The name “Kaminario” intrigues me and I don’t know the meaning of it. But there is a nice roll off the tongue until you say it a few times, fast and your tongue get twisted in a jiffy.
Kaminario is one of the few prominent startups in the all-flash storage space, getting USD$15 million Series C funding from big gun VCs of Sequoia and Globespan Capital Partners in 2011. That brought their total to USD$34 million, and also bringing them the attention of storage market.
I am beginning my research into their technology and their product line, the K2 and see why are they special. I am looking for an angle that differentiates them and how they position themselves in the market and why they deserved Series C funding.
Kaminario was founded in 2008, with their headquarters in Boston Massachusetts. They have a strong R&D facility in Israel and looking at their management lineup, they are headed by several personalities with an Israel background.
All this shouldn’t be a problem to many except the fact that Malaysia don’t recognize Israel diplomatically and some companies here, especially the government, might have an issue with the Israeli link. But then again, we have a lot of hypocrites in Malaysian politics and I am not going to there in my blog. It’s a waste of my time.
The key technology is Kaminario’s K2 SPEAR Architecture and it defines a fundamental method to store and retrieve performance-sensitive data. Yes, since this is an all-Flash storage solution, performance numbers, speeds and feeds are the “weapons” to influence prospects with high performance requirements. Kaminario touts their storage solution scales up to 1.5 million IOPS and 16GB/sec throughput and indeed they are fantastic numbers when you compare them with the conventional HDDs based storage platforms. But nowadays, if you are in the all-Flash game, everyone else is touting similar performance numbers as well. So, it is no biggie.
The secret sauce to the Kaminario technology is of course, its architecture – SPEAR. SPEAR stands for Scale-out Performance Storage Architecture. While Kaminario states that their hardware is pretty much off-the-shelf, open industry standard, somehow under the covers, the SPEAR architecture could have incorporate some special, proprietary design in its hardware to maximize the SPEAR technology. Hence, I believe there is a reason why Kaminario chose a blade-based system in the enclosures of its rack. Here’s a look at their hardware offering:
The idea using blades is a good idea because blades offers integrated wiring, consolidation, simple plug-and-play, ease-of-support, N+1 availability and so on. But this will also can put Kaminario in a position of all-blades or nothing. This is something some customers in Malaysia might have to get used to because many would prefer their racks. I could be wrong and let’s hope I am.
Each enclosure houses 16 blades, with N+1 availability. As I am going through Kaminario’s architecture, the word availability is becoming louder, and this could be something Kaminario is differentiating from the rest. Yes, Kaminario has the performance numbers, but Kaminario is also has a high-available (are we talking 6 nines?) architecture inherent within SPEAR. Of course, I have not done enough to compare Kaminario with the rest yet, but right now, availability isn’t something that most all-Flash startups trumpet loudly. I could be wrong but the message will become clearer when I go through my list of all-Flash – SolidFire, PureStorage, Virident, Violin Memory and Texas Memory Systems.
Each of the blades can be either an ioDirector or a DataNode, and they are interconnected internally with 1/10 Gigabit ports, with at least one blade acting as a standby blade to the rest in a logical group of production blades. The 10Gigabit connection are used for “data passing” between the blades for purpose of load-balancing as well as spreading out the availability function for the data. The Gigabit connection is used for management reasons.
In addition to that there is also a Fibre Channel piece that is fronting the K2 to the hosts in the SAN. Yes, this is an FC-SAN storage solution but since there was no mention of iSCSI, the IP-SAN capability is likely not there (yet).
Here’s a look at the Kaminario SPEAR architecture:
The 2 key components are the ioDirector and the DataNode. A blade can either have a dedicated personality (either ioDirector or DataNode) or it can share both personalities in one blade. Minimum configuration is 2-blades of 2 ioDirectors for redundancy reasons.
The ioDirector is the front-facing piece. It presents to the SAN the K2 block-based LUNs and has the intelligence to dynamically load balance both Reads and Writes and also optimizing its resource utilization. The DataNode plays the role of fetching, storing, and backup and is pretty much the back-end worker.
