I have spent most of the day looking over the products from TMS (Texas Memory Systems) that IBM just acquired. One of the questions I have always wondered is how to map performance back to a certain technology. When dealing with DRAM and Flash devices there seems to trade offs on each. The first that comes to mind is DRAM requires some sort of battery backup as it will loose the data contents when power is lost. Most DRAM cards and appliances have this under control with some sort of destaging to SSD or they have some sort of battery attached to the IO card that allows the DRAM time to hold information until power is restored.
DRAM is typically faster than its flash cousin as well as reliable and more durable. Typically there is less controller latency due to the lack of complexity of wear leveling and garbage collection. DRAM is still more expensive that Flash and has the problem of needing power all the time.
When looking at systems to find out how to decide which solution fits your environment it comes down to price and IO. The DRAM solutions are usually smaller in size but can push more IO. For Example the TMS 420 is 256GB of storage in a 4U frame that can push 600,000 IOPs. Not bad if you need 256GB of really fast space. This could be used for very high transaction volumes. This can be deployed with traditional storage and used for the frequently used database tables and indexes while lower IO tables can be thrown over to the spinning hard disk side.
In comparison the TMS 820 Flash array delivers a whopping 24TB in a 1U space and can push a meek 450,000 IOPS. This is somewhat incredible as the footprint is small and dense but still gives you the punch needed to beef up parts of your infrastructure. I stared running the numbers to compare this with say a V7000 with all SSD drives and we can’t come close. You could virtualize the system under the SVC code (included in the V7000) and use the Easy Tier function to move hot data to and from the TMS array which gives you the performance needed. I see why IBM decided to acquire TMS now.
So who wins in a DRAM and Flash discussion? The vendors of course, they are the ones who are going after this market aggressively. I think most consumers are trying to figure out if its needed to spend money on moving a database from 1500 disks at sub 20 ms response to using 200 larger disk and adding the DRAM or Flash device to keep the same latency. As an architect I want to keep in mind how much space and environmentals all of those disk eat up and having an alternative even if it costs more up front, is appealing.
Last week IBM published some very interesting results on the Storage Performance Council website. Using the SPC-1 test method, IBM raised more than just a few eye brows. IBM configured 18 200GB SSDs in a mirrored configuration which was able to attain 120k IOPS with less than 5 ms response time even at the 100% load.
IBM used an all SSD array that fit into a single 2U space and mirrored the drives in RAID 1 fashion. These were all put into a pool and 8 volumes were carved out for the AIX server. The previous SPC1 run IBM performed used spinning media and virtualized the systems using the SAN Volume Controller virtualization engine. This gave IBM the top spot in the SPC1 benchmark with over 520,000 IOPS costing a wopping $6.92 per IOP.
This has been compared to the Oracle/Sun ZFS 7420 who published 137k IOPS late last year. When matched to the IBM V7000 results we see the V7000 was $181,029 with roughly $1.50 per IOP compared to the SUN 7420 who came in at $409,933 and $2.99 per IOP. The V7000 was able to perform 88% of the work at 44% of the price. Oracle has not come back with any type of statement but I can only bet they are seething over this and will be in the labs trying to find a way to lower their cost while still performing.
The SPC1 benchmark tests the performance of a storage array doing mainly random I/O in a business environment. It has been called out for not being a very realistic workload as it tends to carter to higher end, cache heavy systems. Now we have a mid range box that not only has the most IOPs trophy but also wins in the ‘Best Bang for your Buck’ category too.
I would have liked to see what the results would have been with the addition of compression to the software feature stack. This going to be a game changer for IBM as the inline compression of block data is way beyond what other vendors are doing. Couple that with the virtualization engine and now I can compress data on anyone’s storage array. The V7000 is definitely becoming a smarter storage solution for a growing storage market.
Every year DCIG scores and ranks over 50 midrange arrays to help storage buyers make an educated decision on future purchases. Like most of these reports, it takes months to put together and by the time the report is released, the information is stale and out of date. I noticed in this year’s report they XiV gen3 system was listed as the number 6 system behind some of the other vendors that were ranked with a higher score. I decided to dig in and see what was the difference between the number 6 and the number 1 array.
The fist thing I noticed is the ranking on the XiV was based on some old/incorrect data. Here is a snapshot of the report with red circles on things that are not correct:
A couple of things we should point is the support for SSD drives. IBM XiV started support for the multi-level cell drives to enhance its read cache up to 6TB on a full populated system. This increases performance by 3X in most cases. Other options listed like mirroring and snapshot licenses included have been part of the system since the Gen2 days.
