Afternoon everyone.
One of the trends that we have noticed in some threads on the forums and other communication methods is that some customers believe we are not investing enough in the platforms. We believe that that this does not reflect reality, but the fact that people think this to be true is our own fault for not keeping you all informed of our plans. In this posting I am hoping to give you an insight into our spending plans, what we want to spend money on, and why. It is important to note that as you will all be aware, we work in a very fast paced industry, time behaves differently in the Broadband industry, as our Chief Operating Officer often reminds us, we operate in mouse years! This means that these plans are of course subject to change, but I hope they give you the clear view that we do plan for the future and are constantly trying to provide the customers with the best platform that we can. Also, some of these plans are more fleshed out than others, the further they are away in time, then the less detail we have. So, without further ado...
Capital Expenditure plans for 2006 / 2007 Over this year we have planned for a total CAPEX spend of over £3 million, of which these are the highlights of items yet to be implemented:
MXCores - £120,000. The MXCores are responsible for the delivery of mail to customer's inboxes.
The current MXCore platform consists of 24 1U Intel based servers, and was designed to scale sideways, i.e we could just throw more kit at the platform as the traffic increased. This design has proved very effective, and was perfect in its day, however 24 servers already take a lot of environmental resource, as well as a lot of management overhead. To scale the platform sideways further is only going to add to this burden. So, to reduce the resource required to run the platform, we are planning to significantly change the architecture.
The plan is to replace the 24 Intel boxes with 4 clusters of 4 x Sun T1000 servers and 1 Sun 3510 Fibre Channel Disk Array. There will be 2 clusters per site providing true site resilience and the option to scale sideways again.
Core Database - £100,000. The Core Database contains all the information related to the customer base and is the heart of the business in many ways.
The technology used in the Core database platform is due to be refreshed to reflect the increased performance and storage requirements we have with an increased user base. As with the MXCore platform, we also want to take advantage of the multi-core technology that powers modern servers, which reduces our rack, power and cooling requirements. In light of this, the plan is to replace the six Sun V440s and associated various storage back ends we currently use, with 5 x Sun T2000 servers and 5 fully populated Sun 3220s for storage.
Core Network - £262,000 We define the Core Network as the equipment in London that takes traffic from the BT/LLU networks, and passes it on to PlusNet's own services, or the Internet, either over Peering or one of our Transit providers. We currently have a switched 10 Gigabit Ethernet layer 2 core in London consisting of Foundry Big Iron Switches and a routed Gigabit Ethernet Layer 3 core consisting of Juniper M10i's. These devices are reaching the edge of their performance and scalability limits so need to be replaced with more powerful equipment. Again we are taking the opportunity provided by needing to scale the platform to change the architecture. We are looking to both simplify and reduce the environmental footprint requirements of the Core Network.
This project's aim then is to collapse these two types of device into a single device and to remove the layer 2 switched core architecture. To do this we will be implementing some new Cisco 7609's as a replacement for both the Juniper routers and the Foundry switches. As with most of our projects, we prefer the phased approach. So this work is being spread over several months, with the first phase already completed.
Customer Termination & Traffic Management - £135,000 Customer Termination equipment is defined as the equipment that customer's DSL connections are terminated on, and of course, as everyone knows we use Ellacoyas for our traffic management.
To scale the Broadband platform we have already placed the Purchase Order with both Juniper and Ellacoya for some new kit. However we have not purchased a new ERX1440, but instead have made the move to the new E320 which is Juniper's next generation of Broadband Termination device. This improves our environmental requirements and gives greater performance for the money.
Portal Servers - £12,000 As part of the long term plan for the Portal platform, we have initially scoped two lots of budget for the portals. This first allocation of cash is really just to sideways scale the platform to stay in line with the increasing number of customers using the portals. So all we are doing here is buying some additional 1U servers with 4Gig of Memory, 2 x 3Ghz processors, hardware raid and 2 x 73GB SCSI disks. I will talk about the next phase of this work later.
