News
 

Bookmark and Share

(17) 

Mosaid Technologies this week introduced HLSSD (HyperLink solid-state drive), a SSD that achieves terabyte-class storage capacity using a single controller device. The solid-state drive utilizes Mosaid's proprietary HLNAND technology. The developer hopes that its HLSSDs will become cost-efficient high-performance storage solutions for enterprises.

Mosaid’s prototype HLSSD utilizes a single FPGA-based Serial ATA III controller to deliver 2TB capacity over four HLNAND channels, with 16 32GB HLNAND flash MCPs (multi-chip packages) per channel. In comparison, a conventional 2TB SSD requires four SATA controllers, plus a Serial ATA III hub controller, operating over 32 channels.

HLNAND flash is a proprietary high-performance solution that combines Mosaid's own HyperLink memory technology with industry standard NAND flash cell technology to deliver advanced feature set, reaching sustained I/O input-output bandwidths more than ten times higher than conventional flash, according to Mosaid.

Mosaid’s terabyte-class HLSSDs are optimized for data centers and the enterprise. Based on HLNAND (HyperLink NAND) technology, Mosaid’s HLSSDs deliver scalability, performance, reliability and cost advantages over SSDs that use conventionally designed NAND flash devices.

Today’s data centers are designed with up to four tiers of storage devices: lower cost, lower performance tape drives and Serial ATA HDDs, and higher cost, higher performance SAS HDDs and SSDs. This architecture necessitates advanced intelligence in the storage arrays to determine which data to allocate to the faster SAS HDDs and SSDs.

With Mosaid’s terabyte-class HLSSDs, only two storage tiers are needed: Serial ATA HDDs and HLSSDs. As a result, the requirement for advanced intelligence is reduced, as is the total costs of ownership, including power, footprint, cooling, maintenance and other elements, according to the company.

“Our affordable and scalable terabyte-class HLSSDs will be an enabling technology in data center and enterprise applications. HLSSDs can be used to build large, fast storage pools that lower the total cost of ownership by reducing the number of storage tiers, and minimizing the need for advanced intelligence in the storage arrays,” said Jin-Ki Kim, vice president of research and development at Mosaid.

In mid-2013, Mosaid plans to release a 2TB/4TB/8TB production HLSSD that will use a purpose-built HLNAND flash controller in 40nm process technology.

Tags: SSD, HLNAND, NAND, HLSSD, Mosaid, Flash

Discussion

Comments currently: 17
Discussion started: 08/03/12 09:31:00 AM
Latest comment: 03/28/14 06:22:31 AM
Expand all threads | Collapse all threads

[1-4]

1. 
8TB in mid 2013? It sounds like SSDs will finally exceed HDDs in terms of available space. If only they were as cheap as HDDs...
4 0 [Posted by: knedle  | Date: 08/03/12 09:31:00 AM]
Reply
- collapse thread

 
SSD based on NAND Flash memory does not hold data forever, so I do not think an 8 TB SSD is a good investment for your data when thinking long term. An 8 TB SSD will be good for a cache, but who needs 8 TB for cache. HDD is better for holding data for long term even though it has moving parts. The platters of HDD can hold data a lot longer than SSD with NAND Flash memory.

It is best use SSD with a capacity less than 200 GB because SSD is best for as OS drives and not for media drives.
1 2 [Posted by: tecknurd  | Date: 08/03/12 06:10:08 PM]
Reply
 
show the post
0 3 [Posted by: moderntheorist  | Date: 08/04/12 01:03:56 PM]
Reply
 
Most of the HDD that I have lasts at least a decade. The recent HDDs that I bought are failing within a year although I am still able to access some parts of failed HDD. HDD manufactures are having quality issues lately. However, the data that I have is still intact on multiple HDD from my old computer that I have not touched or have not power it on for 5+ years, so I am not wrong or I am NOT VERY WRONG.

I did not say about wear and tear of SSD. I am stating that oldest data on a SSD can not be more than 5 to 10 years because of the type of non-volatile memory being used which is NAND Flash memory. SSD are good for short term data, but HDD are better for long term.

