You are on page 1of 4

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245346373

A new revolution in enterprise storage architecture

Article  in  IEEE Potentials · November 2009


DOI: 10.1109/MPOT.2009.934894

CITATIONS READS
12 147

1 author:

David Leong
Republic Polytechnic
4 PUBLICATIONS   18 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

EPSILON (Ɛ): End-to-end Protection for Smart-metering Communication Infrastructure within a Local Field Area Network View project

All content following this page was uploaded by David Leong on 13 February 2018.

The user has requested enhancement of the downloaded file.


A New Revolution in Enterprise Storage Architecture

David Leong
Republic Polytechnic
9 Woodlands Avenue 9
Singapore 738964
{David_Leong@rp.sg}
IEEE Member No: 41631787
Member IEEE, Member IEEE GOLD

Abstract
NAS and SAN are the common enterprise storage architectures that are commonly deployed in
data centers. However as the storage demands escalate over time, none of these storage networks
support aggressive concurrency and per-server throughput requirements for the applications on
highly scalable computing clusters. In recent times, efforts had been set to incorporate the
strengths of existing storage architectures into a unique framework that accomplished the per-
formance, scalability, manageability and high level of data security. The approach had leads to
the design of novel storage architecture that support next generation object-based storage tech-
nology.

Some of the critical concerns in network storage today are high costs and the complexity in man-
aging data and storage devices. Gartner and IDC studies had revealed that the purchasing price of
storage systems comprises not more than 10% of the total cost for enterprise storage. Factors such
as downtime, backup and recovery procedures, as well as administration costs of the storage sys-
tems had made up greater part of the total cost of ownership. The problem is further expedited by
the exponential growth of data and massive information lifecycle that entails more storage devic-
es and data to be managed. For simplification of data management and enhancing data sharing,
enterprises desire that their large database files could be available to their global customers across
multiple platforms simultaneously. In view of these, IT industry is often driven to look for more
effective solutions to information management.
In recent times, DAS, NAS and SAN are the common enterprise storage architectures that were
commonly deployed. As the usage of distributed systems and large-scale clusters of commodity
computers increased, significant research has been designated towards the designing of distribut-
ed and highly scalable storage systems. However, all these existing storage architectures have
severe limitations when clusters become larger. One of the main drawbacks of cluster file systems
is as cluster nodes, shared-storage capacity and the I/O workload escalates, it tends to create se-
vere performance bottleneck.
Neither NAS nor SAN supports aggressive concurrency and per-server throughput requirements
for the applications on highly scalable computing clusters. Recently, researchers have set their
efforts to incorporate the strengths of existing storage architectures into a unique framework, so
that an ideal storage architecture that combines performance, scalability, manageability and high
level of data security can be accomplished. This novel approach leads to object-based storage ar-
chitecture.
One of the main design features in object-based storage architecture is the separation of metadata
from the data management. This is to avoid the metadata server (MDS) from becoming the per-
formance bottleneck, as seen in NAS architecture. MDS does not contain any user data. It is es-

1
sentially used as a global resource to locate objects, mitigating secure access to objects, and assist
objects management. Data are transferred directly between the application servers (also referred
as clients) and the object-based storage devices (OSD) during its read/write operations without
any intervention from MDS. A MDS cluster allows enterprise storages to scale to a very large
finite number of objects (possibly 1030) that can be managed simultaneously, and at the same time
sustaining their access performance. In addition, object-based storage architecture allows applica-
tion servers to access its storage devices in parallel to cultivate high performance data transfers.
Object-based storage architecture deals with storage as objects, rather than as files or blocks of
data. An object is a combination of data and the attributes that describe its context. While classi-
cal storage devices manage independent data blocks, it does not know the type of applications
that utilizing it. As storage capacity grows exponentially in time, the tasks to manage the loca-
tions of the files and associated data blocks would grow tremendously as well. For instant, a regu-
lar 300 GB disk drive contains 600 million of data blocks, increasing the capacity to 500 TB
would required 1 trillion blocks of data to be managed. Managing these huge amounts of data
blocks is a great burden to servers. In contrast, the object-based storage architecture is capable of
off-loading the management of these individual objects to OSDs, effectively shifting that respon-
sibility away from the server’s file system.

Application MDS Cluster


Servers

Storage OSD
Network Clusters

Control Paths
OSD Data Paths

Figure 1: An object-based storage architecture comprising multiple OSD clusters, a


metadata server (MDS) cluster and several application servers. Multiple OSDs may
be coupled together using high-speed network switches to create delegated enterprise
storage clusters in a structured peer-to-peer (P2P) network environment.

The combination of object attributes and data allows storage controller to comprehend some of
the relationships between the data blocks on the storage device. The attributes could provide
some hints about the object’s characteristics and behaviors. OSD could use this vital information
to implement intelligent, low-level optimizations in the storage devices that includes self man-
agement, per-object security, quality of service (QoS) level agreements, and data-aware caching
to improve storage performance. Object-based storage architecture can meet the challenge of en-
terprise storage demands by providing high availability, high flexibility and better manageability
where traditional DAS, NAS, and SAN-based storages have fell short.

2
While many major storage players such as IBM, EMC, HP, Sun Microsystems, NetApp and
Seagate Technology have show interests or engaged some form of development works in the ob-
ject-based storage, Panasas bundled the parallel storage solutions with its own commodity-based
hardware. In order to meet the demands of massively parallel data movement occur in high per-
formance computing environments, Panasas offers ActiveStor 3000 and ActiveStor 5000 unified
clustered storages designed for data centers via an object-based approach. The storage solutions
are enhanced by a matured third generation ActiveScale object-based file system. The Panasas
storage system also encompasses a number of key technologies for the health monitoring and
self-healing of its data objects and storage devices.
With the helps from several research groups and universities, the standard for OSD-1 command
set to control the operations of the object-based storage devices was ratified by ANSI in 2004.
The Storage Network Industry Association (SNIA) OSD Technical Work Group is currently
working on the standardization of the command set extensions for OSD-2.

References
[1] A. Raghuveer, S.W. Schlosser and S. Iren, Enabling database-aware storage with OSD, Pro-
ceedings of the 24th IEEE Conference on Mass Storage Systems and Technologies (MSST07), pp.
129-142, Sept. 2007.
[2] A. Devulapalli, D. Dalessandro, P. Wyckoff, N Ali and P. Sadayappan, Integrating parallel
file systems with object-based storage devices, Proceedings of the 2007 ACM/IEEE Conference
on Supercomputing (SC2007), Article No. 27, 2007.
[3] L. Qin and D. Feng, Active Storage Framework for Object-based Storage Devices, Proceed-
ings of the 20th International Conference on Advanced Information Networking and Applications
- Volume 2 (AINA'06) - Volume 02, pp. 97-101, 2006.
[4] J.C. Wu and S.A. Brandt, QoS support in Object-Based Storage Devices, Proceeding of the
3rd International Workshop on Storage Network Architecture and Parallel I/Os (SNAPI05),
pp.41-48, Sept. 2005.
[5] D. Nagle, D. Serenyi and A. Matthews, The Panasas ActiveScale Storage Cluster – Delivering
Scalable High Bandwidth Storage, Proceedings of the 2004 AGM/IEEE Conference on Super-
computing (SC2004), pg. 53, Nov. 2004.
[6] M. Mesnier, G.R. Ganger and E. Riedel, Object-Based Storage, IEEE Communications Maga-
zine, 41(8), pp. 84-90, Aug. 2003.

Biography

David Leong is an academic staff with the School of InfoComm Technol-


ogy at Republic Polytechnic, Singapore. He facilitates operating system
concepts, data communications and digital forensic to tertiary students.
David’s professional interests include mass storage technology and net-
work computing.

View publication stats

You might also like