With this description, there are 2 layers in the SPEAR architecture. And interestingly, while I mentioned that Kaminario is an all-Flash storage player, it actually has HDDs as well. The HDDs do not participate in the primary data serving and serve as containers for backup for the primary data in the SSDs, which can be MLC-Flash or DRAMs. The back-end backup layer comprising of HDDs is what I said earlier about availability. Kaminario is adding data availability as part of its differentiating features.
That’s the hardware layout of SPEAR, but the more important piece is its software, the SPEAR OS. It has 3 patent-pending capabilities, with not so cool names (which are trademarked).
- Automated Data Distribution
- Intelligent Parallel I/O Processing
- Self Healing Data Availability
The Automated Data Distribution of the SPEAR OS acts as a balancer. It balances the data by dynamically and randomly (in an random equilibrium fashion, I think) to spread out the data over the storage capacity for efficiency, SSD longevity and of course, optimized performance balancing.
The second capability is Intelligent Parallel I/O Processing. The K2 architecture is essentially a storage grid. The internal 10Gigabit interconnects basically ties all nodes (ioDirectors and DataNodes) together in a grid-like fashion for the best possible intra-node communications. The parallelization of the I/O Read and Write requests spreads across the nodes in the storage grid, giving the best average response and service times.
Last but not least is the Self Healing Data Availability, a capability to dynamically reconfigure accessibility to the data in the event of node failure(s). Kaminario claims no single point of failure, which is something I am very interested to know if given a chance to assess the storage a bit deeper. So far, that’s the information I am able to get to.
The Kaminario K2 product line comes in 3 model – D, F, and H.
D is for DRAM only and F is for Flash MLC only. The H model is a combination of both Flash and DRAM SSDs. Here how Kaminario addresses each of the 3 models:
Kaminario is one of the early all-Flash storage systems that has gained recognition in 2011. They have been named a finalist in both Storage Magazine and SearchStorage Storage Product of the Year competitions for 2011. This not only endorses a brand new market for solid state storage systems but validates an entirely new category in the storage networking arena.
Kaminario can be one to watch in 2012 as with others that I plan to review in the coming weeks. The battle for Flash racks is coming!
BTW, Dell is a reseller of Kaminario.
Happy Lunar New Year! The Chinese around world has just ushered in the Year of the Water Dragon yesterday. To all my friends and family, and readers of my blog, I wish you a prosperous and auspicious Chinese New Year!
Over the holidays, I have been keeping up with the progress of Solid State Drives (SSDs). I am sure many of us are mesmerized by SSDs and the storage vendors are touting the best of SSDs have to offer. But let me tell you one thing – you are probably getting the least of what the best SSDs have to offer. You might be puzzled why I say things like this.
Let me share with a common sales pitch. Most (if not all) storage vendors will tout performance (usually IOPS) as the greatest benefits of SSDs. The performance numbers have to be compared to something, and that something is your regular spinning Hard Disk Drives (HDDs). The slowest SSDs in terms of IOPS is about 10-15x faster than the HDDs. A single SSD can at least churn 5,000 IOPS when compared to the fastest 15,000 RPM HDDs, which churns out about 200 IOPS (depending on HDD vendors). Therefore, the slowest SSDs can be 20-25x faster than the fastest HDDs, when measured in IOPS.
But the intent of this blogger is to share with you more about SSDs. There’s more to know because SSDs are not built the same. There are write-bias SSDs, read-bias SSDs; there are SLC (single level cell) and MLC (multi level cell) SSDs and so on. How do you differentiate them if Vendor A touts their SSDs and Vendor B touts their SSDs as well? You are not comparing SSDs and HDDs anymore. How do you know what questions to ask when they show you their performance statistics?
SNIA has recently released a set of methodology called “Solid State Storage (SSS) Performance Testing Specifications (PTS)” that helps customers evaluate and compare the SSD performance from a vendor-neutral perspective. There is also a whitepaper related to SSS PTS. This is something very important because we have to continue to educate the community about what is right and what is wrong.
In a recent webcast, the presenters from the SNIA SSS TWG (Technical Working Group) mentioned a few questions that I think we as vendors and customers should think about when working with an SSD sales pitch. I thought I share them with you.
- Was the performance testing done at the SSD device level or at the file system level?