The XiV system also supports VASA, that crazy acronym inside of another acronym. The VMWare API for Storage Awareness allows for automation of storage operations and monitor storage alerts. Most storage vendors have come to the same conclusion that IBM has that you have support VASA if you are going to be part of the future.
As you prepare to sit down and start to decide which here are a few things to consider when looking at the XiV system:
1. Management is close to zero. This system is not only easy to deploy but it self tunes itself. The problems with other traditional systems like hot spots or provisioning is made easier with XiV technology.
2. All features are included. No need to worry if you bought the right licenses as all of them are included in the price.
3. XiV is true grid based storage. You really have to compare apples to apples and look under the covers at other storage systems. They may claim some type of grid architecture but in reality they are still using active / active pairs instead of 15 modules all working together on a infiband backbone.
While this is no dig on the DCIG team, it is a little misleading potential customers. I spoke with the analyst that puts these together and he admitted that it was done back in the September time frame and may not represent the true feature and functionality that IBM offers today. I also learned they are working on a more interactive report that can be updated more regularly.
The differences in what they published and what is offered today would more than surpass some of the other storage arrays listed ahead of the XiV. I also would like to see them consider the XiV more of a enterprise array and not so much of the midrange but that’s another blog article entirely.
One thing that has gotten a lot of attention in the last 6 months is how much cache you can put into a solution. Fusion IO has caught fire with its cache cards that you put into the host. Texas Memory and Violin have solid state arrays that can help accelerate your SAN environment. But those only mask the real problem.
Most of the time people are not saturating their 8GB FC link or even their 10GB CNA link to between their host system and their storage. So network is usually not that big of a problem (if configured correctly of course).
The cache cards that are installed in the host systems do offload some of the reads and writes that normally go to the storage array but unless the cache is shared, then you can get into situations where you are keeping multiple copies of the same data in multiple systems. This is effective for each host but not for the entire data center.
Then there is the all mighty “lets put faster disk in between the real disk and the hosts” solution that has been around since we started getting off Ultra Wide SCSI disk arrays. This does again solve the issue of performance but becomes either an one off solution or involves time to get up and running and only benefits the data that lands on those disks.
At IBM, the XIV Storage System has been evolving since we purchased the technology a few years ago. When I first learned of the system I thought the information was wrong, no way could a box that was just a bunch of servers with no raid out perform some of the tried and tested systems not only at IBM but in the storage market. After doing the research and talking to customers that run XIV in their datacenter, I was starting to believe. Was I drinking the Kool-Aid (no this is not a Netapp Blog) or was this a real shift in the storage technology?
Fast forward a couple of years and now IBM is deploying the third generation of XIV. One of the biggest factors way customers switch to XIV is how it lowers the amount of administration overhead. This shows up when other vendors are trying to put so many features and functionality into their product but they are missing the fact that most people want it to be easier. As their data footprint grows, they know the amount of time to keep up with were data lives or the time it takes to provision a lun will grow exponentially.
It was only after talking with a customer who previously not purchased IBM storage and hearing their story of how much work went into keeping up with the day to day running of the storage that I understood that the XIV was more than just a pretty face/GUI.
Today, IBM announced the addition of solid state drives to its XIV system. But wait, doesn’t XIV only use near line SAS drives and touts the whole ‘no tiers’ strategy? Correct. The XIV system is adding these SSD drives not for another tier but to increase the amount of cache available to the entire storage array (unless you thing cache is really tier 0, then ok whatever.). This improves performance of the entire array three times the original XIV GEN3 SPC-2/E benchmark. But wait, there’s more!
Because XIV spreads the load across all of the modules in the array, all of the data that is stored gets a bigger benefit than just adding cache to a host. The XIV can now also exclude certain volumes from benefitting from the improved performance allowing you to maintain efficient use of the SSD technology. The XIV executes an intelligent prefetch of data that is triggered when random reads are sensed. For random operations, XIV SSD cache will prefetch pages to increase the cache hit ratio. With XIV, SSD hotspots are monitored and managed by the storage system. Data placement algorithms from IBM Research are designed to improve wear leveling and optimize performance. XIV SSD Caching is the first commercial implementation of this IBM research project. The future of storage is going to change were customers react to requirements, projects and deployments that are easier to manage on a larger scale.A storage system like XIV that drives the tangible benefits of easy to use with the intangibles (seeing your kids on the weekends instead of doing a data migration on a volume that is out of space and you can’t add space on the fly or even roll back to a snapshot cause if you turn them on the box will melt) shows how IBM is ahead of the game. For more information about XIV and the 3x improvement go here
As is customary for many bloggers at the end of the year we take time to reflect on the past year and make some predictions for next year. This is always fun because I get a chance to see what people predicted for this year and who was right / wrong. Some people are more right than wrong but its fun to guess at what will happen next year none the less.
