Skip to main content

The buzz and fuss on SSD


With the focus on faster IOPS and lower power consuming devices, there has been a lot of interest and buzz around Solid State Drives (SSD). We explore some of the key talking points around SSD over here.

For a high level understanding perspective, most (not all) SSD’s key characteristics are :
•          Semiconductor (NAND flash, non-volatile)
•          No mechanical read/write interface, no rotating parts eliminating seek time or rotational delays
•          Electronically erasable medium
•          Random access storage

Through the various publications, there is a general consensus that the flash memory-based SSDs provide lower read/access latencies, higher read bandwidths, and minimal/negligible seek overheads.
 


Some of the key gains which make SSD a popular choice for column oriented database in NoSQL generation are :
“(1) SSDs scale up linearly with concurrent execution of database queries and outperform disks by up to a factor of two, 
(2) the low seek cost on SSDs makes column storage a better choice for laying out data on a variety of flash devices,”  (source : http://pages.cs.wisc.edu/~swift/papers/fdb-tr.pdf )


From various experiments, it has become obvious that SSD can help reduce computations and I/Os in MapReduce significantly. The observed advantage is significant in sort map and sort reduce phases up to a tune of 30-40% on a average. The advantage observed are primarily due to faster SSD throughput, shorter IOWait while making more RAM available for processing. SSDs are best suited for cache-unfriendly data with high access densities (IOps/GB) requiring low response times.

Typical use case of SSD deployment include ATM, Online Banking, ATM, Currency Trading, Point-of-Sale Transaction and data mining. Further from a green footprint, SSD offer an additional benefit of lower energy consumption, cooling and space requirements


However, flash memory architecture has also been criticized for “wear” (loss of charge due to isolation defects) and low performance during “mix” operation involving both read and write. There has also been a concern on high cost per GB for SSD versus HDD. However, the cost gap is fast reducing with market forces. With the concerns on enterprise adoption of SSD still a small question mark, there has been a tendency to move to a hybrid architecture involving both HDD and SSD with SSD taking the role of frontend or controller to HDD.  There have also been initiatives for Tiered storage where data centers utilize different types of storage throughout the storage infrastructure. For instance, SSD in Tier 0, FC/SAS in Tier 1, SATA in Tier 2 layers. There is a key trend on moving to SSD for use cases where read optimized database are deployed and primary use is to scan, select and fetch data. 
Over the years, SSD has made a key market presence and is expected to grow up in maturity and price point to offer a valid option to enterprise for much more storage options depending upon application architecture.

Comments

Popular posts from this blog

Low latency SQL querying on HBase

HBase has emerged as one of the most popular NoSQL database offering distributed, versioned, non-relational tables hosted on commodity hardware. However, with a large set of users coming from a relational SQL world, it made sense to bring the SQL back in this NoSQL. With Apache Phoenix, database professionals get a convenient way to query HBase through SQL in a fast and efficient manner. Continuing our discussion with James Taylor, the founder of Apache Phoenix, we focus on the functional aspects of Phoenix in this second part of interaction.
Although Apache Phoenix started off with distinct low latency advantage, have the other options like Hive/Impala (integrated with HBase) caught up in terms of performance?
No, these other tools such as Hive and Impala have not invested in improving performance against HBase data, so if anything, Phoenix's advantage has only gotten bigger as our performance improves.  See this link for comparison of Apache Phoenix with Apache Hive and Cloudera Im…

Data deduplication tactics with HDFS and MapReduce

As the amount of data continues to grow exponentially, there has been increased focus on stored data reduction methods. Data compression, single instance store and data deduplication are among the common techniques employed for stored data reduction.
Deduplication often refers to elimination of redundant subfiles (also known as chunks, blocks, or extents). Unlike compression, data is not changed and eliminates storage capacity for identical data. Data deduplication offers significant advantage in terms of reduction in storage, network bandwidth and promises increased scalability.
From a simplistic use case perspective, we can see application in removing duplicates in Call Detail Record (CDR) for a Telecom carrier. Similarly, we may apply the technique to optimize on network traffic carrying the same data packets.
Some of the common methods for data deduplication in storage architecture include hashing, binary comparison and delta differencing. In this post, we focus on how MapReduce and…

Pricing models for Hadoop products

A look at the various pricing models adopted by the vendors in the Hadoop ecosystem. While the pricing models are evolving in this rapid and dynamic market, listed below are some of the major variations utilized by companies in the sphere.
1) Per Node:Among the most common model, the node based pricing mechanism utilizes customized rules for determining pricing per node. This may be as straight forward as pricing per name node and data node or could have complex variants of pricing based on number of core processors utilized by the nodes in the cluster or per user license in case of applications.
2) Per TB:The data based pricing mechanism charges customer for license cost per TB of data. This model usually accounts non replicated data for computation of cost.
3) Subscription Support cost only:In this model, the vendor prefers to give away software for free but charges the customer for subscription support on a specified number of nodes. The support timings and level of support further …