Actually SSD can fail all of the sudden with out warning. HDD provides warnings that sectors are going, becoming noisier or louder, taking more time to access data. I can tell a HDD is failing, but I can not tell if solid-state components are failing. Digital just goes like snap of finger when they fail while analog parts shows warning signs of failure is soon.
1 2 [Posted by: tecknurd  | Date: 08/05/12 10:56:19 AM]
Reply

2. 
SATA is not a very good bus for flash storage from a performance (both throughput and latency) standpoint--particularly at this size--but it's convenient. Robin Harris of StorageMojo.com has a great set of articles on this:

http://storagemojo.com/20...-based-arrays-a-bad-idea/
http://storagemojo.com/20...-architectural-critiques/
http://storagemojo.com/20...write-cliff-in-real-life/
0 1 [Posted by: bluvg  | Date: 08/03/12 01:37:55 PM]
Reply
- collapse thread

 
SATA is a standard connection. This means if SATA controller works with your desire OS, just about any SATA drive will work. Could use PCIe, but you will need drivers for your desire OS for SSD that hook up to PCIe.

PCIe has latency issues too. More cards hook up to this bus and latency increases. It is best to use less devices on a bus then many. Since all motherboards comes with sound cards, NIC, USB, and graphics the bus latency will be high.

600 megabytes per second for SATA-III is plenty for anybody and I do mean anybody. High definition video does not need that amount of bandwidth for RAW recording. If you care too much bandwidth when all your data is less than 100 megabytes in size, you mess accessing time as the most important spec.

When getting to less than 1 millisecond of response time or accessing time, it is harder to measure. There is cache, buffering, memory latency issues, and bus latency that have to be dealt with. Just think of SSD and HDD as higher level memory cache. When data is access from these devices, they first have to be put in physical memory or RAM before the data is ever read by the CPU.

You can use SSD in an array. The throughput of SSD is already high, very high. RAID-1, or RAID-10, RAID-5, RAID-6 could be use to provide some fail safe.

What Robin Harris is doing is using common sense to tell you about data storage mediums like SSD.
0 0 [Posted by: tecknurd  | Date: 08/03/12 05:52:42 PM]
Reply
 
We're not talking consumer mobos here: "Mosaid’s terabyte-class HLSSDs are optimized for data centers and the enterprise." In that environment, there are lots of applications that push--or would benefit if able to push--well over the throughput of an uncompressed HD video stream.

Mr. Harris's point is much more about the fact that SSDs and SATA are not optimized formats for flash storage. They weren't designed with flash in mind. To put it another way, the performance potential of flash storage is held back when delivered via an SSD. There are other ways to package flash storage that are better matched to its performance characteristics.
0 2 [Posted by: bluvg  | Date: 08/03/12 09:21:53 PM]
Reply
 
SATA is a standard while SSD using PCIe is not. A PCIe based SSD requires drivers to be installed for the OS. Companies only make Windows drivers for those devices. SATA on the other hand is already supported by other OS besides SATA. Top of all, SATA can be in AHCI mode for increase compatibility of the OS that supports it. I prefer going with a standard instead going proprietary.

The only "applications" in the server market that need high throughput is virtual machines and huge databases. Throughput can easily be increase through striping RAID levels.

SSD can use other solid-state non-volatile memory besides just using NAND Flash memory. NAND Flash memory is one of memories that are used in SSD. Harris has no tips what can improve NAND Flash memory performance. He seems more like an amateur that is stating common sense instead of scientific facts. You want to back him up, fine, but SSD companies have been using NAND Flash memory for decades. Every company that makes SSD are researching by themselves and not sharing what improves or penalizes SSD using NAND Flash memory. I suggest take Harris explanations that SSD and SATA is penalizing NAND Flash memory as a grain of salt.

I will say NAND Flash memory is not held back by SATA and SSD. After a few decades of NAND Flash research, NAND Flash memory is being pushed to its limits. It is not a good replacement for HDD because it can not hold data longer than 5 to 10 years depending the quality. It has slow writes because of a certain hardware that is require to write data to NAND Flash memory.