2011 was a great year for storage and IT as a whole. A couple of highlights I think were important points this year:
- SSD pricing drops significantly to approximately $3 per GB. With the flooding in Thailand, the price for spinning drives went sky high back in October. Since then, the prices have started to decline but not as quickly as the SSDs.
- Tape is still around and is giving people options. There are only a handful of vendors than even like to talk about tape as another storage tier. Those who do have it in the bag of options levered this as something the others can not provide as a full solution.
- Archive and backup were debated, debated again and hopefully the marketing people have learned the difference. I think there are times were backups can be archives but not the other way around. There are people out there that backup their archives but that is a whole blog article to it’s self.
- Mobile apps were plentiful from Fruit Ninja to Facebook to business app like Quick Office flooded the market place. Not only were people developing for the iPad, iPod and iPoop platforms but we saw the rise of the Droid (Google) and Blackberry (RIM) even Windows is now reporting over 50,000 apps available for downloads.
- Clouds got a little more defined and people are starting to see the benefits of having the right ‘cloud’ for the right job. This time year the future of clouds was a little cloudier than what we see them as today. The funny thing I believe most people realized this year was we have been clouding in IT for a long time just under different names.
- Social media was the biggest part of IT in 2011 in my opinion. I saw a fundamental shift in how people got information and how that influenced their decisions. From CEOs blogging to Charlie Sheen going up in flames on Twitter, the warlocks and trolls out there were craving something more. Social media is the new dot com era and now we wait for the bubble to burst.
Now for the good part. Here is what I think 2012 will bring to the IT / Storage world. Note: If any of these do or don’t come true I will deny any involvement.
- Big Data moves into full swing. If you think you heard a lot of people talking about Big Data in 2011 then prepare yourselves for the avalanche of bloggers, researchers, analysts, marketers, sales people, you name it to bombard you with not just what Big Data is but what to do with it. I suspect technologies like Hadoop and Analytics will drive products more than typical structured data storage.
- Protection of remote data on mobile devices like tablets and phones will be a bigger concern. With the rise of these devices people have started to move from the traditional desktop or even laptop. There is already an uptick on the number of viruses in the wild that are designed for mobile users. As more data is generated and stored on them the higher the risk companies face in loosing data. I predict companies will either rely on public clouds like Amazon S3 or Dropbox to help protect from data loss or even go private and force users to back their data to a central repository.
- Software will continue to drive innovation over hardware. Virtualization was a big part of 2011 and that will only continue to grow bigger in the datacenter. Storage systems for the most part are made up of the same parts from the same suppliers. It’s the software that is put to use that drives the efficiency, performance and utilization of the hardware. Those storage vendors who can get out of the hardware speeds and features will show you how their solution solves the problems of today. There are still some customers who want to know how many 15K RPM drives are in the solution. I think there will always be gear-heads who want to know these things, but they are getting fewer.
- Scale out grid clustering with global name space will dominate the new dynamic of how to deploy units. Netapp should be releasing the long anticipated Cluster Mode of Data Ontap sometime in 2012. I hear not everyone will be getting carte blanche on the download so be ready to be told its still not prime time. Even though it took Netapp nine years to get a real product in the market for general consumption. Other vendors like EMC, IBM and HP will all be touting their own version of scale out / grid / you name it as the best way to drive up the efficiency number. Do your research and make sure to compare apples to apples. Just because they all say the same thing doesn’t mean its done the same way.
- Tiering will be on everyone’s mind. Even with the price of SSDs coming down they still are a bit pricey for a SSD only solution. Companies like Violin Memory or Fusion IO will help keep I/O up at the server instead of hitting the storage system. Automatic file based tiering like that of the Active Cloud Engine from IBM will determine from policies where data needs to be written and moved down the $/GB slope as it ages. IBM also has a great automatic tiering solution called Easy Tier that is on the DS8000, SVC and the V7000 takes the guess work out of when to put data on faster disks.
- Consolidation of smaller systems with fewer islands of storage will be a key initiative in 2012. As the economy flat lines on the global operating table IT budgets will be looking for ways to cut line items out of the cost of running the shop. This is a continuance from 2011 but with the push of ‘Big Data’ companies will be asked to take on more data demand with the same budget as last year. As customers pool their resources together to meet these demands they will pool their storage platforms together for better utilization. Look for data migration tools like SVC from IBM to help make these transitions easier.
Finally, I send you the best for a new year full of exciting challenges. Happy New Year!