of snapshots over time, we are able to go back to former snapshots and explore these snapshots for malware.   Perhaps we may extract data from our snapshots before the encryption activates.  Of course, data sometimes is lost.  It is up to management to decide to pay faceless hackers for your data or try to recover it via backups and snapshots.

Snapshots have been in the storage technology tool bag for a while.  The technology has matured so that most storage array vendors are offering this functionality.  Over years of working with clients, we have discovered many innovative ways that people are using snapshots.  In this article, I have shared what I have seen, but I am interested in what you are doing with your snapshots.  Feel free to share and let everyone know how they can use snapshots within their storage appliance.

 

Share this:

How Do I connect to my Storage Appliance?

Fiber Channel attached Storage Appliance
Image courtesy of cookie__cutter at FreeDigitalPhotos.net

In this article, we are examining the third question asked in our original article The Beginners Guide to what to know before you shop for a Storage Appliance.  That question in a nutshell is “How do I intend to connect my storage so that all of my applications can get to it?”  Well, that question begs a good look at your current environment.  Based on what you find, we will determine if you should connect to your existing environment or connect through other dedicated technologies within your existing environment.  There are also other less common methods to connect to your storage.

Using Existing network infrastructure

Is your network stable?  Every network administrator or sysadmin knows who the problem children are in their network.  Do you have any segments or switches in your environment that are currently congested or causing delays now?  Adding storage to it will only exacerbate the problem.  On the flip side of that coin, a well-running network makes adding storage easy and inexpensive.

In addition, the speed of your existing network will come into play.  Depending on your current storage needs, I would recommend that no one attach storage at speeds of less than 1 Gigabit Ethernet. As 10 GigE becomes more affordable and more pervasive in networks, it is never a bad idea to increase bandwidth to your storage.  Fortunately, many manufacturers enable upgrading with field replaceable units.  Speak with the vendor about this ability in the units you are investigating.

Most storage appliances will support a variety of connection protocols.  For storage area networks (SAN), it is important that iSCSI be supported in the unit.  iSCSI will support most of the externally mounted volumes or LUNs (Logical Unit Number).  For Network Attached Storage (NAS), NFS is a popular way of attaching storage for most virtualization shared storage and *nix computing.  These storage protocols may all be supported, or only some of them.  SMB/CIFS should be supported for full functionality in a Microsoft network.

Using Dedicated connection technologies

There are situations where the use of the existing network may not be advisable.  If the network is older or slow, putting the data needs of shared storage on the network will just exacerbate an already slow situation.  In this case, there are dedicated connection technologies that may come to the rescue.

Ethernet connectivity is still a very viable alternative, using dedicated switches and VLANs.  VLANs are Virtual Local Area Networks that allow for the logical partitioning of ports within a switch to create virtual switches and LANs.  This lets you segregate data traffic and dedicate resources to the various ports that may be passing your data traffic.

Fiber Channel (FC) is a mature, well established connection technology.  FC uses glass fibers to connect physical servers to physical storage using light.  While this technology is a bit more expensive than traditional ethernet switches it does have advantages over ethernet.  There is tremendous support for this protocol in software and hardware because it is a very stable protocol developed specifically for storage.  Fiber Channel allows for data to be consistently delivered with very low overhead.  Fiber Channel switches are available to connect servers to storage in a logical mesh setup, but it is also a regular practice to directly connect servers with FC Host Bus Adapters (HBA’s – think of an HBA as a fiber channel version of a network card).  This will cut out the expense of a Fiber Channel switch for smaller deployments.

Exotic Connection methods

In addition to the well established protocols of Fiber Channel and iSCSI, there are other ways to connect storage.  There are storage appliances out there that will allow connection to servers via specialized technologies like InfiniBand, or SAS ports.  There is eSATA that is available.  These various ways to connect range from the super fast (InfiniBand – and expensive by the way) to the fairly common and slow.  “Exotic” connection technologies serve special cases and are outside the scope of this article.  These connection technologies will limit your field of vendors, but not disqualify you from a storage appliance.

Considerations of Connectivity

In addition to the connection methods discussed above, there are also other connectivity possibilities to consider.  Bonded connections is one.  Bonded connections make multiple physical paths (read cables or ports) to appear as one logical path to data.  In essence, two 1GB Ethernet connections becomes one logical 2 GB Ethernet connection.  A single path of bandwidth to the storage appliance will be quickly overwhelmed.  There will be many servers and users trying to connect to the storage.  Bonding allows several ports to simultaneously send out data.  Bonding also helps with failover.

Another consideration of connectivity is failover.  Although it may not happen often, if a cable, NIC, or port fails on the storage appliance or on the connectivity side, all servers using that storage are suddenly unable to access data.  Or all of your virtual machines may come down at once.  You have placed all of your proverbial eggs in that one proverbial basket.  Failover mitigates this risk accordingly.

This is often accomplished through the use of different controllers or “heads”.  Two (or more) controllers allows for multiple disparate paths to the data.  It allows for one head to crash and you still have access to your data.  It allows for one power supply to fail, and you still have access to your data.  Many manufacturers will vary on how they support this functionality, so it is important to research this carefully.  Make sure that the storage appliance will run on one power supply.  Verify that the controller heads support failover.  Implement bonded connections.

Summary

In this article, we have discussed the final question raised in our original article about finding the best storage appliance for your environment.  We have gone over considerations of attaching the shared storage to your existing network, the prospect of attaching the NAS or SAN via new connectivity, or even attaching via a special, non-standard or exotic connectivity mode.  Many vendors support these differing connectivity methods.  Specialized connectivity will limit the number of storage appliances that you have to choose from. Most users know that they are required from the start and can plan accordingly.