Mosaid's HLSSD is basically SSD with their own chipset just like Sandforce that has their chipset. Other SSD has their own chipset. It is not any different. All these companies pick a standard which is SATA. Some may include PCIe for some of their SSD models, but that requires drivers which is only for Windows.
1 0 [Posted by: tecknurd  | Date: 08/03/12 11:18:48 PM]
Reply
 
It might be worth clarifying terminology. SSD in this case means Solid State Disk, i.e. flash storage delivered in the same format as a traditional HDD container and connector along with a flash controller. PCIe flash storage solutions don't really use SSDs.

I'm not sure what point you're trying to make by saying that the "only 'applications' in the server market that need high throughput is virtual machines and huge databases." Only? VMs are the de facto standard for servers at many organizations these days (outpacing physical servers in 2011), and VDI is also taking off. Huge databases are also critical applications for many businesses. "Big data"--whether a traditional relational database, Hadoop, or whatever--is an exploding market. HPC solutions are also growing significantly.

Where you lose me, though, is when you state that Harris "seems more like an amateur." Wow. That is quite a criticism to make of someone who has been in the storage industry for decades, once worked at Sun, is brought in as a consultant by ultra-high-performance and other storage solution providers such as Violin Memory, Nimble, Parascale, IBRIX, Amplidata, etc., and has one of the most-read storage blogs in the industry. As he wrote, if we didn't have the legacy of HDDs as our storage mechanism of choice for the past few decades, would we use the same approach for solid state storage? Directly from his article (did you read any of it before beginning to criticize it?):

Latency. Low compared to disks, but substantial compared to flash. SAS/SATA stacks were never optimized because disk latency was the big problem.
SSD bandwidth. There are wider options, especially close to the CPU.
Reliability. SSDs replace the head/media assembly in disk drives with NAND chips. The rest of the SSD has all the tender bits of a regular disk – bits that account for about half of all disk failures. Compare DIMM and disk replacement rates.
Cost. SSDs cost 50%-100% more than the raw flash, even after using all the high-volume disk components. Mounting directly on PC boards, like DIMMs or PCIe cards, is much more cost effective.
Flexibility. The good news with SSDs is that they take advantage of the huge tech infrastructure that supports disks. But that’s the bad news too, if an optimized clean-sheet architecture is the goal.

I'm not really sure what you're trying to argue. The points are pretty clear. SSDs are convenient and obviously much faster in most use cases than HDDs, and that's without getting into the various kinds of MLC and SLC. But an optimized approach to implementing flash storage? No, it's not. You seem to be arguing against further performance gains, as though more performance is a problem, regardless whether it's "necessary" for some uses or not. But calling Robin Harris an "amateur"... that is a comment you may want to reconsider.
0 2 [Posted by: bluvg  | Date: 08/04/12 09:51:57 PM]
Reply
 
A consultant or sales or marketing do not know anything about engineering. Anybody can be a consultant by doing some research and give out their own opinions. Engineers have their own opinions when someone does not know that saying and doing is two different things. Making NAND Flash memory to have the lowest latency and high throughput is a lot harder than saying. The physics says that NAND Flash memory is already pushed to its limits. NAND Flash memory is not going to be equal to DRAM performance on terms on latency and throughput. I think the fastest NAND Flash memory chip is 60 ns and the bandwidth is about 10 megabytes per second.

The point I am stating is SSD will not get any lower of 1 millisecond because where they are position in the memory level. HDD and SSD are position in the level 20 memory while physical memory or RAM is position at level 10 memory. CPU has level 1, level 2, and level 3 memory because they are closer. Further or higher level of the memory loses performance.

I arguing because people do not need super high throughput. Also latency for file storage can not be any lower than 1 millisecond. Mr. Harris thinks it can. Again saying is different than doing. Yes, SSD are better than HDD when it comes to quicker access of applications, but not better for long term. Though I prefer to not have a pure SSD, but to have a hybrid because a hybrid provides a place for my data to last years on the HDD portion and SDD for caching the most used data.
0 0 [Posted by: tecknurd  | Date: 08/05/12 11:54:38 AM]
Reply
 
A generic statement saying a consultant in a sales or marketing role knows nothing about engineering is pointless--for one, it's just plain 100% wrong to make as a categorical statement, and more importantly, we're not talking about a trend, we're talking about Robin Harris specifically. If you're going to call him an amateur, consider your credibility lost.

