From markd at ripe.net Thu Apr 12 15:39:09 2007 From: markd at ripe.net (Mark Dranse) Date: Thu, 12 Apr 2007 15:39:09 +0200 Subject: [tt-tf] TTM futures - new draft Message-ID: <461E367D.9070508@ripe.net> Dear Task Force, Based on feedback previously provided, below you will find the new outline of our plans for future direction of the TTM network. We appreciate you taking the time to read this draft proposal, and are very eager to receive any and all feedback you may have on the general plan, or any specific element of it. We expect this discussion phase to be completed by RIPE 54 in May, and therefore would appreciate you feedback by 27th April. Vision ~~~~~~ Our goal is to expand and maintain the TTM network, increasing the value of the service to the community at large and to the owners of the probes. This will be achieved by making improvements in the following areas: - Enhancing the measurement architecture - Enhancing the TTM network architecture - Improving alarms and reporting - Improving the business model These changes will require a radical rethink in probe and measurement architecture, resulting in new structures for the probes themselves, and the tests they carry out. TTM network infrastructure ~~~~~~~~~~~~~~~~~~~~~~~~~~ The TTM network is the foundation of the TTM and associated services. Therefore, these services strongly depend on reliability and coverage of the network. In particular it is essential that there are many active probes at interesting locations, that probes are well maintained, and that coverage can be further increased by deploying simpler devices. Let's consider these goals separately. 1. For the services that the TTM provides, it is essential that there are enough active probes at interesting locations. At present, we are relatively limited by the availability, and geography,of those willing to sponsor probes, and as a result, some rather denseand unhelpful clustering has occurred. A possible solution to mitigate this problem is to install a number of probes at important locations for free. The NCC will fund the hardware and service contract, while the host will support the probe with power, connectivity and remote-hands. Rough criteria for hosting a box: - Major AS or IX - Only one free box per AS ? if a site wants more, they have to pay - Commitment from the site to operate the box for a minimum 3 year period - Selection will be a ?beauty contest?, and at the discretion of the NCC 2. However this alone will not ensure sustained growth of the network. This needs to be addressed by increasing the number of sponsored probes and ensuring that those probes are well maintained and upgraded when necessary. Therefore It is important to make the TTM services attractive to sponsors of the TTM probes to increase stability and steady growth of the network. Several incentives can be provided: - Allow the sponsor to define their own measurement mesh and specific routine measurements (level of detail and frequency). Management of gathered data will be the responsibility of the sponsor. - Support the sponsor in creating dynamic meshes to monitor their own applications and services, providing a global picture of availability. 3. To further expand the network in the areas where support and technical expertise are limited (e.g. small/residential customers), a probe itself should be as lightweight as possible (no GPS, embedded systems). The CBM proposal fits well in this category. This setup doesn't allow one-way measurements, although an aggregated view from a cluster of such lightweight probes can enhance the TTM and add-on services. TTM Measurements ~~~~~~~~~~~~~~~~ 1. Near real-time measurement of application services The TTM network will continue to act as a measurement platform for critical Internet public service and infrastracture applications (e.g. root and ccTLD DNS monitoring and of multicast performance). These independent tests are of benefit to the service administrators, the end users, and to the hosts of TTM probes. We will improve this service by - Considering measuring new applications as the need arises - Making reporting clearer (see below) - Enhancing alarm functionality (see below) - Implementing the CBM proposal (see infrastructure section) 2. Global one-way measurement platform The TTM network will continue its one-way measurements of the delay/jitter/loss between the distributed probes. These independent tests are of benefit to the hosts of the TTM probes, to the broader community, and to specific interested organisations who rely on neutral views (e.g. regulators and governments). We will improve this service by - Introducing custom measurement meshes (see infrastructure section) - Introducing support for OWAMP. This will allow the TTM network to interact with other compatible networks, and will facilitate measurements on demand (subject to access restrictions) - Setting up data quality checks to ensure the sanity of presented data - Making reporting clearer (see below) - Enhancing alarm functionality (see below) 3. General purpose distributed measurement platform for ad-hoc experiments The TTM network will be developed to enable interested parties (e.g. CAIDA) to conduct a variety of time-limited, global, local, or mid-range experiments (for example the k-root anycast studies). An outcome will typically be an analytical report that will be open to the community. Reporting ~~~~~~~~~ We consider the current reports to be complex and overwhelming. We will simplify the presentation of results to include fewer pre-generated plots, and shift our focus to the identification of trends and changes in gathered data. We will continue to support tools to generate plots on demand, and raw data will continue to be available for analysis. DQM will be used to verify data sanity. In addition to this, we are investigating ways to merge TTM data with that from other services - such as RIS - combining data from these powerful platforms to display broader and more representative pictures of Internet behaviour in near real-time. Alarms ~~~~~~ We consider the current alarm functionality to be complex and overwhelming. This will be re-engineered to provide more powerful network and application alarms. We will enable users to run a variety of NCC defined tests, limited by number, frequency and longevity, so as to ensure that these tests do not impact public TTM network operation. As with reporting, we will simplify presentation of generated alarms to make them clearer. Admin and pricing model ~~~~~~~~~~~~~~~~~~~~~~~ As a result of feedback from existing TTM hosts, we intend to move to a single up front payment. This payment will cover hardware, setup, and three years of service fees. After three years, the host may re-subscribe, or opt to stop participating. -- Mark Dranse Information Services Manager RIPE NCC From markd at ripe.net Mon Apr 23 15:33:52 2007 From: markd at ripe.net (Mark Dranse) Date: Mon, 23 Apr 2007 15:33:52 +0200 Subject: [tt-tf] TTM