Parbin Refresh - £22,000 Most of the Parbin brands (Port 995, Just The Name etc) that we purchased along with Metronet still reside in the original Redbus Co-Lo facility in London and although we have made some additions to the platforms running the services, i.e. additional memory, we need to completely replace the hardware to bring them in line with our current hardware standards and move the equipment into our main co-lo facilities in London and Sheffield as appropriate.
Out of Band Management Network - £50,000 Out-of-band management refers to system console access provided, even in the event of primary network subsystem (hard and/or software) failure.
We currently do not have a consistent separate network over which to control and manage our servers and network devices. This project is to build an Out Of Band Management Network at every Point of Presence, be it in Sheffield or London. This will entail implementing a pair of Cisco 3825 Integrated Service Routers into each PoP to provide security and serial access to devices and ISDN routers to which we will attach ISDN lines for access from remote locations as well as all the internal LAN infrastructure.
Improved Load Balancing - £20,000
We are planning on implementing Global Server Load Balancing across our Data Centers so that if we have a platform failure in the Primary site, the traffic will automatically fail over to the stand-by.
Data Centre Resilience Improvements - £45,000 The aim here is to ensure that no back-end system is a single point of failure and provide a truly resilient platform.
Component Add Refresh - £6,000This is a piece of development work to improve the functionality of our Component Add system and improve the servers it runs on. This system is used every time you add a component to your account and has become slower to run as we've added additional functionality. This work will speed up its operation and allow us to remove any legacy code.
CBC/VMBU CBC stands for Capacity Based Charging, and is the current system we use to bill our Pay As You Go customers.
This is a piece of development work to allow us to decommission the CBC servers and move the functionality to VMBU. Currently PAYG customers are billed based on RADIUS data, which is processed by the CBC servers. Once this work has been completed all customers will be billed/managed by the VMBU system. There will be some prerequisites fixes to VMBU to ensure the validity of the data.
Authoritative DNS This is a piece of development work to ensure that our current Authoritative DNS servers that contain all of our customer domains are able to scale and handle the increasing number of domains that we host. This includes work to make sure that the servers are able to restart in a timely manner whilst loading up the domain list.
Monitoring - £5k This work is to provide a dashboard view internally and also provide resilience of this view. Currently we only have the very detailed technical view of what's going on with the Infrastructure that can be difficult to read and see quickly what the health of the network is.
New Building We are moving from our current two buildings into a brand new office in the Centre of Sheffield, this is the link to the new developers website
http://velocity-space.com. As part of this we are fully reviewing and replacing our internal infrastructure to give the best possible secure access to everyone in the office.
Mail Storage - £170k This is the new Sun NAS system that has been purchased to provide additional capacity for customer e-mail storage to cope with the growth in traffic.
Scale of the platform in line with increased customer numbers There is an ongoing task to constantly upgrade or improve the current platform to cope with the increases in customer numbers and volumes of data/traffic being passed or processed.
CGI Back-UpsThere is a plan being formulated to improve the current back-up infrastructure for the customer CGI platform. This is to ensure that the infrastructure can handle the increasing customer traffic on the platform.
Capital Expenditure plans for 2007
The total expenditure amounts to more than £3 million again, of which the highlights are as follows:
Customer Termination and Traffic Management - £520,000
Core Network - £151,000
Data Centre Improvements - £21,000
New Portal Server Hardware - £50,000
This is what I mentioned earlier, and follows the same idea as the MXCore changes we are making in 2006. Basically the idea is to reduce footprint, while improving performance.
The attached GANT Chart shows the current view of the order of work.
I hope you have found this information interesting and goes someway to showing that we do have plans for investing in the platforms. As we get close to the date of implementation, I will post individual threads so that everyone knows what our plans are and has a chance to comment and question.
Best regards
Simon
** Modded by IW to add some bold tags.