Arguing that people don't need higher performance is like Bill Gates "saying" (he never actually said that--in fact, he stressed the opposite, talking about how much pain the limit caused--but this is one of those statements said often enough that people accept it as true) that "no one will ever need more than 640 KB" RAM. Why would anyone argue *against* lower latency? Our entire computing paradigm for decades has been based on layers of workarounds to deal with the latency of spinning disk. Why insist, in effect, that we *need* it to be this way? It's nonsensical. There are vendors showing products delivering latencies well below 1 ms. The industry responses (since you question Robin's credentials, these were coming from the vendors themselves) *confirmed* Robin's assessment that SSDs offered convenience and flexibility for the customer, but did NOT offer the highest performance flash can offer.

The point is simply that the SSD form factor is a product of decades of legacy spinning disk. It's not an optimized architecture for flash, it's a capitulation to legacy convenience. This is not in dispute, even by the vendors whose products are based SSDs. The only person disputing it is you. It's somewhat like going from a horse to a car, then insisting on putting a governor on the car so that it can't go faster than a horse because "people do not need" to go faster.
0 0 [Posted by: bluvg  | Date: 08/06/12 12:57:14 AM]
Reply

3. 
show the post
0 5 [Posted by: tks  | Date: 08/03/12 09:11:51 PM]
Reply

4. 
I'm curious about one thing that I honestly haven't done much research into and am hoping one of readers here can shed some light on. Maybe it's a bit off-topic, but seeing other arguments in previous posts deviating off-topic quite substantially, could someone please tell me what would be the options of data retrieval in case of hardware failure on SSDs? I'm well aware how it's done with HDDs and, considering most of failures with HDDs would be limited to mechanical parts or burnt PCBs (with non-mobile parts at least), data retrieval would be "relatively easy" providing a clean room and availability of spare, programmable parts isn't an issue. My question is then - what would be the most common failure types on SSDs and how easy (if at all possible) would be retrieving the data off them? Are we, in general, talking about simply replacing the circuitry (PCB), or more frequently of lower level and harder to reverse/irreversible failures (like blown flash chips, etc.)? Thanks!
1 2 [Posted by: MyK  | Date: 08/05/12 05:21:14 AM]
Reply
- collapse thread

 
* A cell can wear out and no way of constructing from a journal, an image, parity, or ECC, data can not be re-constructed.

* The file is older than 5 or 10 years (depending the quality of the NAND Flash memory used), data can not be re-constructed.

* If during the write process of NAND Flash memory and power goes out, SSD can fail to work upon next power up. This is the probably the most worry-some.

* A defective NAND Flash memory chip could work for a year and then stop working. Can lose several gigabytes to hundreds of gigabytes.

Basically there is no data retrieval for SSD. You just pay the price for no data retrieval when trying to increase performance. If want to pay it safe, go with hybrid drive, Intel's Smart Response (Windows only), or Highpoint's RocketHybrid (Windows only).
1 0 [Posted by: tecknurd  | Date: 08/05/12 12:43:49 PM]
Reply

[1-4]

Add your Comment




Related news

Latest News

Tuesday, July 22, 2014

10:40 pm | ARM Preps Second-Generation “Artemis” and “Maya” 64-Bit ARMv8-A Offerings. ARM Readies 64-Bit Cores for Non-Traditional Applications

7:38 pm | AMD Vows to Introduce 20nm Products Next Year. AMD’s 20nm APUs, GPUs and Embedded Chips to Arrive in 2015

4:08 am | Microsoft to Unify All Windows Operating Systems for Client PCs. One Windows OS will Power PCs, Tablets and Smartphones

Monday, July 21, 2014

10:32 pm | PQI Debuts Flash Drive with Lightning and USB Connectors. PQI Offers Easy Way to Boost iPhone or iPad Storage

10:08 pm | Japan Display Begins to Mass Produce IPS-NEO Displays. JDI Begins to Mass Produce Rival for AMOLED Panels

12:56 pm | Microsoft to Fire 18,000 Employees to Boost Efficiency. Microsoft to Perform Massive Job Cut Ever Following Acquisition of Nokia