futures - new draft In-Reply-To: <46274AEF.6040900@heanet.ie> References: <461E367D.9070508@ripe.net> <46274AEF.6040900@heanet.ie> Message-ID: <462CB5C0.80904@ripe.net> Hi Brian, Thanks for your kind comments. I'll attempt to address each of your points in turn. On 19/04/2007 12:56, Brian Nisbet wrote: >> Vision >> ~~~~~~ >> >> Our goal is to expand and maintain the TTM network, increasing the value >> of the service to the community at large and to the owners of the >> probes. This will be achieved by making improvements in the following >> areas: >> - Enhancing the measurement architecture >> - Enhancing the TTM network architecture >> - Improving alarms and reporting >> - Improving the business model > > Are there any notions of time-scales along with these proposals? At present, this is just a draft. We need consensus from the community, which I hope will come from these on-list discussions, and from the WG meeting at Tallinn. After that, we can take the plan to the Board for executive approval. Once approved, we will begin the real work on this, which will lead to a proper project plan with milestones and timescales. That's the political answer. Back in the real world, I'd like it if most of this could be done over the coming year. But please don't take that as a firm commitment. >> TTM network infrastructure >> ~~~~~~~~~~~~~~~~~~~~~~~~~~ >> 1. For the services that the TTM provides, it is essential that there >> are enough active probes at interesting locations. At present, we are >> relatively limited by the availability, and geography,of those willing >> to sponsor probes, and as a result, some rather dense and unhelpful >> clustering has occurred. A possible solution to mitigate this problem is >> to install a number of probes at important locations for free. The NCC >> will fund the hardware and service contract, while the host will support >> the probe with power, connectivity and remote-hands. Rough criteria for >> hosting a box: >> >> - Major AS or IX >> - Only one free box per AS ? if a site wants more, they have to pay >> - Commitment from the site to operate the box for a minimum 3 year period >> - Selection will be a ?beauty contest?, and at the discretion of the NCC > > This is vital and will, hopefully, attract a decent amount of > respondents. I hope so too. Does anyone else have any comments regarding this? >> 2. However this alone will not ensure sustained growth of the network. >> This needs to be addressed by increasing the number of sponsored probes >> and ensuring that those probes are well maintained and upgraded when >> necessary. Therefore It is important to make the TTM services attractive >> to sponsors of the TTM probes to increase stability and steady growth of >> the network. Several incentives can be provided: >> >> - Allow the sponsor to define their own measurement mesh and specific >> routine measurements (level of detail and frequency). Management of >> gathered data will be the responsibility of the sponsor. >> - Support the sponsor in creating dynamic meshes to monitor their own >> applications and services, providing a global picture of availability. > > While I think that it will be a pity (as I have voiced previously) > to lose the fully-meshed nature of the TTM network, I would not > for one moment suggest that this step is not necessary, and it is > one that will hopefully present a clearer business case for TTM. By way of clarity, we're not proposing losing the full mesh. This is a fundamental and integral element of the architecture, and fully meshed measurements will continue for the foreseeable future (the only argument I can see for them stopping is if the sheer volume of measurements becomes too high as a result of enormous network growth - a nice problem to have at some point in the future!) What is being proposed is the addition of support for locally customised meshes so that probe sponsors can carry out measurements of specific local interest to them. It's very much an augmentation, not a replacement of current functionality. >> TTM Measurements >> ~~~~~~~~~~~~~~~~ >> >> 1. Near real-time measurement of application services >> The TTM network will continue to act as a measurement platform for >> critical Internet public service and infrastracture applications (e.g. >> root and ccTLD DNS monitoring and of multicast performance). These >> independent tests are of benefit to the service administrators, the end >> users, and to the hosts of TTM probes. We will improve this service by > > The near real-time measurements are of great importance to us in > HEAnet and I am very pleased to see they are not in danger. > >> Reporting >> ~~~~~~~~~ >> >> We consider the current reports to be complex and overwhelming. We will >> simplify the presentation of results to include fewer pre-generated >> plots, and shift our focus to the identification of trends and changes >> in gathered data. We will continue to support tools to generate plots on >> demand, and raw data will continue to be available for analysis. DQM >> will be used to verify data sanity. In addition to this, we are >> investigating ways to merge TTM data with that from other services - >> such as RIS - combining data from these powerful platforms to display >> broader and more representative pictures of Internet behaviour in near >> real-time. > > This really does seem to be the best of all possible worlds. > >> Admin and pricing model >> ~~~~~~~~~~~~~~~~~~~~~~~ >> >> As a result of feedback from existing TTM hosts, we intend to move to a >> single up front payment. This payment will cover hardware, setup, and >> three years of service fees. After three years, the host may >> re-subscribe, or opt to stop participating. > > This makes a lot of sense. It's possible that this is a question for > the infrastructure section, but I would very much like to hear more > detail (although perhaps not appropriate to this discussion) on the > hardware plans for the full TTM boxes. Some of the hardware out there > is getting old at this point and might it be useful to have an upgrade > path available for those who have been part of the project for some > time? I'm not really asking on behalf of HEAnet here, more wondering > if continued participation by commercial entities could be ensured > by sweetening such a deal? You raise an interesting and valid point. Hopefully I'll have an answer to this by the time we all meet in Tallinn. > Once again, thanks for all the work that obviously went into this > proposal. And thank you for your comments and queries. I'm interested in seeing some more traffic on this topic prior to RIPE54, so I encourage anyone else who reads these emails to share their thoughts. If you're part of the silent majority, please try to share just a few comments with us - be they in support, or (perhaps more importantly) otherwise. Mark -- Mark Dranse RIPE Network Coordination Center Information Services Manager Singel 258 Amsterdam NL http://www.ripe.net +31 20 535 4444