- Was the SSD pre-conditioned before the testing? If so, how?
- Was the performance results taken at a steady state?
- How much data was written during the testing?
- Where was the data written to?
- What data pattern was tested?
- What was the test platform used to test the SSDs?
- What hardware or software package(s) used for the testing?
- Was the HBA bandwidth, queue depth and other parameters sufficient to test the SSDs?
- What type of NAND Flash was used?
- What is the target workload?
- What was the percentage weight of the mix of Reads and Writes?
- Are there warranty life design issue?
I thought that these questions were very relevant in understanding SSDs’ performance. And I also got to know that SSDs behave differently throughout the life stages of the device. From a performance point of view, there are 3 distinct performance life stages
- Fresh out of the box (FOB)
- Steady State
As you can see from the graph below, a SSD, fresh out of the box (FOB) displayed considerable performance numbers. Over a period of time (the graph shown minutes), it transitioned into a mezzanine stage of lower IOPS and finally, it normalized to the state called the Steady State. The Steady State is the desirable test range that will give the most accurate type of IOPS numbers. Therefore, it is important that your storage vendor’s performance numbers should be taken during this life stage.
Another consideration when understanding the SSDs’ performance numbers are what type of tests used? The test could be done at the file system level or at the device level. As shown in the diagram below, the test numbers could be taken from many different elements through the stack of the data path.
Performance for cached data would given impressive numbers but it is not accurate. File system performance will not be useful because the data travels through different layers, masking the true performance capability of the SSDs. Therefore, SNIA’s performance is based on a synthetic device level test to achieve consistency and a more accurate IOPS numbers.
There are many other factors used to determine the most relevant performance numbers. The SNIA PTS test has 4 main test suite that addresses different aspects of the SSD’s performance. They are:
- Write Saturation test
- Latency test
- IOPS test
- Throughput test
A few days ago, Apple paid USD$500 million to buy an Israeli startup, Anobit, a maker of flash storage technology.
Obviously, one of the reasons Apple did so is to move up a notch to differentiate itself from the competition and positions itself as a premier technology innovator. It has won the MP3 war with its iPod, but in the smartphones, tablets and notebooks space, Apple is being challenged strongly.
Today, flash storage technology is prevalent, and the demand to pack more capacity into a small real-estate of flash will eventually lead to reliability issues. The most common type of NAND flash storage is the MLC (multi-level cells) versus the more expensive type called SLC (single level cells).
But physically and the internal-build of MLC and SLC are the exactly the same, except that in SLC, one cell contains 1 bit of data. Obviously this means that 2 or more bits occupy one cell in MLC. That’s the only difference from a physical structure of NAND flash. However, if you can see from the diagram below, SLCs has advantages over MLCs.
NAND Flash uses electrical voltage to program a cell and it is always a challenge to store bits of data in a very, very small cell. If you apply too little voltage, the bit in the cell does not register and will result in something unreadable or an error. If you apply too much voltage, the adjacent cells are disturbed and resulting in errors in the flash. Voltage leak is not uncommon.
The demands of packing more and more data (i.e. more bits) into one cell geometry results in greater unreliability. Though the reliability of the NAND Flash storage is predictable, i.e. we would roughly know when it will fail, we will eventually reach a point where the reliability of MLCs will no longer be desirable if we continue the trend of packing more and more capacity.
That’s when Anobit comes in. Anobit has designed and implemented architectural changes of the way NAND Flash storage is used. The technology in laymen terms comes in 2 stages.
- Error reduction – by understanding what causes flash impairment. This could be cross-coupling, read disturbs, data retention impairments, program disturbs, endurance impairments
- Error Correction and Signal Processing – Advanced ECC (error-correcting code), and introducing the patented (and other patents pending) Memory Signal Processing (TM) to improve the reliability and performance of the NAND Flash storage as show in the diagram below:
In a nutshell, Anobit’s new and innovative approach will result in
- More reliable MLCs
- Better performing MLCs
- Cheaper NAND Flash technology
This will indeed extend the NAND Flash technology into greater innovation of flash storage technology in the near future. Whatever Apple will do with Anobit’s technology is anybody’s guess but one thing is certain. It’s going to propel Apple into newer heights.