 |
|
Outsourcing Network Connectivity Process
iPuzzlebiz Reports:
Phase 1: The data center is currently connected AT&T via an OC-3 connection as well as to AT&T Canada also with an OC-3 connection.
Phase 2: All aspects of the network will be redundant. The core switches will be Cisco switches with Gigabit Ethernet connectivity to the database and fileservers and 100Base-T Fast Ethernet to the webservers.
Backups
Phase 1: Incremental backups of any changed files are performed every night and full backups of all machines are performed at least once a week. Tapes are rotated every week and kept in an offsite Media Fire-proof safe. A Dell server utilizing Veritas’ Net Backup Business edition software handles the backups. For Oracle Database backups the Veritas add-on for Oracle is utilized to provide hot-backups of all databases.
Phase 2: The ability to restore any item of the infrastructure in the event of a failure is key so Veritas Netbackup will be utilized to safeguard the infrastructure. Nightly incremental backups will be conducted with full backups occurring once a week. The backups will be written to an ADIC Scalar 100 tape library and tapes will be rotated weekly to a secure offsite location.
Monitoring
Phase 1: Various software is currently used to monitor the status of the network and the our infrastructure. Some of these tools inlclude: MRTG (Multi Router Traffic Grapher) to store bandwidth statistics, MON a tool for tracking network connectivity and the built in service checks of the F5 labs load balancers. We have also developed our own service checks that test the availability of the database and fileserver through PHP and PERL.
Phase 2: We implement HP’s Openview Network Node Manager product for monitoring all aspects of the network and infrastructure. This will proactively notify the Network Operations staff of any problems or failure on the equipment.
Security
For security we will be implementing the Cisco Secure PIX firewall running in an active/standby fail over configuration. The firewall system will protect the infrastructure from any unauthorized connections as well as the prevention of Denial of Service attacks.
| |
 |
Phase one & two will concentrate on providing optimum performance, availability and reliability in a single location on the West coast. Phase three will involve mirroring the infrastructure on the East coast providing disaster recovery and optimum path routing based on geographic region.
Phase Three Scaling
The core infrastructure for iPuzzlebiz is designed not only for the highest performance and availability but also scalability. With scalability in mind when our customer base increases we will not have to be performing fork-lift upgrades rather we can build in additional capacity with our existing equipment.
The initial system will be sized with a one terabyte RAID (Redundant Array of Inexpensive Disks) array of disks. An array of this size is more than adequate for our initial rollout. If additional space is required as we go forward we can scale in this cabinet to 15 terabytes simply by adding more disks.
If we grow beyond that size we can add another drive array into the system and expand from there. Since we will be utilizing Fibre Channel technology to interconnect all of the drive equipment we can extend the connectivity up to several kilometers if necessary, meaning we can locate the drive array in the vicinity of the initial infrastructure without problem.
The webserver capacity issue can be overcome simply by adding additional servers to the infrastructure. Scalability is not really an issue as far as the servers themselves but rather will depend on the fileserver and the database machines capacity.
The database cluster provides us with a great deal of flexibility for a number of reasons. Our initial configuration will consist of some Sun Enterprise servers.
Each Server chassis can scale to 14 processors and 14 gigabytes of RAM. This will increase our capacity a great deal simply by adding processing power to our existing cluster. As our needs grow even further we can bring more servers online to share the load up to a maximum of 32 servers in one cluster.
The fileserver cluster will be built relatively the same as the database cluster. The initial configuration will consist of some Sun Enterprise servers.
Each Server chassis can scale to 14 processors and 14 gigabytes of RAM. Again capacity increase will simply mean the addition of processors and RAM.
Storage space can be increased by the addition of disks to the Hitachi Data Systems drive array. The fileserver cluster will utilize Veritas’ Cluster Server Fileserver edition to provide maximum performance and reliability.
The network will scale easily with the rest of the infrastructure since it is designed around the idea of expandability and modularity.
The two core switches both Cisco switches will allow almost any combination of interface cards, anything from 8 port Gigabit Ethernet cards to 48 port 100BaseT Ethernet cards.
Each switch supports 9 interface cards and initially the switches will only require two cards each for a total of 16 Gigabit Ethernet ports per switch.
As the server infrastructure grows we will add more Cisco 3548XL switches for aggregation of the webserver infrastructure. These switches will then uplink via gigabit Ethernet to the 6509 switches.
Additionally as the load increases on the load balancers we can add another set of balancers to provide additional capacity into the network.
- Infrastructure Estimates at 1,000,000 customers
The following is only an estimate of needed resources to support 1,000,000 customers. Obviously it is difficult to project the requirements on a strictly theoretical basis so this should serve only as a guide.
- Webservers: Approximately (100) Webservers. An additional (4) servers for Domain Name Service
- Fileservers: (2) Sun Enterprise 4500 Servers with 14 Processors and 14 GB of RAM a piece
- Database: (4) Sun Enterprise 4500 Servers with 14 Processors and 14 GB of RAM a piece.
- Network: (10) Cisco Catalyst 3548XL switches
- (2) Cisco 6509 Catalyst Switches with (4) 8-Port Gigabit Ethernet Cards and (1) 24-Port 100BaseT card per switch
- (4) BIG/IP HA+ load balancers
- (1) Cisco 7576 Router with (4) 1 Port Fast Ethernet Cards and (3) OC-3 Packet over Sonet cards
- Storage: (1) Hitache Data System 7700E Drive array for a total of 3.4 Terabytes of disk storage.
- Datacenter: To support the 1,000,000 customer system we will need space for approximately 8 Cabinets or a 16ft x 7ft area.
- Bandwidth: 100Megabits/Sec of bandwidth.
Disaster Recovery
For phases one and two of implementation we have offsite backups of all infrastructure in case of a catastrophic event. Further redundancy is provided by the availability of spare equipment.
Phase three of implementation will include a second site with a mirrored copy of site A.
This will be accomplished by using Hitachi Data Systems’ Remote Data Facility software to update each site with any content changes. This will ensure that in the event either site is destroyed the remaining site will be able to serve communities with limited downtime.
Data Center
iPuzzlebiz utilizes the Class A data center facilities of Fibercloud. Fibercloud utilizes various forms of security including programmed key entry, retinal scanning, and closed circuit video surveillance as well as state-of-the art fire detection and suppression systems.
To ensure maximum uptime, FiberCloud has designed the center to include multiple fiber connections coming into and going out of the facility through multiple providers, dual HVAC units, backup UPS and generator and redundant routing equipment configurations.