 

 

 

Share this:

The Storage Study – or How do I determine what my environment is using?

Storage Study
© Ultramarine5 | Dreamstime Stock Photos & Stock Free Images

In a previous article, we discussed three important questions to answer about YOUR environment before jumping into a storage appliance.  In this article, we will delve deeper into the first question we asked, “How fast and how much storage do you need?”.  This article is designed for the IT generalist, someone who is looking for some insight on how to do one.

So – how do I tell what I need?  The first step is to do a storage study.  The storage study is done in your environment for a period of around seven days.  Why seven days?  Because that will capture an entire work week of your environment.  And by work week, I mean those weekends that systems guys work and backups run on as well.  Is Saturday a full-backup day?  You want to see what the impact is on your systems.  Perhaps accounting prepares reports for payroll on Wednesdays.  Usually, a seven day sampling of your storage needs will account for standard practices within your environment, and not create information stores that are massive in size.

If you would like to capture more days than seven, break it out into multiple capture files of seven days.  Perhaps doing multiple sampling weeks during significant system events would reveal more details about your environment.  End of the quarter accounting processing?  Start of a new production cycle?  You decide.

The storage study should include several important take-aways collated and also broken out by host or server.  These four important metrics are IOP/s, Latency, Storage Footprint, and information on new (or “hot”) data.  We will delve a bit deeper into what each of these means below.

Input/output OPerations per Second (IOP/s)

What is an IOP, and what does it mean to my environment?  IOP/s simply put are a measurement of how many storage operations your host is doing every second.  IOP/s can be misleading, though.  While a single read operation generally takes 1 IOP, writes to disk can use up to 6 IOP/s for the same bit of information.  Why this happens is a bit technical, so your relevant question should be “How do I account for this?”

In addition to the overall IOP/s number, most studies will include a read vs. write percentage. This is usually written as 65/35, or 65% reads across this study and 35% writes.  This percentage determines how exactly to account for the IOP/s that were collected.  Of primary importance to the IOP/s study, though is the IOP/s over time.  This will help determine when the busy parts of the day are.  You should see numbers for absolute peak (meaning that this was the largest IOP/s event during the entire sampling period), and several percentiles.

The 85th percentile number is what is usually used to determine how to size your system.  You can certainly size your system to accommodate your peak IOP/s, but usually this is more appliance than you really need.  It follows the same logic of building a house that is above a 500 year flood plain.  Sure, your house won’t get flooded out (statistically speaking) for 500 years, but will the house even be standing by then?

Latency

OK – we know about how many IOP/s our systems are using in the course of our storage study.  Now, how long is it taking those IOP/s to be serviced?  In essence, your systems are issuing commands to your storage, but how long does it take your storage to complete the command?  Is that number acceptable?  Milliseconds are the usual time.  Lower is better.

Peak and trending latency are important.  If peak latency reaches 100ms, there is cause to investigate further.  Most applications are tolerant of high latency.  High latency is noticed in database record access times, or the spinning wheel/hourglass of uncommitted data.  It can be a bit tricky to run down exactly where the slowdown may be occurring.  Our primary concern with this storage study is that it is NOT happening along the disk I/O path.  Common culprits are slower disks, inadequate system RAM, and older CPUs.

If you start to see this number trending up or if you see spikes during the day, this is indicative of concerns in your system. While your disk storage may not be the bottleneck, we would like to be able to disqualify it.  Your planned storage appliance should be sized to accommodate any extra load.

Overall Storage footprint

Overall footprint is straightforward.  How much stuff do you have stored on all of your systems?  You will see this represented both by the server and also the entire environment that you collected.  This is often represented by a total amount of space in the environment – all the space on all the hard drives.  The amount of used v. free space is important.  This lets you know how much of all the spinning disks that you are have filled with your data.  A small amount of data on a fast, expensive disk or disks is not cost effective.

If you conduct multiple storage studies, compare the amount of used space from one study to another. This will give you an idea of how quickly your environment is growing.  Most of the storage study tools out there will collect information on each disk individually.  This allows you to drill down to the application level.  Find those greedy disk hogging applications quickly.

This metric will help you to determine how much overall space you should put into your storage appliance.

“Hot Data”

Hot data is data that is accessed, changed, or newly written by your systems within the storage study collection period.  In essence, this is the data that your applications used during the study.  All other data is not accessed, touched, or read during this time, but may be necessary to keep.  This hot data contains clues into how much your overall data needs may be growing every week.

Hot Data also answers the question of how fast your storage needs to be.  Writing data puts more of a strain on a system than reading data.  Hence, we need a faster system the more writing we do.  Hot data also gives us a rough estimate of what new data was written on the system.  This allows us to extrapolate what your storage needs will look like in a quarter or a year given your current rate of growth.

One important aspect that hot data drives is the speed of the storage appliance.  The higher the percentage of hot data the faster the storage appliance needs to be.  The larger the overall amount of hot data, the faster the storage appliance needs to be.  These are important considerations in correctly sizing storage appliances for both size and speed.

Accurate growth rates allow us to properly size the overall storage capacity of a storage appliance.  No one wants to buy too little space right off the bat.  But it is also pertinent that we not buy too much storage at the onset of the project.  Storage prices go down every year as capacity goes up.  It is financially cheaper to only buy the storage when you need it, not buy it all up front.

Summary

A storage study is the first step in determining your needs in a storage appliance.  This report generates many of the metrics that are required to correctly size a storage appliance.  The numbers generated will give us ideas of how much disk space overall we need, and how fast that disk space needs to be.

We have discussed IOP/s, Latency, Storage Footprint, and information on new (or “hot”) data in this article.  Once you have collected these metrics, analyze them.  Next, let’s see how various applications affect our storage needs.

Share this: