|
Peer-to-Peer
Data warehousing at RCOM
Reliance Communications has deployed a data warehouse to
expedite the process of loading data and fulfilling queries from legal and regulatory
agencies, writes Nikita Upadhyay
 Facing
the predicament of managing massive amounts of customer data, responding to
ad hoc queries from legal authorities and loading data, Reliance Communications
(RCOM) has deployed a Greenplum Database, running on a Sun Data Warehouse Appliance
to power a range of applications ranging from legal and regulatory compliance
to detailed record analysis.
Greenplum is a data warehouse software and services company and its goal is
to reinvent database software for the purpose of business intelligence. It collaborated
with Sun Microsystems in 2007.
Business challenges
Rising demand for Reliance’s services was resulting in explosive growth
of the company’s systems and infrastructure. The need to provide accurate,
timely analytics to all parts of the business was undeniable. One area of the
company that required rapid delivery of relevant analysis was Reliance’s
Law Enforcement Department.
The Indian government requires telcos to produce detailed phone records for
specific subscribers to improve national security and reduce crime. Such queries
can pop up at any point of time and the concerned service provider has to revert
to these agencies within 24 hours. Call Detail Records (CDRs) for all subscribers
of RCOM are stored as required by regulatory authorities and local governments.
Reliance’s rapid growth, coupled with the inability of traditional database
systems to scale and perform, was reducing the company’s ability to deliver
on these requests. Responding to a request for records was taking several days.
Even loading a day’s worth of data into the system was taking about two
to three hours. To maintain compliance and deliver results, Reliance needed
a breakthrough.
With conventional methods coming a cropper, the company chose a Greenplum Database
and Sun Data Warehouse Appliance to solve its problems by speeding up queries.
As Reliance continues to thrive with new services and subscribers, the company
increasingly looks to Greenplum and the Sun Data Warehouse Appliance to solve
its toughest data problems.
Making the switch
|
 "The
deployment delivered such value that in a short span of time Reliance
decided to expand the system. This was done to include another important
data set to add CDRs from interconnect traffic"
- Srivalsan Ponnachath
Director-Solution Sales, Sun Microsystems India
|
Reliance deployed this technology during the first quarter
of 2008. Offered as a Sun Customer Ready designed system, the server, storage,
and database were integrated at the factory and RCOM got a system ready to run
on its arrival at the company’s data center. Using the Sun Customer Ready
Program, RCOM reduced deployment from two weeks and multiple resources, to a
single day installation requiring a single system administrator. The project
was delivered, just four months after initiation.
The solution utilizes the massive parallel processing capability
of the Greenplum Database (about 20 TB of data is combed through to respond
to queries) combined with the Sun Data Server, to offer near-zero latency access
to data. This has resulted in faster retrieval while responding to inquiries,
with RCOM delivering canned requests for call detail records within seven to
ten minutes in the existing dataset, compared with an hour previously. The parallel
loading enables data to be added to the Greenplum database at a faster load
rate, resulting in three to five times faster data loading performance (one
to two hours).
RCOM has also recovered a significant amount of space and saved money on that
account by deploying the Sun OpenStorage technology, with six rows of data center
rack space saved and power and cooling reduced significantly, lowering overall
energy consumption for a smaller and therefore better ecological footprint.
The solution
The company chose the S1004 model of the Sun Data Warehouse Appliance which
integrates a Greenplum Database with Sun OpenStorage Sun Fire X4500 servers,
and Sun Fire X4200 servers to provide a fast, efficient scalable system with
significantly reduced density, cooling and power requirements.
The combination of Sun server technology, including the Sun Fire X4200 and Sun
Fire X4500 servers and the Greenplum database, has resulted in an extremely
powerful appliance. All systems in the solution run the Solaris 10 Operating
System and its file system, ZFS.
The systems provide 20 CPU cores driving 192 SATA disk drives and 40 terabytes
of usable database capacity.
| Company |
Reliance Communications |
| Solution |
Greenplum and the Sun Data Warehouse Appliance |
| Business challenges |
- RCOM’s explosive growth in systems
and infrastructure.
- Ability to scale the existing database
while maintaining performance.
- Achieve compliance with government regulations.
|
| Timeline |
Q1 2008 |
| Aim of the implementation
|
RCOM deployed a high performance data
warehouse system to improve performance in loading and querying of the CDR
Datamart, and mitigate the risk of non-compliance with legal and regulatory
agencies. |
| Benefits |
- Achieved three to five times faster performance
for data loading and ad hoc data query responses in context of RCOM
needs in CDR database.
- Helped in RCOM's goal of compliance with
regulatory Service Level Agreements.
- Reduced data center rack space and lowered
overall energy consumption.
- Saved up to two weeks of implementation
through Sun Customer Ready Program.
|
Cognizable benefits
RCOM generates about 160 GB of CDR data per day and loading this CDR data from
the production environment into a datamart takes anything from six to eight
hours. Complex query requests on a datamart are fulfilled in one to two hours.
To speed information retrieval and remain compliant, many requests were fulfilled
by queries, directly on the production environment rather than on the datamart.
“This has resulted in faster retrieval times, near zero-latency for inquiries,
with RCOM delivering canned requests for call detail records within seven to
ten minutes in the existing dataset, compared with one hour previously,”
explained Pankaj Shah, Decision Support Systems, Reliance Communications. The
Ad hoc Query and data load is three to five times faster than in the previous
system. It is more compact, with significant savings in storage space and energy
consumption, while providing high performance.
After the deployment of the S1004, the time required to revert to a request
for detailed call records shrank by over 80%, from multiple days to a few hours.
Compared to Reliance’s previous database system, the Greenplum system reduced
the average time to load a day’s worth of data by over 90%, from two hours
to less than 10 minutes. Deploying the Greenplum solution improved response
times and mitigated Reliance’s risk of non-compliance.
System expansion
The deployment delivered such value that, in a short span of time, Reliance
decided to expand the system. “This was done to include another important
data set to add CDRs from interconnect traffic,” informed Srivalsan Ponnachath,
Director-Solution Sales, Sun Microsystems India.
These records represent all of the calls that navigate the Reliance network
and one or more networks to link initiating subscribers to receiving subscribers.
Interconnect CDRs are used to monitor SLAs and to perform billing among co-existent
Telco carriers on the broader network.
In an industry marked by rapid growth and new market entrants, the scale and
complexity of Interconnect data had become too much for traditional database
technology that Reliance had used in the past. Taking advantage of Greenplum’s
ability to scale horizontally, Reliance simply added two X4500 systems to the
existing S1004, to construct what is effectively an S1006 appliance. Today,
the system supports 60 terabytes of usable capacity. Reliance’s two primary
data sets represent over 28 terabytes of actual data, growing at a pace of 1
terabyte per week.
nikita.upadhyay@expressindia.com
|
